Clinical Trajectories, Healthcare Resource Use, and Costs of Diabetic Nephropathy Among Patients with Type 2 Diabetes: A Latent Class Analysis

Clinical Trajectories, Healthcare Resource Use, and Costs of Diabetic Nephropathy Among Patients... Diabetes Ther (2018) 9:1021–1036 https://doi.org/10.1007/s13300-018-0410-8 ORIGINAL RESEARCH Clinical Trajectories, Healthcare Resource Use, and Costs of Diabetic Nephropathy Among Patients with Type 2 Diabetes: A Latent Class Analysis . . . . Ruixuan Jiang Ernest Law Zhou Zhou Hongbo Yang Eric Q. Wu Raafat Seifeldin Received: January 31, 2018 / Published online: March 29, 2018 The Author(s) 2018 diagnosis of T2DM and C 2 urine albumin tests ABSTRACT within the Truven MarketScan database (2004–2014), based on T2DM-related compli- Introduction: Patients with type 2 diabetes cations, comorbidities, and therapies. DN mellitus (T2DM) are clinically heterogeneous in severity categories (normoalbuminuria, moder- terms of disease severity, treatment, and ately increased albuminuria, and severely comorbidities, potentially resulting in differen- increased albuminuria) were determined based tial diabetic nephropathy (DN) progression on urine albumin measure. The risks of DN courses. In this exploratory study we used latent progression and reversal (change to a more/less class analysis (LCA) to identify patient groups severe DN category) were compared among all with distinct clinical profiles of T2DM severity identified latent classes using Kaplan–Meier and explored the association between disease analyses and log-rank tests. All-cause and DN- severity, DN progression or reversal, and related costs and HRU were assessed and com- healthcare resource use (HRU) and costs. pared during the study period among the Methods: Latent class analysis was used to identified latent classes. group adults with C 2 medical claims with a Results: Four clinically distinct profiles were identified among the 23,235 eligible patients: Enhanced content To view enhanced content for this low comorbidity/low treatment (46.5%), low article go to https://doi.org/10.6084/m9.figshare. comorbidity/high treatment (29.0%), moderate comorbidity/high insulin use (9.7%), and high Electronic supplementary material The online comorbidity/moderate treatment (14.8%). The version of this article (https://doi.org/10.1007/s13300- 018-0410-8) contains supplementary material, which is 5-year DN progression rates for these clinically available to authorized users. distinct profiles were 11.8, 18, 16.5, and 27.7%, respectively. Compared with the low comor- R. Jiang  E. Law (&) bidity/low treatment group, all other groups Department of Pharmacy Systems, Outcomes, and were associated with an increased risk of DN Policy, College of Pharmacy, University of Illinois at progression (all p \ 0.001). Increasing comor- Chicago, Chicago, IL, USA e-mail: elaw3@uic.edu bidity was significantly associated with higher all-cause and DN-related HRU and costs, pri- Z. Zhou  H. Yang  E. Q. Wu marily driven by higher pharmacy and inpa- Analysis Group, Inc., Boston, MA, USA tient costs. R. Seifeldin Conclusion: Patients with T2DM who have Formerly of Takeda Development Center Americas, more comorbidities experienced higher rates of Inc., Deerfield, IL, USA 1022 Diabetes Ther (2018) 9:1021–1036 DN progression and HRU and incurred higher Prospective Diabetes Study (UKPDS) examined healthcare costs compared with patients with the progression of DN among 5100 T2DM low comorbidity profiles. Future prospective patients followed for over a decade [12]. The studies are needed to confirm the significance of UKPDS reported the rates and time to DN pro- these groups on DN progression, HRU, and gression among patients with normoalbumin- costs. uria, moderately increased albuminuria, or Funding: Takeda Development Center Americ- severely increased albuminuria, and found that as, Inc. the more severe the albuminuria, the shorter the time to progression to the next stage of DN [12]. In addition, in T2DM patients, long dura- Keywords: Clinical outcomes; Costs; Diabetic tion and poor control have been associated with nephropathy; Healthcare resource use; Type 2 DN progression [13], while the treatment of diabetes metabolic syndrome has been shown to be independently associated with lesser progres- INTRODUCTION sion of DN in T2DM [14]. A prospective study of Japanese patients with T2DM examined the frequency of and reasons for DN reversal and Diabetic nephropathy (DN), often referred as diabetic kidney disease, is a common and seri- found that DN duration, DN treatment type, glycated hemoglobin (HbA1c) levels, and heal- ous complication of type 2 diabetes mellitus (T2DM), and presents as albuminuria and/or thy blood pressure were linked to likelihood of reversal [15]. The relationship between factors decreased glomerular filtration rate (GFR) [1–3]. In addition to being a leading cause of end-stage related to treatment type and comorbidity cannot be assumed to lie on a continuum, but renal disease (ESRD), DN greatly increases the risk of all-cause and cardiovascular-related rather may co-vary in different ways in different patients, resulting in distinct clinical patterns. mortality, cardiovascular disease and coronary atherosclerosis, and kidney failure among These patterns, or their relationship to mean- ingful outcomes, have not been explored. patients with diabetes [4, 5]. DN affects up to Understanding the relationships between 40% of patients with T2DM [6] and had an estimated prevalence of 3.3% (6.9 million DN progression and T2DM comorbidity and severity, as well as which patients are at partic- adults) within the USA during 2005–2008 [7]. The overall prevalence of DN among U.S. adults ular risk of progression or poor outcomes, may help improve patient-centered disease manage- with diabetes did not change significantly from 1988 to 2014 [8]. The disease progression of DN ment. The use of real-world patient data from claims databases to characterize T2DM could is classified by the following stages based on urine albumin levels: normoalbuminuria contribute to this goal, as real-world patients have been shown to be more heterogeneous in (\ 30 mg/24 h or an albumin/creatinine ratio terms of both disease severity and diabetic [ACR] of \ 30 lg/mg), moderately increased treatments received [16], and claims data are a albuminuria (30–300 mg/24 h or an ACR of large and accessible resource. Latent class anal- 30–300 lg/mg), and severely increased albu- minuria ([ 300 mg/24 h or an ACR [ 300 lg/ ysis (LCA), a statistical method for grouping individuals who share common characteristics mg) [9]. Moderately increased albuminuria pre- cedes severely increased albuminuria, and and thereby allowing distinct ‘‘clusters’’ to be identified [17], can integrate the limited disease without treatment, the GFR progressively declines and can ultimately result in ESRD [10]. information available in claims databases. LCA has been previously used to examine hetero- However, the severity of T2DM is clinically heterogeneous among affected patients, which geneity in patient populations, to identify sub- groups at high potential risk of disability or may also result in differential courses of DN adverse effects, or to predict patients who may progression [11]. Several prior studies have identified risk fac- benefit from particular interventions in various disease areas [18–21]. tors for DN progression. The United Kingdom Diabetes Ther (2018) 9:1021–1036 1023 In the study reported here, we used LCA to with continuous enrollment for C 12 months integrate multiple indicators of T2DM severity prior to and 6 months following the test date; that are readily available in commonly used and (3) without ESRD during the 12 months claims databases to identify patient groups with prior to the test date (Fig. 1). The index date was distinct clinical profiles. Differences in the randomly selected among all eligible urine clinical trajectory of DN (i.e., DN progression albumin test dates. and reversal) were then assessed among the identified latent classes. Additionally, as there is Study Measures and Outcomes limited information regarding the healthcare economic burden among patient groups with Type 2 diabetes mellitus severity indicators (di- distinct clinical profiles of T2DM, we assessed abetic comorbidities, complications, and treat- the health resource utilization and healthcare ments) included in the LCA were assessed costs among the patient groups identified with during the 12-month period prior to the index LCA and identified the patient subgroup with date. T2DM-related complications and comor- the highest economic burden. bidities included retinopathy, neuropathy, ischemic heart disease, cerebrovascular disease, chronic heart failure, hypertension, chronic METHODS kidney disease (CKD)-related symptoms (hy- perkalemia, high parathyroid hormone level, Data Source and high phosphorus level), and metabolic disorder (lipid disorders and other metabolic This study used data from the Truven Health disorders) (see Appendix A in the Electronic Analytics MarketScan Commercial and Medi- Supplementary Material [ESM]). Diabetic treat- care Supplemental and Lab databases (1 January ments included insulin, metformin, sulphony- 2003 to 31 December 2014), which represent lureas, dipeptidyl peptidase-4 inhibitor, approximately 25 million U.S. employees, glucagon-like peptide-1-based therapy, and dependents, and Medicare-eligible retirees cov- other antidiabetic agents (amylin analogs, ered by over 130 health plans and self-insured amino acid derivatives, meglitinide analogs, employers. A subset of the covered lives (ap- aldose reductase inhibitors, alpha-glucosidase prox. 1 million) have recorded laboratory tests inhibitors, dopamine receptor agonists, insulin (mainly ordered in office-based practice) in the sensitizing agents and antidiabetic combina- MarketScan Lab Database. Data were de-iden- tions) (see ESM Appendix B). tified and comply with the patient confiden- tiality requirements of the U.S. Health Baseline Characteristics Insurance Portability and Accountability Act; Baseline characteristics (demographics, disease thus, no institutional review was required. characteristics, and DN-related treatments) were assessed during the 12-month period prior to Study Population and Study Cohorts the index date. Patient demographic informa- tion collected included age, sex, and type of The study sample included adult patients (aged health insurance. The time from the first C 18 years) with C 2 distinct medical claims observed T2DM diagnosis to the index date and with a diagnosis of T2DM (International Clas- patients’ DN severity at the index date (nor- sification of Diseases, 9th revision-Clinical moalbuminuria, moderately increased albu- Modification [ICD-9-CM]: 250.x0, 250.x2) and minuria, or severely increased albuminuria) with C 2 urine albumin test results after the first were assessed. Normoalbuminuria was defined observed T2DM diagnosis. Patients with at least as urinary albumin excretion of\30 mg/24 h or one eligible urine albumin test who met the an ACR of \ 30 lg/mg; moderately increased following criteria were identified: (1) with C 1 albuminuria was defined as excretion of additional follow-up urine albumin test; (2) 30–300 mg/24 h or an ACR of 30–300 lg/mg; 1024 Diabetes Ther (2018) 9:1021–1036 Fig. 1 Sample selection. ESRD End-stage renal disease, T2DM type 2 diabetes mellitus Progression or Reversal of DN severely increased albuminuria was defined as excretion of[ 300 mg/24 h or an ACR[ 300 lg/ The time to DN progression and time to DN reversal were assessed from the index date until mg. In addition, for each patient the Charlson Comorbidity Index (CCI) score [22] and the use inpatient mortality, the end of continuous eli- of DN-related treatments during the 12-month gibility, or the end of data availability, which- period before the index date were recorded. DN- ever came first. DN severity was classified into related treatments included angiotensin con- four disease stages ranging from normal to most verting enzyme inhibitors, angiotensin receptor severe: normoalbuminuria, moderately blockers, diuretics, calcium channel blockers, increased albuminuria, severely increased albu- minuria, and presence of ESRD/dialysis/renal and other antihypertensive agents [23] (see ESM Appendix B). transplantation procedure (see ESM Appendix C). DN disease progression was defined as the Diabetes Ther (2018) 9:1021–1036 1025 presence of a urine albumin test, diagnosis or Information Criterion and Akaike Information procedure indicating a more severe disease stage Criterion) and interpretability of groups. Once than the index disease stage, while DN disease the model was selected, subjects were assigned reversal was defined as the presence of a urine to the latent class based on their probability of albumin test indicating a less severe disease being in that class. Analyses were conducted in stage than the index disease stage. DN disease SAS version 9.4 software using the PROC LCA reversal was only assessed in patients with procedure [25]. moderately increased albuminuria and severely For baseline characteristics, means and increased albuminuria at the index date. standard deviations (SD) were reported for continuous characteristics; frequencies and percentages were reported for categorical char- Healthcare Resource Use and Costs acteristics. Characteristics were compared Economic outcomes were assessed at the per- between the selected reference latent class ver- patient-per-year (PPPY) level from the index sus other latent classes using Wilcoxon rank- date until 2 years from the index date, the end sum tests for continuous variables and Chi- of continuous eligibility, end of data availabil- square tests for categorical variables. Time to ity, or inpatient mortality, whichever came first. disease progression and time to disease reversal Healthcare resource use (HRU) information was were evaluated for each latent class using collected for all-cause and DN-related medical Kaplan–Meier methods and compared between visits, including inpatient, emergency room the reference latent class versus other classes (ER), outpatient, and other medical visits. DN- using log-rank tests. The incidence rates of each related HRU was defined as medical services type of healthcare visit PPPY and the annual associated with a diagnosis of DN or kidney healthcare costs were described for each latent disease, or a procedure for dialysis/hemodialysis class and compared between the reference or renal transplantation. latent class versus other classes using Wilcoxon All-cause and nephropathy-related health- rank-sum tests. A p value of B 0.05 was consid- care costs were calculated from a U.S. payer’s ered to be statistically significant. perspective and inflated to 2016 U.S. dollars All procedures followed were in accordance using the annual medical care component of with the ethical standards of the responsible the Consumer Price Index [24]. Cost compo- committee on human experimentation (insti- nents included medical costs (inpatient, ER, tutional and national) and with the Helsinki outpatient, and other medical services costs) Declaration of 1964, as revised in 2013. Only and pharmacy costs. DN-related costs were de-identified data was used in this study, thus defined as costs associated with a diagnosis code no institutional review was required. This article of diabetic nephropathy or costs associated with is based on previously conducted studies and a procedure for dialysis/hemodialysis or renal does not contain any studies with human par- transplantation. ticipants or animals performed by any of the authors. Statistical Analysis Latent class analysis was used to identify groups RESULTS of patients with clinically distinct T2DM sever- ity profiles. Class membership was determined Latent Class Analysis based on the T2DM disease severity and treat- ment indicators, and individual patients could A total of 23,235 patients with T2DM fulfilled belong to only one group. Models with a vary- all study criteria and were included in the ing number of classes were estimated, and the analysis, including 18,409 patients with nor- best-fitting model was chosen. Model selection moalbuminuria, 3863 with moderately was based on the consideration of several cri- increased albuminuria, and 963 with severely teria, including model fit statistics (Bayesian increased albuminuria (Fig. 1). In the LCA, a 1026 Diabetes Ther (2018) 9:1021–1036 four-class model yielded the best fit, and four low comorbidity/low treatment group (all clinically distinct T2DM patient profiles were p \ 0.05) (Table 2). Compared with the low identified based on distributions of complica- comorbidity/low treatment group, all other tions/comorbidities (e.g., microvascular and groups had significantly longer mean time since cardiovascular disease, CKD-related symptoms, first observed T2DM diagnosis (low comorbid- and metabolic disorder) and use of diabetic ity/low treatment: 32.9 [SD 25.1] months; low treatment (e.g., insulin, metformin, etc.) comorbidity/high treatment: 42.0 [26.8] (Table 1). The four clinically distinct profiles months; moderate comorbidity/high insulin were defined as: (1) Latent Class 1 (46.5% of the use: 42.6 [26.9] months; high comorbid- sample [N = 10,812]; reference latent class), low ity/moderate treatment: 46.6 [29.6] months; all comorbidity/low treatment; (2) Latent Class 2 p \ 0.01). Patients in the high comorbid- (29.0% [N = 6728]), low comorbidity/high ity/moderate treatment group had the highest treatment; (3) Latent Class 3 (9.7% [N = 2255]), mean CCI (mean 2.9 [SD 1.6]), followed by the moderate comorbidity and high insulin use; (4) moderate comorbidity/high insulin group (1.7 Latent Class 4 (14.8% [N = 3440]), high [1.1]), the low comorbidity/high treatment comorbidity and moderate treatment. group (1.6 [1.0]), and the low comorbidity/low Retinopathy was more common among Latent treatment group (1.4 [1.0]). All other groups had Classes 3 (26.8%) and 4 (24.4%) compared with significantly higher mean CCI compared with Latent Classes 1 (2.6%) and 2 (13.3%), and the low comorbidity/low treatment group (all Latent Class 4 had a higher incidence of all p \ 0.05). other comorbidities in comparison with the other three latent classes. Time to DN Disease Progression Baseline Demographics and Disease The median years of follow-up for DN progres- Characteristics sion among the low comorbidity/low treat- ment, low comorbidity/high treatment, The baseline demographics and disease charac- moderate comorbidity/high insulin use, and teristics of the four latent classes are described high comorbidity/moderate treatment groups in Table 2; multiple significant differences were 2.1, 1.9, 2.1, and 1.6 years, respectively. existed between the characteristics of the latent The respective 1-, 3-, and 5-year DN progression classes. For example, patients with moderate rates were 9.3, 20.8, and 27.7% for the high comorbidity/high insulin use were younger comorbidity/moderate treatment group; 5.9, (mean [SD] 49.1 [13.3] years) than those in the 13.3, and 16.5% for the moderate comorbidity/ low comorbidity/low treatment group (54.5 high insulin group; 5.6, 13.9, and 18.0% for the [9.2] years), and patients with high comorbid- low comorbidity/high treatment group; and ity/moderate treatment were older (58.2 [8.7] 4.0, 9.6, and 11.8% for the low comorbidity/low years) and more likely to be male than those in treatment group (Fig. 2). Compared to the low the low comorbidity/low treatment group (male comorbidity/low treatment group, all other 57 vs. 54%, respectively; all p \ 0.05). Patients groups were associated with a significantly from the southern USA were over-represented in increased risk of progression to a more severe every class (range 38–44%). Preferred provider stage of DN (all p \ 0.01). The high comorbid- organization was the most common type of ity/moderate treatment group was associated health plan among all classes (range 72–76%). with the highest risk of disease progression The majority of patients in all groups were among all latent groups. classified with normoalbuminuria at the index date (range 68–84%); however, higher propor- Time to DN Disease Reversal tions of patients were classified with moderately increased albuminuria or severely increased The median years of follow-up for DN reversal albuminuria in other groups compared with the among the low comorbidity/low treatment, low Diabetes Ther (2018) 9:1021–1036 1027 Table 1 Item–response probabilities for the four-class model: probability of each patient in a given latent class Item Latent class (%) Latent Class 1 (low Latent Class 2 (low Latent Class 3 Latent Class 4 (high comorbidity/low comorbidity/high (moderate comorbidity/moderate treatment group) treatment group) comorbidity/high treatment group) insulin use group) Comorbidity Microvascular disease Retinopathy 2.6 13.1 26.8 24.4 disease Neuropathy 5.2 13.3 16.4 31.7 disease Cardiovascular disease Ischemic heart 6.8 6.2 4.7 40.7 disease Cerebrovascular 2.4 1.3 2.0 19.0 disease Chronic heart 0.6 0.1 1.6 11.0 failure Hypertension 58.2 64.0 32.8 94.7 CKD-related 0.5 0.9 1.0 5.3 disease Metabolic 64.1 72.1 43.3 92.8 disorder Use of diabetic treatment Metformin 59.8 88.2 18.3 56.3 Sulphonylureas 13.5 48.3 3.5 25.9 Insulin 0.0 21.9 87.7 44.1 DPP4 inhibitor 4.6 19.3 0.7 12.8 GLP1-based 1.5 15.2 3.1 9.5 therapy Other antidiabetic 7.6 35.5 31.3 26.5 agents Item–response probabilities were calculated among all patients, given the probability of each patient being in that latent class CKD Chronic kidney disease, DPP4 dipeptidyl peptidase-4, GLP-1 glucagon-like peptide-1 CKD-related disease included hyperkalemia, high parathyroid hormone level, and high phosphorus level Metabolic disorder included lipid disorders and other metabolic disorders Other antidiabetic agents included amylin analogs, amino acid derivatives, meglitinide analogs, aldose reductase inhibitors, alpha-glucosidase inhibitors, dopamine receptor agonists, insulin sensitizing agents, and antidiabetic combinations 1028 Diabetes Ther (2018) 9:1021–1036 Table 2 Patient baseline demographics and disease characteristics according to latent class Patient baseline demographics Latent Class Latent Latent Latent p value and disease characteristics 1 Class 2 Class 3 Class 4 [2] vs. [3] vs. [4] vs. (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [1] [1] [1] Age at index date (years) 54.5 ± 9.2 54.4 ± 8.6 49.1 ± 13.3 58.2 ± 8.7 \ 0.01 \ 0.001 Male 5795 (54%) 3789 (56%) 1199 (53%) 1965 (57%) \ 0.05 \ 0.001 U.S. region Northeast 2016 (19%) 1196 (18%) 394 (17%) 837 (24%) \ 0.001 North-Central 2826 (26%) 1629 (24%) 722 (32%) 828 (24%) \ 0.05 \ 0.01 \ 0.001 South 4533 (42%) 2976 (44%) 855 (38%) 1367 (40%) \ 0.05 \ 0.01 \ 0.001 West 1434 (13%) 927 (14%) 283 (13%) 408 (12%) \ 0.001 Insurance plan type Preferred provider organization 8094 (75%) 5145 (76%) 1678 (74%) 2483 (72%) \ 0.05 \ 0.001 Non-capitated point-of-service 781 (7%) 538 (8%) 166 (7%) 225 (7%) Exclusive provider organization 501 (5%) 343 (5%) 82 (4%) 228 (7%) \ 0.01 \ 0.001 Comprehensive 1025 (9%) 502 (7%) 198 (9%) 442 (13%) \ 0.05 \ 0.001 Consumer-driven health plan 314 (3%) 141 (2%) 99 (4%) 43 (1%) \ 0.05 \ 0.01 \ 0.001 High-deductible health plan 97 (1%) 59 (1%) 32 (1%) 19 (1%) \ 0.01 \ 0.001 Time from first observed T2DM 32.9 ± 25.1 42.0 ± 26.8 42.6 ± 26.9 46.6 ± 29.6 \ 0.05 \ 0.01 \ 0.001 diagnosis in the database to the index date (months) Diabetic nephropathy disease status at the index date Normal 9083 (84%) 5249 (78%) 1728 (77%) 2349 (68%) \ 0.05 \ 0.01 \ 0.001 Moderately increased 1483 (14%) 1208 (18%) 400 (18%) 772 (22%) \ 0.05 \ 0.01 \ 0.001 albuminuria Severely increased albuminuria 246 (2%) 271 (4%) 127 (6%) 319 (9%) \ 0.05 \ 0.01 \ 0.001 Charlson Comorbidity Index 1.4 ± 1.0 1.6 ± 1.0 1.7 ± 1.1 2.9 ± 1.6 \ 0.05 \ 0.01 \ 0.001 Nephropathy-related treatments 6892 (64%) 5369 (80%) 1352 (60%) 2975 (86%) \ 0.05 \ 0.01 \ 0.001 ACE inhibitor 3045 (28%) 2823 (42%) 782 (35%) 1524 (44%) \ 0.05 \ 0.01 \ 0.001 Diuretic 1646 (15%) 1291 (19%) 327 (15%) 1201 (35%) \ 0.05 \ 0.001 Calcium channel blocker 1339 (12%) 1017 (15%) 225 (10%) 856 (25%) \ 0.05 \ 0.01 \ 0.001 ARB 1208 (11%) 981 (15%) 262 (12%) 704 (20%) \ 0.05 \ 0.001 Diabetes Ther (2018) 9:1021–1036 1029 Table 2 continued Patient baseline demographics Latent Class Latent Latent Latent p value and disease characteristics 1 Class 2 Class 3 Class 4 [2] vs. [3] vs. [4] vs. (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [1] [1] [1] Other antihypertensive agent 2716 (25%) 1978 (29%) 360 (16%) 1108 (32%) \ 0.05 \ 0.01 \ 0.001 Data in table are presented as the mean ± standard deviation (SD) or as an absolute number with the percentage in parenthesis The four latent classes are described in Table 1 and in section ‘‘Latent Class Analysis’’ T2DM Type 2 diabetes mellitus,ACE angiotensin converting enzyme, ARB angiotensin receptor blockers Latent Classes 2–4 were compared to Latent Class 1, respectively Other antihypertensive agents included direct renin inhibitors, antiadrenergic antihypertensives, selective aldosterone receptor antagonists, agents for pheochromocytoma, vasodilators, monoamine oxidase inhibitors, and antihypertensive combinations comorbidity/high treatment, moderate comor- associated with a significantly higher rate of bidity/high insulin use, and high comorbid- disease reversal compared to the low comor- ity/moderate treatment groups were 2.1, 2.0, bidity/low treatment patients (all p \ 0.05). 2.2, and 1.7 years, respectively. The respective 1-, 3-, 5-year DN reversal rates were 7.1, 12.0, All-Cause and DN-Related HRU and 13.9% for the high comorbidity/moderate treatment group; 5.0, 8.9, and 8.9% for the Increasing comorbidity was associated with moderate comorbidity/high insulin group; 6.1, significant increases in the annual frequency of 9.9, and 10.5% for the low comorbidity/high all-cause healthcare visits, with a consistent treatment group; and 4.5, 7.9, and 8.9% for the trend across inpatient, outpatient, ER, and low comorbidity/low treatment group (Fig. 3). other medical services visits (p \ 0.01 in all The low comorbidity/high treatment and high pairwise comparisons vs. low comorbidity/low comorbidity/moderate treatment groups were treatment group) (Table 3). The high Fig. 2 Time to progression to a more severe diabetic nephropathy (DN) disease stage. The respective 1-, 3-, Fig. 3 Time to reversal to a less severe DN disease stage. 5-year DN progression rates were 9.27, 20.79, and 27.70% The respective 1-, 3-, 5-year DN reversal rates were 7.09, for the high comorbidity/moderate treatment group 11.99, and 13.92% for the high comorbidity/moderate (purple); 5.86, 13.26, and 16.46% for the moderate treatment group (purple); 5.00, 8.91, and 8.91% for the comorbidity/high insulin group (green); 5.60, 13.93, and moderate comorbidity/high insulin group (green); 6.11, 17.97% for the low comorbidity/high treatment group 9.90, and 10.54% for the low comorbidity/high treatment (red); and 3.95, 9.54, and 11.78% for the low comorbidity/ group (red); and 4.51, 7.88, and 8.89% for the low low treatment group (blue) comorbidity/low treatment group (blue) 1030 Diabetes Ther (2018) 9:1021–1036 Table 3 All-cause and diabetic nephropathy-related healthcare resource use and healthcare costs Per patient per year HRU and Latent Class 1 Latent Class 2 Latent Class 3 Latent Class 4 p value costs, mean – SD (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [2] vs. [3] vs. [4] vs. [1] [1] [1] All-cause HRU Inpatient admissions 0.09 ± 0.29 0.11 ± 0.36 0.14 ± 0.48 0.26 ± 0.62 \ 0.05 \ 0.01 \ 0.001 Inpatient days 0.44 ± 2.23 0.62 ± 3.18 0.83 ± 3.75 1.71 ± 6.30 \ 0.05 \ 0.01 \ 0.001 Emergency room services 0.39 ± 1.12 0.42 ± 1.04 0.58 ± 2.83 0.80 ± 2.12 \ 0.05 \ 0.01 \ 0.001 Outpatient services 12.96 ± 11.26 13.94 ± 11.66 14.87 ± 13.11 21.07 ± 16.71 \ 0.05 \ 0.01 \ 0.001 Other 1.65 ± 3.56 2.13 ± 3.55 4.22 ± 4.93 4.60 ± 7.57 \ 0.05 \ 0.01 \ 0.001 DN-related HRU Inpatient admissions 0.00 ± 0.06 0.01 ± 0.12 0.02 ± 0.17 0.05 ± 0.28 \ 0.05 \ 0.01 \ 0.001 Inpatient days 0.03 ± 0.55 0.11 ± 1.80 0.15 ± 1.47 0.53 ± 4.35 \ 0.05 \ 0.01 \ 0.001 Emergency room services 0.00 ± 0.07 0.00 ± 0.09 0.01 ± 0.15 0.03 ± 0.25 \ 0.01 \ 0.001 Outpatient services 0.22 ± 1.63 0.35 ± 1.49 0.60 ± 3.49 1.46 ± 8.17 \ 0.05 \ 0.01 \ 0.001 Other 0.01 ± 0.59 0.02 ± 0.49 0.05 ± 0.63 0.20 ± 3.67 \ 0.05 \ 0.01 \ 0.001 All-cause healthcare costs (2016 USD) Total healthcare costs 10,172 ± 19,518 13,836 ± 23,696 17,411 ± 25,149 24,564 ± 36,317 \ 0.05 \ 0.01 \ 0.001 Total medical costs 7,383 ± 18,161 8,709 ± 22,059 11,314 ± 23,555 17,812 ± 34,637 \ 0.05 \ 0.01 \ 0.001 Inpatient admission costs 2300 ± 11,886 3107 ± 16,511 3521 ± 14,939 7616 ± 27,065 \ 0.05 \ 0.01 \ 0.001 Emergency room service costs 645 ± 2760 703 ± 3329 997 ± 9189 1581 ± 6625 \ 0.05 \ 0.01 \ 0.001 Outpatient service costs 3973 ± 9413 4252 ± 8327 4520 ± 9125 6745 ± 10984 \ 0.05 \ 0.01 \ 0.001 Other medical service costs 466 ± 2036 646 ± 2048 2275 ± 4455 1870 ± 5730 \ 0.05 \ 0.01 \ 0.001 Total pharmaceutical costs 2788 ± 4920 5127 ± 5828 6097 ± 6003 6753 ± 8028 \ 0.05 \ 0.01 \ 0.001 DN-related healthcare costs (2016 USD) Total healthcare costs 378 ± 2088 475 ± 2088 700 ± 5015 1687 ± 11,386 \ 0.05 \ 0.001 Total medical costs 113 ± 2017 166 ± 2000 474 ± 4977 1292 ± 11,369 \ 0.05 \ 0.01 \ 0.001 Inpatient admission costs 58 ± 1500 90 ± 1589 299 ± 3800 701 ± 5867 \ 0.05 \ 0.01 \ 0.001 Emergency room service costs 4 ± 127 6 ± 137 9 ± 206 28 ± 326 \ 0.01 \ 0.001 Outpatient service costs 49 ± 1191 68 ± 744 141 ± 1746 458 ± 6318 \ 0.05 \ 0.01 \ 0.001 Other medical service costs 2 ± 90 3 ± 84 25 ± 366 105 ± 2251 \ 0.05 \ 0.01 \ 0.001 Total pharmaceutical costs 265 ± 514 309 ± 530 226 ± 444 395 ± 624 \ 0.05 \ 0.01 \ 0.001 Data in table are presented as the mean ± SD The four latent classes are described in Table 1 DN Diabetic neuropathy, HRU health resource utilization, USD U.S. dollars Latent Classes 2–4 were compared to Latent Class 1, respectively Other medical services included durable medical equipment, home care, skilled nursing facility care, and dental or vision care comorbidity/moderate treatment group experi- had 13.0 (SD 11.3) visits (p \ 0.001). In addi- enced a mean of 0.3 (SD 0.6) inpatient admis- tion, the high comorbidity/moderate treatment sions annually, with a mean length of stay of group had increased all-cause HRU compared 1.7 (SD 6.3) days, while the low comorbidity/ with the low comorbidity/high treatment or low treatment group experienced 0.1 (SD 0.3) moderate comorbidity/high insulin use groups admissions with a mean stay of 0.4 (SD 2.2) days in each HRU category. (p \ 0.001). For outpatient services, the high A similar trend was observed regarding DN- comorbidity/moderate treatment group experi- related HRU; patients with high comorbid- enced a mean of 21.1 (SD 16.7) visits annually, ity/moderate treatment had a notably high while the low comorbidity/low treatment group incidence of nephropathy-related inpatient, ER, Diabetes Ther (2018) 9:1021–1036 1031 outpatient, and other medical services visits in Regarding the components that were the largest comparison with the other groups (all p \ 0.05). contributors to total healthcare costs, a similar For example, the high comorbidity/moderate trend was observed in DN-related costs as with all- treatment group had a mean of 1.5 (SD:8.2) DN- cause costs, with pharmacy costs contributing the related outpatient visits annually, compared to majority of the cost difference between the low 0.22 (SD 1.36) visits for the low comorbidity/ comorbidity/low treatment and low comorbid- low treatment group (p \ 0.001). ity/high treatment groups. Higher inpatient admission costs and outpatient costs drove the increase in DN-related healthcare costs among All-Cause and DN-Related Healthcare patients with increasing comorbidity (i.e., low Costs comorbidity/low treatment group vs. the moder- ate comorbidity/high insulin group and the high Following from the observed trend of higher HRU comorbidity/moderate treatment group). associated with higher comorbidity burden, In this study, we did not adjust for baseline higher comorbidity was also associated with sig- characteristics in the outcome comparisons. nificantly higher all-cause healthcare costs The reason for this decision is that latent (Table 3). The incremental differences in total all- patient groups, as the main effect in the model, cause healthcare costs PPPY between the low were identified by LCA using complications, comorbidity/low treatment group and the low comorbidity, and therapies for T2DM during comorbidity/high treatment, moderate comor- the baseline period. Therefore, these character- bidity/high insulin use, and high comorbid- istics should not be adjusted for when outcomes ity/moderate treatment groups were US$3664, are compared across latent patient groups. In US$7239, and US$14,392, respectively (all addition, baseline demographics are in general p\ 0.05). Pharmaceutical cost differences comparable across groups. (US$2339) accounted for approximately 60% of the all-cause cost differences between the low comorbidity/high treatment and low comorbid- DISCUSSION ity/low treatment groups (the next largest com- ponent was inpatient admission costs [22%; cost A large proportion of patients with diabetes difference US$807]). Similarly, pharmaceutical develop DN, leading to progressive increases in cost differences (US$3309) accounted for approx- albuminuria, declining GFR, and risk of ESRD imately50% of thedifferencebetween themod- [1]. Thus, an important step towards improving erate comorbidity/high insulin use and low the management, treatment, and clinical out- comorbidity/low treatment groups (the next lar- comes of patients with T2DM is to identify gest components were other medical services patients at high risk of DN progression. To the [25%; cost difference US$1809] and inpatient best of our knowledge, this is the first study that admission [17%; US$1221] costs). Conversely, the has utilized LCA to identify subgroups of T2DM majority of the cost differences between high patients based on their diabetes comorbidity comorbidity/moderate treatment and low comor- and treatment profile within a claims database, bidity/low treatment groups was due to differences and that has compared the clinical trajectories in inpatient admission costs (37%; cost difference of DN, HRU, and costs among the identified US$5316), followed by pharmaceutical (28%; subgroups. Patients with larger burdens of US$3965) and outpatient (19%; US$2772) costs. comorbidity had significantly increased risk of The incremental differences in total DN-re- progression to a more severe stage of DN, as well lated costs PPPY between the low comorbidity/ as higher all-cause and DN-related HRU and low treatment group and the low comorbidity/ healthcare costs. In particular, patients with high treatment, moderate comorbidity/high high comorbidity and moderate treatment use insulin use, and high comorbidity/moderate had the highest 5-year DN progression rate, treatment groups were $US97 (p\ 0.05), US$322, HRU, and costs compared to all of the other and US$1309 (p \ 0.001), respectively (Table 3). groups. The current results demonstrate the 1032 Diabetes Ther (2018) 9:1021–1036 feasibility of using LCA to identify clinical sub- therapy [33]. These prior findings may help groups using many aspects of diabetes severity, explain why group of patients in the current and the validity of the identified subgroups was study with moderate comorbidity/high insulin supported by the differential DN progression, use had lower rates of DN progression (5-year resource use, and costs observed across the rate 16.5%) than the group with low comor- subgroups. bidity/high treatment use (5-year rate 18.0%). Previous research has helped establish the A small number of studies have previously association between DN progression and dia- examined the relationship between T2DM betes complications or anti-diabetic treatments. severity and healthcare costs, reporting findings Hypertension, a common complication of dia- similar to the present results. A 2013 systematic betes, plays a major role in the onset and pro- literature review by Banerji et al. synthesized gression of DN, and anti-hypertensive studies reporting the impact of glycemic control treatment can reduce albuminuria and slow the and treatment adherence, which are important progression of DN [26, 27]. Remission of very contributors to DN and other diabetes compli- advanced DN has been observed among type 1 cations, on the healthcare costs of T2DM diabetes mellitus (T1DM) and T2DM patients patients [36]. These authors observed that the undergoing aggressive anti-hypertensive treat- healthcare resource utilization and costs asso- ment [28–31]. In the present study, the high ciated with T2DM management were reduced comorbidity/moderate treatment group had the when glycemic levels and comorbidity were highest prevalence of hypertension (94.7%) and better controlled, although comorbid condi- the highest rates of DN progression (5-year rate tions were still prevalent and anti-diabetic 27.7%) among all groups. The low comorbidity/ medication adherence was largely suboptimal high treatment group had the next highest [37–40]. A retrospective study in 2010 by Men- prevalence of hypertension (64%), as well as the zin et al. reported that the hospitalization rate next highest rate of DN progression (5-year rate and healthcare costs were significantly higher 18.0%). A previous retrospective analysis for T1DM and T2DM patients with poorly con- reported that the concomitant presence of trolled versus well-controlled blood glucose retinopathy was significantly associated with ([ 10 vs. \ 7% HbA1c) [41]. Furthermore, DN progression in patients with T2DM [32], patients with severe complications or comor- perhaps due to similar underlying contributing bidities related to T2DM have been shown to pathology. In the present study, the low incur healthcare costs that are threefold higher comorbidity/low treatment group had a much than those of matched patients without lower prevalence of retinopathy (2.6%) than did comorbidities [42], similar to the present the other groups (range 13.1–26.8%), as well as results. Our results emphasize the magnitude of the lowest rate of DN progression (5-year rate this cost difference and may present opportu- 11.8% vs. [range] 16.5–27.7%). nities to identify patients at risk of higher In addition, several studies have shown that healthcare expenditures. DN progression can be slowed with intensive The results of this study have potential glycemic control [33–35]. The randomized trial implications for the management of T2DM and ACCORD reported that therapy targeting DN as well as for the further application of this HbA1c to levels of \ 6.0 vs. 7.0–7.9% delayed method to identify high-risk patients based on the onset of severely increased albuminuria and claims data. The heterogeneous nature of T2DM some microvascular complications, but complicates clinical management, and the cur- increased risk mortality and other complica- rent results indicate that some patients (e.g., the tions (e.g., hypoglycemia) [34]. A similar, high comorbidity/moderate treatment group) prospective study reported that anti-diabetic may have received suboptimal anti-diabetic therapy targeting HbA1c to levels of \ 6.5% treatment or suboptimal blood pressure control, resulted in delayed onset of DN, retinopathy, which led to a high risk of DN progression. and neuropathy in comparison to conventional However, although patients with the highest Diabetes Ther (2018) 9:1021–1036 1033 comorbidity were at the highest risk of DN Diabetes Complication Severity Index (DCSI), a progression, the 5-year rates of DN reversal 13-point scale for scoring patients’ diagnostic, among both the low comorbidity/high treat- pharmacy, and laboratory data, has also been ment group (10.5%) and the high comorbid- used to assess healthcare costs and comorbidity ity/moderate treatment group (13.9%) were management [45–47]. However, deriving the higher than those of the low comorbidity/low DSCI from claims data is again restrained by the treatment and moderate comorbidity/high availability of laboratory information in claims insulin use groups (both 8.9%). This result databases. In the present study, we derived dif- suggests that anti-diabetic treatments can be ferent indicators of T2DM disease severity from effective in slowing or reversing DN even in the claims database, which is also subject to the patients with more severe T2DM. Additionally, limitations introduced by the data source. First, the LCA results highlight that there may be DN severity may be misclassified due to the use interactions between comorbid conditions, of the results of a single urine albumin test to T2DM treatment, and patient characteristics as define moderately increased and severely there are no clear indicators of group member- increased albuminuria. To maximize availability ship, i.e., group membership as a proxy for data to assess DN progression or reversal and diabetes severity is not clearly defined by num- because the urine albumin test data were lim- ber or type of T2DM medications or number of ited, we did not require a confirmatory urine co-morbidities, as would be the case in tradi- albumin test. Second, patients with T2DM were tional claims database studies. Future research is identified by ICD-9-CM codes in medical ser- warranted to identify any unmet treatment vices claims; diagnoses or procedures related to needs in this patient population, to characterize ESRD, transplantation, or dialysis were also patients with high risk of DN progression, and identified using diagnostic or procedural codes to identify groups with the largest economic in the claims database. However, such diagnoses burden related to T2DM and DN. In addition, and procedures may be over- or under-reported the current study demonstrates the potential for in the data. Similarly, rates of DN progression LCA to be used to identify classes of patients may have been underestimated, as the data do with clinically distinct T2DM profiles using not contain complete information regarding information available in commercial claims laboratory values. Finally, only patients with a data. This is an important contribution, as urine albumin test were included in the study, claims databases represent large and readily which may not capture all patients with DN accessible sources of healthcare information but disease status since alternative laboratory tests, may lack data typically used to indicate prog- such as proteinuria and estimated GFR decline, nosis (e.g., HbA1c). Future research using LCA may also indicate DN [48]. Although the data- could further contribute to understanding and base is geographically representative of the USA, reducing the risk of DN progression and the data include only commercially insured improving clinical outcomes for patients with patients and those who have commercial T2DM. insurance in supplement to their Medicare insurance. Thus, the current results may not be generalizable across different study populations Limitations that may be relevant, such as various socioeco- nomic groups or the uninsured. Estimating T2DM disease severity using data from claims databases can be challenging due to the limited amount and types of clinical infor- CONCLUSIONS mation they contain. Laboratory measures, such as HbA1c levels, are useful indicators of This exploratory study demonstrated the feasi- diabetes severity [43, 44], although claims bility of using LCA to identify patient groups databases contain limited proportions of with distinct clinical profiles of T2DM severity patients with HbA1c lab values available. The and explored the association between T2DM 1034 Diabetes Ther (2018) 9:1021–1036 disease severity, DN progression or reversal, and consulting fees from Takeda Pharmaceuticals. economic outcomes. Increasing levels of At the time of analysis, Raafat Seifeld was an comorbidity were generally associated with employee of Takeda, and held Takeda stock or higher HRU, healthcare costs, and risk of DN stock options. progression, while anti-diabetic treatment Compliance with Ethics Guidelines. All appeared to slow DN progression. procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and ACKNOWLEDGEMENTS national) and with the Helsinki Declaration of 1964, as revised in 2013. Only de-identified data was used in this study, thus no institutional Funding. Sponsorship for this study and review was required. This article is based on article processing charges were funded by previously conducted studies and does not con- Takeda Development Center Americas, Inc. All tain any studies with human participants or authors had full access to all of the data in this animals performed by any of the authors. study and take complete responsibility for the integrity of the data and accuracy of the data Data Availability. The datasets generated analysis. during and/or analyzed during the current study are not publicly available due to confi- Medical Writing and Editorial Assis- dentiality agreement. tance. Editorial assistance in the preparation of this manuscript was provided by Dr. Shelley Open Access. This article is distributed Batts of Analysis Group, Inc. The authors would under the terms of the Creative Commons like to thank Jing Zhao from Analysis Group for Attribution-NonCommercial 4.0 International significant contribution towards medical writ- License (http://creativecommons.org/licenses/ ing and analytical support. The authors would by-nc/4.0/), which permits any non- also like to thank Melvin Munsaka for his sup- commercial use, distribution, and reproduction port in data analysis. This support was funded in any medium, provided you give appropriate by Takeda Development Center Americans, Inc. credit to the original author(s) and the source, provide a link to the Creative Commons license, Authorship. All named authors meet the and indicate if changes were made. International Committee of Medical Journal Editors (ICMJE) criteria for authorship for this manuscript, take responsibility for the integrity REFERENCES of the work as a whole, and have given final approval to the version to be published. 1. Tuttle KR, Bakris GL, Bilous RW, et al. Diabetic kidney disease: a report from an ADA consensus Disclosures. Ruixuan Jiang was supported conference. Diabetes Care. 2014;37:2864–83. by the University of Illinois/Takeda Health 2. Chen J. Diabetic nephropathy: scope of the prob- Economics and Outcomes Research Fellowship. lem. In: Lerma EV, Batuman V, eds. Diabetes and Ernest Law was also supported by the University Kidney Disease. Heidelberg: Springer; 2014. p. 1–14. of Illinois/Takeda Health Economics and Out- 3. Lim A. Diabetic nephropathy—complications and comes Research Fellowship. Zhou Zhou is an treatment. Int J Nephrol Renovasc Dis. employee of Analysis Group, Inc., which has 2014;7:361–81. received consulting fees from Takeda Pharma- ceuticals. Hongbo Yang is also an employee of 4. Valmadrid CT, Klein R, Moss SE, Klein BE. The risk of cardiovascular disease mortality associated with Analysis Group, Inc., which has received con- microalbuminuria and gross proteinuria in persons sulting fees from Takeda Pharmaceuticals. with older-onset diabetes mellitus. Arch Intern Additionally, Eric Q. Wu is an employee of Med. 2000;160:1093–100. Analysis Group, Inc., which has received Diabetes Ther (2018) 9:1021–1036 1035 5. Kim JJ, Hwang BH, Choi IJ, et al. A prospective two- in U.S. integrated health care delivery systems: center study on the associations between microalbu- 2005–2011. Diabetes care. 2015;39(3):363–70. minuria, coronary atherosclerosis and long-term clin- ical outcome in asymptomatic patients with type 2 17. Lanza ST, Rhoades BL. Latent class analysis: an diabetes mellitus: evaluation by coronary CT angiog- alternative perspective on subgroup analysis in pre- raphy. Int J Cardiovasc Imaging. 2015;31:193–203. vention and treatment. Prev Sci. 2013;14:157–68. 6. Ahmad J. Management of diabetic nephropathy: 18. Virtanen M, Vahtera J, Head J, et al. Work disability recent progress and future perspective. Diabetes among employees with diabetes: latent class anal- Metab Syndr. 2015;9:343–58. ysis of risk factors in three prospective cohort studies. PLoS One. 2015;10:e0143184. 7. de Boer IH, Rue TC, Hall YN, Heagerty PJ, Weiss NS, Himmelfarb J. Temporal trends in the prevalence of 19. Fitzpatrick SL, Coughlin JW, Appel LJ, et al. Appli- diabetic kidney disease in the United States. JAMA. cation of latent class analysis to identify behavioral 2011;305:2532–9. patterns of response to behavioral lifestyle inter- ventions in overweight and obese adults. Int J 8. Afkarian M, Zelnick LR, Hall YN, et al. Clinical Behav Med. 2015;22:471–80. manifestations of kidney disease among US adults with diabetes, 1988–2014. JAMA. 2016;316:602–10. 20. Dey A, Chakraborty A, Majumdar K, Mandel A. Application of latent class analysis to estimate sus- 9. Haneda M, Utsunomiya K, Koya D, et al. A new ceptibility to adverse health outcomes based on classification of diabetic nephropathy 2014: a several risk factors. Int J Com Med Pub Health. report from joint committee on diabetic 2016;3:3423–9. nephropathy. J Diabetes Investig. 2015;6:242–6. 21. Jiang L, Beals J, Zhang L, et al. Latent class analysis 10. de Boer IH, Rue TC, Cleary PA, et al. Long-term of stages of change for multiple health behaviors: renal outcomes of patients with type 1 diabetes results from the special diabetes program for indi- mellitus and microalbuminuria: an analysis of the ans diabetes prevention program. Prev Sci. diabetes control and complications trial/epidemi- 2012;13:449–61. ology of diabetes interventions and complications cohort. Arch Intern Med. 2011;171:412–20. 22. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM 11. Karalliedde J, Gnudi L. Diabetes mellitus, a complex and ICD-10 administrative data. Med Care. and heterogeneous disease, and the role of insulin 2005;43:1130–9. resistance as a determinant of diabetic kidney dis- ease. Nephrol Dial Transplant. 2016;31:206–13. 23. Gross JL, de Azevedo MJ, Silveiro SP, Canani LH, Caramori ML, Zelmanovitz T. Diabetic nephropa- 12. Adler AI, Stevens RJ, Manley SE, et al. Development thy: diagnosis, prevention, and treatment. Diabetes and progression of nephropathy in type 2 diabetes: Care. 2005;28:164–76. the United Kingdom Prospective Diabetes Study (UKPDS 64). Kidney Int. 2003;63:225–32. 24. Bureau of Labor Statistics (U.S. Department of Labor). Consumer price index 2016 [Jan 1, 2017]. 13. Lizicarova D, Krahulec B, Hirnerova E, Gaspar L, Available from: https://www.bls.gov/cpi/. Celecova Z. Risk factors in diabetic nephropathy progression at present. Bratisl Lek Listy. 25. Lanza ST, Collins LM, Lemmon DR, Schafer JL. 2014;115:517–21. PROC LCA: a SAS procedure for latent class analysis. Struct Equ Model. 2007;14:671–94. 14. Duran-Perez EG, Almeda-Valdes P, Cuevas-Ramos D, Campos-Barrera E, Munoz-Hernandez L, Gomez- 26. Rossing K. Progression and remission of nephropa- Perez FJ. Treatment of metabolic syndrome slows thy in type 2 diabetes: new strategies of treatment progression of diabetic nephropathy. Metab Syndr and monitoring. Dan Med Bull. 2007;54:79–98. Relat Disord. 2011;9:483–9. 27. Parving HH, Hovind P, Rossing K, Andersen S. 15. Araki S, Haneda M, Sugimoto T, et al. Factors asso- Evolving strategies for renoprotection: diabetic ciated with frequent remission of microalbumin- nephropathy. Curr Opin Nephrol Hypertens. uria in patients with type 2 diabetes. Diabetes. 2001;10:515–22. 2005;54:2983–7. 28. Hebert LA, Bain RP, Verme D, et al. Remission of 16. Pathak RD, Schroeder EB, Seaquist ER, et al. Severe nephrotic range proteinuria in type I diabetes. hypoglycemia requiring medical intervention in a Collaborative study group. Kidney Int. large cohort of adults with diabetes receiving care 1994;46:1688–93. 1036 Diabetes Ther (2018) 9:1021–1036 29. Hovind P, Rossing P, Tarnow L, Toft H, Parving J, 39. Ho PM, Rumsfeld JS, Masoudi FA, McClure DL, Parving HH. Remission of nephrotic-range albu- Plomondon ME, Steiner JF, et al. Effect of medica- minuria in type 1 diabetic patients. Diabetes Care. tion nonadherence on hospitalization and mortal- 2001;24:1972–7. ity among patients with diabetes mellitus. Arch Intern Med. 2006;166:1836–41. 30. Hovind P, Tarnow L, Rossing P, Carstensen B, Parving HH. Improved survival in patients obtain- 40. Wild H. The economic rationale for adherence in ing remission of nephrotic range albuminuria in the treatment of type 2 diabetes mellitus. Am J diabetic nephropathy. Kidney Int. 2004;66:1180–6. Manag Care. 2012;18:S43–8. 31. Rossing K, Christensen PK, Hovind P, Parving HH. 41. Menzin J, Korn JR, Cohen J, et al. Relationship Remission of nephrotic-range albuminuria reduces between glycemic control and diabetes-related risk of end-stage renal disease and improves survival hospital costs in patients with type 1 or type 2 in type 2 diabetic patients. Diabetologia. diabetes mellitus. JMCP. 2010;16:264–75. 2005;48:2241–7. 42. Gandra SR, Lawrence LW, Parasuraman BM, Darin 32. Alwakeel JS, Isnani AC, Alsuwaida A, et al. Factors RM, Sherman JJ, Wall JL. Total and component affecting the progression of diabetic nephropathy health care costs in a non-Medicare HMO popula- and its complications: a single-center experience in tion of patients with and without type 2 diabetes Saudi Arabia. Ann Saudi Med. 2011;31:236–42. and with and without macrovascular disease. JMCP. 2006;12:546–54. 33. Ohkubo Y, Kishikawa H, Araki E, et al. Intensive insulin therapy prevents the progression of diabetic 43. Lipska KJ, Warton EM, Huang ES, et al. HbA1c and microvascular complications in Japanese patients risk of severe hypoglycemia in type 2 diabetes: the with non-insulin-dependent diabetes mellitus: a diabetes and aging study. Diabetes Care. randomized prospective 6-year study. Diabetes Res 2013;36:3535–42. Clin Pract. 1995;28:103–17. 44. Zhao W, Katzmarzyk PT, Horswell R, Wang Y, 34. Ismail-Beigi F, Craven T, Banerji MA, et al. Effect of Johnson J, Hu G. HbA1c and coronary heart disease intensive treatment of hyperglycaemia on risk among diabetic patients. Diabetes Care. microvascular outcomes in type 2 diabetes: an 2014;37:428–35. analysis of the ACCORD randomised trial. Lancet. 2010;376:419–30. 45. Selby JV, Karter AJ, Ackerson LM, Ferrara A, Liu J. Developing a prediction rule from automated clin- 35. Callaghan BC, Little AA, Feldman EL, Hughes RA. ical databases to identify high-risk patients in a Enhanced glucose control for preventing and large population with diabetes. Diabetes Care. treating diabetic neuropathy. Cochrane Database 2001;24:1547–55. Syst Rev. 2012;6:CD007543. 46. Rosenzweig JL, Weinger K, Poirier-Solomon L, 36. Banerji MA, Dunn JD. Impact of glycemic control Rushton M. Use of a disease severity index for on healthcare resource utilization and costs of type evaluation of healthcare costs and management of 2 diabetes: current and future pharmacologic comorbidities of patients with diabetes mellitus. approaches to improving outcomes. Am Health Am J Manag Care. 2002;8:950–8. Drug Benefits. 2013;6:382–92. 47. Young BA, Lin E, Von Korff M, et al. Diabetes 37. Stark Casagrande S, Fradkin JE, Saydah SH, Rust KF, complications severity index and risk of mortality, Cowie CC. The prevalence of meeting A1C, blood hospitalization, and healthcare utilization. Am J pressure, and LDL goals among people with dia- Manag Care. 2008;14:15–23. betes, 1988–2010. Diabetes Care. 2013;36:2271–9. 48. Porrini E, Ruggenenti P, Mogensen CE, et al. Non- 38. Cramer JA. A systematic review of adherence with proteinuric pathways in loss of renal function in medications for diabetes. Diabetes Care. patients with type 2 diabetes. Lancet Diabetes 2004;27:1218–24. Endocrinol. 2015;3:382–91. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diabetes Therapy Springer Journals

Clinical Trajectories, Healthcare Resource Use, and Costs of Diabetic Nephropathy Among Patients with Type 2 Diabetes: A Latent Class Analysis

Free
16 pages

Loading next page...
 
/lp/springer_journal/clinical-trajectories-healthcare-resource-use-and-costs-of-diabetic-E7YBWl014u
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Internal Medicine; Diabetes; Cardiology; Endocrinology
ISSN
1869-6953
eISSN
1869-6961
D.O.I.
10.1007/s13300-018-0410-8
Publisher site
See Article on Publisher Site

Abstract

Diabetes Ther (2018) 9:1021–1036 https://doi.org/10.1007/s13300-018-0410-8 ORIGINAL RESEARCH Clinical Trajectories, Healthcare Resource Use, and Costs of Diabetic Nephropathy Among Patients with Type 2 Diabetes: A Latent Class Analysis . . . . Ruixuan Jiang Ernest Law Zhou Zhou Hongbo Yang Eric Q. Wu Raafat Seifeldin Received: January 31, 2018 / Published online: March 29, 2018 The Author(s) 2018 diagnosis of T2DM and C 2 urine albumin tests ABSTRACT within the Truven MarketScan database (2004–2014), based on T2DM-related compli- Introduction: Patients with type 2 diabetes cations, comorbidities, and therapies. DN mellitus (T2DM) are clinically heterogeneous in severity categories (normoalbuminuria, moder- terms of disease severity, treatment, and ately increased albuminuria, and severely comorbidities, potentially resulting in differen- increased albuminuria) were determined based tial diabetic nephropathy (DN) progression on urine albumin measure. The risks of DN courses. In this exploratory study we used latent progression and reversal (change to a more/less class analysis (LCA) to identify patient groups severe DN category) were compared among all with distinct clinical profiles of T2DM severity identified latent classes using Kaplan–Meier and explored the association between disease analyses and log-rank tests. All-cause and DN- severity, DN progression or reversal, and related costs and HRU were assessed and com- healthcare resource use (HRU) and costs. pared during the study period among the Methods: Latent class analysis was used to identified latent classes. group adults with C 2 medical claims with a Results: Four clinically distinct profiles were identified among the 23,235 eligible patients: Enhanced content To view enhanced content for this low comorbidity/low treatment (46.5%), low article go to https://doi.org/10.6084/m9.figshare. comorbidity/high treatment (29.0%), moderate comorbidity/high insulin use (9.7%), and high Electronic supplementary material The online comorbidity/moderate treatment (14.8%). The version of this article (https://doi.org/10.1007/s13300- 018-0410-8) contains supplementary material, which is 5-year DN progression rates for these clinically available to authorized users. distinct profiles were 11.8, 18, 16.5, and 27.7%, respectively. Compared with the low comor- R. Jiang  E. Law (&) bidity/low treatment group, all other groups Department of Pharmacy Systems, Outcomes, and were associated with an increased risk of DN Policy, College of Pharmacy, University of Illinois at progression (all p \ 0.001). Increasing comor- Chicago, Chicago, IL, USA e-mail: elaw3@uic.edu bidity was significantly associated with higher all-cause and DN-related HRU and costs, pri- Z. Zhou  H. Yang  E. Q. Wu marily driven by higher pharmacy and inpa- Analysis Group, Inc., Boston, MA, USA tient costs. R. Seifeldin Conclusion: Patients with T2DM who have Formerly of Takeda Development Center Americas, more comorbidities experienced higher rates of Inc., Deerfield, IL, USA 1022 Diabetes Ther (2018) 9:1021–1036 DN progression and HRU and incurred higher Prospective Diabetes Study (UKPDS) examined healthcare costs compared with patients with the progression of DN among 5100 T2DM low comorbidity profiles. Future prospective patients followed for over a decade [12]. The studies are needed to confirm the significance of UKPDS reported the rates and time to DN pro- these groups on DN progression, HRU, and gression among patients with normoalbumin- costs. uria, moderately increased albuminuria, or Funding: Takeda Development Center Americ- severely increased albuminuria, and found that as, Inc. the more severe the albuminuria, the shorter the time to progression to the next stage of DN [12]. In addition, in T2DM patients, long dura- Keywords: Clinical outcomes; Costs; Diabetic tion and poor control have been associated with nephropathy; Healthcare resource use; Type 2 DN progression [13], while the treatment of diabetes metabolic syndrome has been shown to be independently associated with lesser progres- INTRODUCTION sion of DN in T2DM [14]. A prospective study of Japanese patients with T2DM examined the frequency of and reasons for DN reversal and Diabetic nephropathy (DN), often referred as diabetic kidney disease, is a common and seri- found that DN duration, DN treatment type, glycated hemoglobin (HbA1c) levels, and heal- ous complication of type 2 diabetes mellitus (T2DM), and presents as albuminuria and/or thy blood pressure were linked to likelihood of reversal [15]. The relationship between factors decreased glomerular filtration rate (GFR) [1–3]. In addition to being a leading cause of end-stage related to treatment type and comorbidity cannot be assumed to lie on a continuum, but renal disease (ESRD), DN greatly increases the risk of all-cause and cardiovascular-related rather may co-vary in different ways in different patients, resulting in distinct clinical patterns. mortality, cardiovascular disease and coronary atherosclerosis, and kidney failure among These patterns, or their relationship to mean- ingful outcomes, have not been explored. patients with diabetes [4, 5]. DN affects up to Understanding the relationships between 40% of patients with T2DM [6] and had an estimated prevalence of 3.3% (6.9 million DN progression and T2DM comorbidity and severity, as well as which patients are at partic- adults) within the USA during 2005–2008 [7]. The overall prevalence of DN among U.S. adults ular risk of progression or poor outcomes, may help improve patient-centered disease manage- with diabetes did not change significantly from 1988 to 2014 [8]. The disease progression of DN ment. The use of real-world patient data from claims databases to characterize T2DM could is classified by the following stages based on urine albumin levels: normoalbuminuria contribute to this goal, as real-world patients have been shown to be more heterogeneous in (\ 30 mg/24 h or an albumin/creatinine ratio terms of both disease severity and diabetic [ACR] of \ 30 lg/mg), moderately increased treatments received [16], and claims data are a albuminuria (30–300 mg/24 h or an ACR of large and accessible resource. Latent class anal- 30–300 lg/mg), and severely increased albu- minuria ([ 300 mg/24 h or an ACR [ 300 lg/ ysis (LCA), a statistical method for grouping individuals who share common characteristics mg) [9]. Moderately increased albuminuria pre- cedes severely increased albuminuria, and and thereby allowing distinct ‘‘clusters’’ to be identified [17], can integrate the limited disease without treatment, the GFR progressively declines and can ultimately result in ESRD [10]. information available in claims databases. LCA has been previously used to examine hetero- However, the severity of T2DM is clinically heterogeneous among affected patients, which geneity in patient populations, to identify sub- groups at high potential risk of disability or may also result in differential courses of DN adverse effects, or to predict patients who may progression [11]. Several prior studies have identified risk fac- benefit from particular interventions in various disease areas [18–21]. tors for DN progression. The United Kingdom Diabetes Ther (2018) 9:1021–1036 1023 In the study reported here, we used LCA to with continuous enrollment for C 12 months integrate multiple indicators of T2DM severity prior to and 6 months following the test date; that are readily available in commonly used and (3) without ESRD during the 12 months claims databases to identify patient groups with prior to the test date (Fig. 1). The index date was distinct clinical profiles. Differences in the randomly selected among all eligible urine clinical trajectory of DN (i.e., DN progression albumin test dates. and reversal) were then assessed among the identified latent classes. Additionally, as there is Study Measures and Outcomes limited information regarding the healthcare economic burden among patient groups with Type 2 diabetes mellitus severity indicators (di- distinct clinical profiles of T2DM, we assessed abetic comorbidities, complications, and treat- the health resource utilization and healthcare ments) included in the LCA were assessed costs among the patient groups identified with during the 12-month period prior to the index LCA and identified the patient subgroup with date. T2DM-related complications and comor- the highest economic burden. bidities included retinopathy, neuropathy, ischemic heart disease, cerebrovascular disease, chronic heart failure, hypertension, chronic METHODS kidney disease (CKD)-related symptoms (hy- perkalemia, high parathyroid hormone level, Data Source and high phosphorus level), and metabolic disorder (lipid disorders and other metabolic This study used data from the Truven Health disorders) (see Appendix A in the Electronic Analytics MarketScan Commercial and Medi- Supplementary Material [ESM]). Diabetic treat- care Supplemental and Lab databases (1 January ments included insulin, metformin, sulphony- 2003 to 31 December 2014), which represent lureas, dipeptidyl peptidase-4 inhibitor, approximately 25 million U.S. employees, glucagon-like peptide-1-based therapy, and dependents, and Medicare-eligible retirees cov- other antidiabetic agents (amylin analogs, ered by over 130 health plans and self-insured amino acid derivatives, meglitinide analogs, employers. A subset of the covered lives (ap- aldose reductase inhibitors, alpha-glucosidase prox. 1 million) have recorded laboratory tests inhibitors, dopamine receptor agonists, insulin (mainly ordered in office-based practice) in the sensitizing agents and antidiabetic combina- MarketScan Lab Database. Data were de-iden- tions) (see ESM Appendix B). tified and comply with the patient confiden- tiality requirements of the U.S. Health Baseline Characteristics Insurance Portability and Accountability Act; Baseline characteristics (demographics, disease thus, no institutional review was required. characteristics, and DN-related treatments) were assessed during the 12-month period prior to Study Population and Study Cohorts the index date. Patient demographic informa- tion collected included age, sex, and type of The study sample included adult patients (aged health insurance. The time from the first C 18 years) with C 2 distinct medical claims observed T2DM diagnosis to the index date and with a diagnosis of T2DM (International Clas- patients’ DN severity at the index date (nor- sification of Diseases, 9th revision-Clinical moalbuminuria, moderately increased albu- Modification [ICD-9-CM]: 250.x0, 250.x2) and minuria, or severely increased albuminuria) with C 2 urine albumin test results after the first were assessed. Normoalbuminuria was defined observed T2DM diagnosis. Patients with at least as urinary albumin excretion of\30 mg/24 h or one eligible urine albumin test who met the an ACR of \ 30 lg/mg; moderately increased following criteria were identified: (1) with C 1 albuminuria was defined as excretion of additional follow-up urine albumin test; (2) 30–300 mg/24 h or an ACR of 30–300 lg/mg; 1024 Diabetes Ther (2018) 9:1021–1036 Fig. 1 Sample selection. ESRD End-stage renal disease, T2DM type 2 diabetes mellitus Progression or Reversal of DN severely increased albuminuria was defined as excretion of[ 300 mg/24 h or an ACR[ 300 lg/ The time to DN progression and time to DN reversal were assessed from the index date until mg. In addition, for each patient the Charlson Comorbidity Index (CCI) score [22] and the use inpatient mortality, the end of continuous eli- of DN-related treatments during the 12-month gibility, or the end of data availability, which- period before the index date were recorded. DN- ever came first. DN severity was classified into related treatments included angiotensin con- four disease stages ranging from normal to most verting enzyme inhibitors, angiotensin receptor severe: normoalbuminuria, moderately blockers, diuretics, calcium channel blockers, increased albuminuria, severely increased albu- minuria, and presence of ESRD/dialysis/renal and other antihypertensive agents [23] (see ESM Appendix B). transplantation procedure (see ESM Appendix C). DN disease progression was defined as the Diabetes Ther (2018) 9:1021–1036 1025 presence of a urine albumin test, diagnosis or Information Criterion and Akaike Information procedure indicating a more severe disease stage Criterion) and interpretability of groups. Once than the index disease stage, while DN disease the model was selected, subjects were assigned reversal was defined as the presence of a urine to the latent class based on their probability of albumin test indicating a less severe disease being in that class. Analyses were conducted in stage than the index disease stage. DN disease SAS version 9.4 software using the PROC LCA reversal was only assessed in patients with procedure [25]. moderately increased albuminuria and severely For baseline characteristics, means and increased albuminuria at the index date. standard deviations (SD) were reported for continuous characteristics; frequencies and percentages were reported for categorical char- Healthcare Resource Use and Costs acteristics. Characteristics were compared Economic outcomes were assessed at the per- between the selected reference latent class ver- patient-per-year (PPPY) level from the index sus other latent classes using Wilcoxon rank- date until 2 years from the index date, the end sum tests for continuous variables and Chi- of continuous eligibility, end of data availabil- square tests for categorical variables. Time to ity, or inpatient mortality, whichever came first. disease progression and time to disease reversal Healthcare resource use (HRU) information was were evaluated for each latent class using collected for all-cause and DN-related medical Kaplan–Meier methods and compared between visits, including inpatient, emergency room the reference latent class versus other classes (ER), outpatient, and other medical visits. DN- using log-rank tests. The incidence rates of each related HRU was defined as medical services type of healthcare visit PPPY and the annual associated with a diagnosis of DN or kidney healthcare costs were described for each latent disease, or a procedure for dialysis/hemodialysis class and compared between the reference or renal transplantation. latent class versus other classes using Wilcoxon All-cause and nephropathy-related health- rank-sum tests. A p value of B 0.05 was consid- care costs were calculated from a U.S. payer’s ered to be statistically significant. perspective and inflated to 2016 U.S. dollars All procedures followed were in accordance using the annual medical care component of with the ethical standards of the responsible the Consumer Price Index [24]. Cost compo- committee on human experimentation (insti- nents included medical costs (inpatient, ER, tutional and national) and with the Helsinki outpatient, and other medical services costs) Declaration of 1964, as revised in 2013. Only and pharmacy costs. DN-related costs were de-identified data was used in this study, thus defined as costs associated with a diagnosis code no institutional review was required. This article of diabetic nephropathy or costs associated with is based on previously conducted studies and a procedure for dialysis/hemodialysis or renal does not contain any studies with human par- transplantation. ticipants or animals performed by any of the authors. Statistical Analysis Latent class analysis was used to identify groups RESULTS of patients with clinically distinct T2DM sever- ity profiles. Class membership was determined Latent Class Analysis based on the T2DM disease severity and treat- ment indicators, and individual patients could A total of 23,235 patients with T2DM fulfilled belong to only one group. Models with a vary- all study criteria and were included in the ing number of classes were estimated, and the analysis, including 18,409 patients with nor- best-fitting model was chosen. Model selection moalbuminuria, 3863 with moderately was based on the consideration of several cri- increased albuminuria, and 963 with severely teria, including model fit statistics (Bayesian increased albuminuria (Fig. 1). In the LCA, a 1026 Diabetes Ther (2018) 9:1021–1036 four-class model yielded the best fit, and four low comorbidity/low treatment group (all clinically distinct T2DM patient profiles were p \ 0.05) (Table 2). Compared with the low identified based on distributions of complica- comorbidity/low treatment group, all other tions/comorbidities (e.g., microvascular and groups had significantly longer mean time since cardiovascular disease, CKD-related symptoms, first observed T2DM diagnosis (low comorbid- and metabolic disorder) and use of diabetic ity/low treatment: 32.9 [SD 25.1] months; low treatment (e.g., insulin, metformin, etc.) comorbidity/high treatment: 42.0 [26.8] (Table 1). The four clinically distinct profiles months; moderate comorbidity/high insulin were defined as: (1) Latent Class 1 (46.5% of the use: 42.6 [26.9] months; high comorbid- sample [N = 10,812]; reference latent class), low ity/moderate treatment: 46.6 [29.6] months; all comorbidity/low treatment; (2) Latent Class 2 p \ 0.01). Patients in the high comorbid- (29.0% [N = 6728]), low comorbidity/high ity/moderate treatment group had the highest treatment; (3) Latent Class 3 (9.7% [N = 2255]), mean CCI (mean 2.9 [SD 1.6]), followed by the moderate comorbidity and high insulin use; (4) moderate comorbidity/high insulin group (1.7 Latent Class 4 (14.8% [N = 3440]), high [1.1]), the low comorbidity/high treatment comorbidity and moderate treatment. group (1.6 [1.0]), and the low comorbidity/low Retinopathy was more common among Latent treatment group (1.4 [1.0]). All other groups had Classes 3 (26.8%) and 4 (24.4%) compared with significantly higher mean CCI compared with Latent Classes 1 (2.6%) and 2 (13.3%), and the low comorbidity/low treatment group (all Latent Class 4 had a higher incidence of all p \ 0.05). other comorbidities in comparison with the other three latent classes. Time to DN Disease Progression Baseline Demographics and Disease The median years of follow-up for DN progres- Characteristics sion among the low comorbidity/low treat- ment, low comorbidity/high treatment, The baseline demographics and disease charac- moderate comorbidity/high insulin use, and teristics of the four latent classes are described high comorbidity/moderate treatment groups in Table 2; multiple significant differences were 2.1, 1.9, 2.1, and 1.6 years, respectively. existed between the characteristics of the latent The respective 1-, 3-, and 5-year DN progression classes. For example, patients with moderate rates were 9.3, 20.8, and 27.7% for the high comorbidity/high insulin use were younger comorbidity/moderate treatment group; 5.9, (mean [SD] 49.1 [13.3] years) than those in the 13.3, and 16.5% for the moderate comorbidity/ low comorbidity/low treatment group (54.5 high insulin group; 5.6, 13.9, and 18.0% for the [9.2] years), and patients with high comorbid- low comorbidity/high treatment group; and ity/moderate treatment were older (58.2 [8.7] 4.0, 9.6, and 11.8% for the low comorbidity/low years) and more likely to be male than those in treatment group (Fig. 2). Compared to the low the low comorbidity/low treatment group (male comorbidity/low treatment group, all other 57 vs. 54%, respectively; all p \ 0.05). Patients groups were associated with a significantly from the southern USA were over-represented in increased risk of progression to a more severe every class (range 38–44%). Preferred provider stage of DN (all p \ 0.01). The high comorbid- organization was the most common type of ity/moderate treatment group was associated health plan among all classes (range 72–76%). with the highest risk of disease progression The majority of patients in all groups were among all latent groups. classified with normoalbuminuria at the index date (range 68–84%); however, higher propor- Time to DN Disease Reversal tions of patients were classified with moderately increased albuminuria or severely increased The median years of follow-up for DN reversal albuminuria in other groups compared with the among the low comorbidity/low treatment, low Diabetes Ther (2018) 9:1021–1036 1027 Table 1 Item–response probabilities for the four-class model: probability of each patient in a given latent class Item Latent class (%) Latent Class 1 (low Latent Class 2 (low Latent Class 3 Latent Class 4 (high comorbidity/low comorbidity/high (moderate comorbidity/moderate treatment group) treatment group) comorbidity/high treatment group) insulin use group) Comorbidity Microvascular disease Retinopathy 2.6 13.1 26.8 24.4 disease Neuropathy 5.2 13.3 16.4 31.7 disease Cardiovascular disease Ischemic heart 6.8 6.2 4.7 40.7 disease Cerebrovascular 2.4 1.3 2.0 19.0 disease Chronic heart 0.6 0.1 1.6 11.0 failure Hypertension 58.2 64.0 32.8 94.7 CKD-related 0.5 0.9 1.0 5.3 disease Metabolic 64.1 72.1 43.3 92.8 disorder Use of diabetic treatment Metformin 59.8 88.2 18.3 56.3 Sulphonylureas 13.5 48.3 3.5 25.9 Insulin 0.0 21.9 87.7 44.1 DPP4 inhibitor 4.6 19.3 0.7 12.8 GLP1-based 1.5 15.2 3.1 9.5 therapy Other antidiabetic 7.6 35.5 31.3 26.5 agents Item–response probabilities were calculated among all patients, given the probability of each patient being in that latent class CKD Chronic kidney disease, DPP4 dipeptidyl peptidase-4, GLP-1 glucagon-like peptide-1 CKD-related disease included hyperkalemia, high parathyroid hormone level, and high phosphorus level Metabolic disorder included lipid disorders and other metabolic disorders Other antidiabetic agents included amylin analogs, amino acid derivatives, meglitinide analogs, aldose reductase inhibitors, alpha-glucosidase inhibitors, dopamine receptor agonists, insulin sensitizing agents, and antidiabetic combinations 1028 Diabetes Ther (2018) 9:1021–1036 Table 2 Patient baseline demographics and disease characteristics according to latent class Patient baseline demographics Latent Class Latent Latent Latent p value and disease characteristics 1 Class 2 Class 3 Class 4 [2] vs. [3] vs. [4] vs. (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [1] [1] [1] Age at index date (years) 54.5 ± 9.2 54.4 ± 8.6 49.1 ± 13.3 58.2 ± 8.7 \ 0.01 \ 0.001 Male 5795 (54%) 3789 (56%) 1199 (53%) 1965 (57%) \ 0.05 \ 0.001 U.S. region Northeast 2016 (19%) 1196 (18%) 394 (17%) 837 (24%) \ 0.001 North-Central 2826 (26%) 1629 (24%) 722 (32%) 828 (24%) \ 0.05 \ 0.01 \ 0.001 South 4533 (42%) 2976 (44%) 855 (38%) 1367 (40%) \ 0.05 \ 0.01 \ 0.001 West 1434 (13%) 927 (14%) 283 (13%) 408 (12%) \ 0.001 Insurance plan type Preferred provider organization 8094 (75%) 5145 (76%) 1678 (74%) 2483 (72%) \ 0.05 \ 0.001 Non-capitated point-of-service 781 (7%) 538 (8%) 166 (7%) 225 (7%) Exclusive provider organization 501 (5%) 343 (5%) 82 (4%) 228 (7%) \ 0.01 \ 0.001 Comprehensive 1025 (9%) 502 (7%) 198 (9%) 442 (13%) \ 0.05 \ 0.001 Consumer-driven health plan 314 (3%) 141 (2%) 99 (4%) 43 (1%) \ 0.05 \ 0.01 \ 0.001 High-deductible health plan 97 (1%) 59 (1%) 32 (1%) 19 (1%) \ 0.01 \ 0.001 Time from first observed T2DM 32.9 ± 25.1 42.0 ± 26.8 42.6 ± 26.9 46.6 ± 29.6 \ 0.05 \ 0.01 \ 0.001 diagnosis in the database to the index date (months) Diabetic nephropathy disease status at the index date Normal 9083 (84%) 5249 (78%) 1728 (77%) 2349 (68%) \ 0.05 \ 0.01 \ 0.001 Moderately increased 1483 (14%) 1208 (18%) 400 (18%) 772 (22%) \ 0.05 \ 0.01 \ 0.001 albuminuria Severely increased albuminuria 246 (2%) 271 (4%) 127 (6%) 319 (9%) \ 0.05 \ 0.01 \ 0.001 Charlson Comorbidity Index 1.4 ± 1.0 1.6 ± 1.0 1.7 ± 1.1 2.9 ± 1.6 \ 0.05 \ 0.01 \ 0.001 Nephropathy-related treatments 6892 (64%) 5369 (80%) 1352 (60%) 2975 (86%) \ 0.05 \ 0.01 \ 0.001 ACE inhibitor 3045 (28%) 2823 (42%) 782 (35%) 1524 (44%) \ 0.05 \ 0.01 \ 0.001 Diuretic 1646 (15%) 1291 (19%) 327 (15%) 1201 (35%) \ 0.05 \ 0.001 Calcium channel blocker 1339 (12%) 1017 (15%) 225 (10%) 856 (25%) \ 0.05 \ 0.01 \ 0.001 ARB 1208 (11%) 981 (15%) 262 (12%) 704 (20%) \ 0.05 \ 0.001 Diabetes Ther (2018) 9:1021–1036 1029 Table 2 continued Patient baseline demographics Latent Class Latent Latent Latent p value and disease characteristics 1 Class 2 Class 3 Class 4 [2] vs. [3] vs. [4] vs. (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [1] [1] [1] Other antihypertensive agent 2716 (25%) 1978 (29%) 360 (16%) 1108 (32%) \ 0.05 \ 0.01 \ 0.001 Data in table are presented as the mean ± standard deviation (SD) or as an absolute number with the percentage in parenthesis The four latent classes are described in Table 1 and in section ‘‘Latent Class Analysis’’ T2DM Type 2 diabetes mellitus,ACE angiotensin converting enzyme, ARB angiotensin receptor blockers Latent Classes 2–4 were compared to Latent Class 1, respectively Other antihypertensive agents included direct renin inhibitors, antiadrenergic antihypertensives, selective aldosterone receptor antagonists, agents for pheochromocytoma, vasodilators, monoamine oxidase inhibitors, and antihypertensive combinations comorbidity/high treatment, moderate comor- associated with a significantly higher rate of bidity/high insulin use, and high comorbid- disease reversal compared to the low comor- ity/moderate treatment groups were 2.1, 2.0, bidity/low treatment patients (all p \ 0.05). 2.2, and 1.7 years, respectively. The respective 1-, 3-, 5-year DN reversal rates were 7.1, 12.0, All-Cause and DN-Related HRU and 13.9% for the high comorbidity/moderate treatment group; 5.0, 8.9, and 8.9% for the Increasing comorbidity was associated with moderate comorbidity/high insulin group; 6.1, significant increases in the annual frequency of 9.9, and 10.5% for the low comorbidity/high all-cause healthcare visits, with a consistent treatment group; and 4.5, 7.9, and 8.9% for the trend across inpatient, outpatient, ER, and low comorbidity/low treatment group (Fig. 3). other medical services visits (p \ 0.01 in all The low comorbidity/high treatment and high pairwise comparisons vs. low comorbidity/low comorbidity/moderate treatment groups were treatment group) (Table 3). The high Fig. 2 Time to progression to a more severe diabetic nephropathy (DN) disease stage. The respective 1-, 3-, Fig. 3 Time to reversal to a less severe DN disease stage. 5-year DN progression rates were 9.27, 20.79, and 27.70% The respective 1-, 3-, 5-year DN reversal rates were 7.09, for the high comorbidity/moderate treatment group 11.99, and 13.92% for the high comorbidity/moderate (purple); 5.86, 13.26, and 16.46% for the moderate treatment group (purple); 5.00, 8.91, and 8.91% for the comorbidity/high insulin group (green); 5.60, 13.93, and moderate comorbidity/high insulin group (green); 6.11, 17.97% for the low comorbidity/high treatment group 9.90, and 10.54% for the low comorbidity/high treatment (red); and 3.95, 9.54, and 11.78% for the low comorbidity/ group (red); and 4.51, 7.88, and 8.89% for the low low treatment group (blue) comorbidity/low treatment group (blue) 1030 Diabetes Ther (2018) 9:1021–1036 Table 3 All-cause and diabetic nephropathy-related healthcare resource use and healthcare costs Per patient per year HRU and Latent Class 1 Latent Class 2 Latent Class 3 Latent Class 4 p value costs, mean – SD (N = 10,812) (N = 6728) (N = 2255) (N = 3440) [2] vs. [3] vs. [4] vs. [1] [1] [1] All-cause HRU Inpatient admissions 0.09 ± 0.29 0.11 ± 0.36 0.14 ± 0.48 0.26 ± 0.62 \ 0.05 \ 0.01 \ 0.001 Inpatient days 0.44 ± 2.23 0.62 ± 3.18 0.83 ± 3.75 1.71 ± 6.30 \ 0.05 \ 0.01 \ 0.001 Emergency room services 0.39 ± 1.12 0.42 ± 1.04 0.58 ± 2.83 0.80 ± 2.12 \ 0.05 \ 0.01 \ 0.001 Outpatient services 12.96 ± 11.26 13.94 ± 11.66 14.87 ± 13.11 21.07 ± 16.71 \ 0.05 \ 0.01 \ 0.001 Other 1.65 ± 3.56 2.13 ± 3.55 4.22 ± 4.93 4.60 ± 7.57 \ 0.05 \ 0.01 \ 0.001 DN-related HRU Inpatient admissions 0.00 ± 0.06 0.01 ± 0.12 0.02 ± 0.17 0.05 ± 0.28 \ 0.05 \ 0.01 \ 0.001 Inpatient days 0.03 ± 0.55 0.11 ± 1.80 0.15 ± 1.47 0.53 ± 4.35 \ 0.05 \ 0.01 \ 0.001 Emergency room services 0.00 ± 0.07 0.00 ± 0.09 0.01 ± 0.15 0.03 ± 0.25 \ 0.01 \ 0.001 Outpatient services 0.22 ± 1.63 0.35 ± 1.49 0.60 ± 3.49 1.46 ± 8.17 \ 0.05 \ 0.01 \ 0.001 Other 0.01 ± 0.59 0.02 ± 0.49 0.05 ± 0.63 0.20 ± 3.67 \ 0.05 \ 0.01 \ 0.001 All-cause healthcare costs (2016 USD) Total healthcare costs 10,172 ± 19,518 13,836 ± 23,696 17,411 ± 25,149 24,564 ± 36,317 \ 0.05 \ 0.01 \ 0.001 Total medical costs 7,383 ± 18,161 8,709 ± 22,059 11,314 ± 23,555 17,812 ± 34,637 \ 0.05 \ 0.01 \ 0.001 Inpatient admission costs 2300 ± 11,886 3107 ± 16,511 3521 ± 14,939 7616 ± 27,065 \ 0.05 \ 0.01 \ 0.001 Emergency room service costs 645 ± 2760 703 ± 3329 997 ± 9189 1581 ± 6625 \ 0.05 \ 0.01 \ 0.001 Outpatient service costs 3973 ± 9413 4252 ± 8327 4520 ± 9125 6745 ± 10984 \ 0.05 \ 0.01 \ 0.001 Other medical service costs 466 ± 2036 646 ± 2048 2275 ± 4455 1870 ± 5730 \ 0.05 \ 0.01 \ 0.001 Total pharmaceutical costs 2788 ± 4920 5127 ± 5828 6097 ± 6003 6753 ± 8028 \ 0.05 \ 0.01 \ 0.001 DN-related healthcare costs (2016 USD) Total healthcare costs 378 ± 2088 475 ± 2088 700 ± 5015 1687 ± 11,386 \ 0.05 \ 0.001 Total medical costs 113 ± 2017 166 ± 2000 474 ± 4977 1292 ± 11,369 \ 0.05 \ 0.01 \ 0.001 Inpatient admission costs 58 ± 1500 90 ± 1589 299 ± 3800 701 ± 5867 \ 0.05 \ 0.01 \ 0.001 Emergency room service costs 4 ± 127 6 ± 137 9 ± 206 28 ± 326 \ 0.01 \ 0.001 Outpatient service costs 49 ± 1191 68 ± 744 141 ± 1746 458 ± 6318 \ 0.05 \ 0.01 \ 0.001 Other medical service costs 2 ± 90 3 ± 84 25 ± 366 105 ± 2251 \ 0.05 \ 0.01 \ 0.001 Total pharmaceutical costs 265 ± 514 309 ± 530 226 ± 444 395 ± 624 \ 0.05 \ 0.01 \ 0.001 Data in table are presented as the mean ± SD The four latent classes are described in Table 1 DN Diabetic neuropathy, HRU health resource utilization, USD U.S. dollars Latent Classes 2–4 were compared to Latent Class 1, respectively Other medical services included durable medical equipment, home care, skilled nursing facility care, and dental or vision care comorbidity/moderate treatment group experi- had 13.0 (SD 11.3) visits (p \ 0.001). In addi- enced a mean of 0.3 (SD 0.6) inpatient admis- tion, the high comorbidity/moderate treatment sions annually, with a mean length of stay of group had increased all-cause HRU compared 1.7 (SD 6.3) days, while the low comorbidity/ with the low comorbidity/high treatment or low treatment group experienced 0.1 (SD 0.3) moderate comorbidity/high insulin use groups admissions with a mean stay of 0.4 (SD 2.2) days in each HRU category. (p \ 0.001). For outpatient services, the high A similar trend was observed regarding DN- comorbidity/moderate treatment group experi- related HRU; patients with high comorbid- enced a mean of 21.1 (SD 16.7) visits annually, ity/moderate treatment had a notably high while the low comorbidity/low treatment group incidence of nephropathy-related inpatient, ER, Diabetes Ther (2018) 9:1021–1036 1031 outpatient, and other medical services visits in Regarding the components that were the largest comparison with the other groups (all p \ 0.05). contributors to total healthcare costs, a similar For example, the high comorbidity/moderate trend was observed in DN-related costs as with all- treatment group had a mean of 1.5 (SD:8.2) DN- cause costs, with pharmacy costs contributing the related outpatient visits annually, compared to majority of the cost difference between the low 0.22 (SD 1.36) visits for the low comorbidity/ comorbidity/low treatment and low comorbid- low treatment group (p \ 0.001). ity/high treatment groups. Higher inpatient admission costs and outpatient costs drove the increase in DN-related healthcare costs among All-Cause and DN-Related Healthcare patients with increasing comorbidity (i.e., low Costs comorbidity/low treatment group vs. the moder- ate comorbidity/high insulin group and the high Following from the observed trend of higher HRU comorbidity/moderate treatment group). associated with higher comorbidity burden, In this study, we did not adjust for baseline higher comorbidity was also associated with sig- characteristics in the outcome comparisons. nificantly higher all-cause healthcare costs The reason for this decision is that latent (Table 3). The incremental differences in total all- patient groups, as the main effect in the model, cause healthcare costs PPPY between the low were identified by LCA using complications, comorbidity/low treatment group and the low comorbidity, and therapies for T2DM during comorbidity/high treatment, moderate comor- the baseline period. Therefore, these character- bidity/high insulin use, and high comorbid- istics should not be adjusted for when outcomes ity/moderate treatment groups were US$3664, are compared across latent patient groups. In US$7239, and US$14,392, respectively (all addition, baseline demographics are in general p\ 0.05). Pharmaceutical cost differences comparable across groups. (US$2339) accounted for approximately 60% of the all-cause cost differences between the low comorbidity/high treatment and low comorbid- DISCUSSION ity/low treatment groups (the next largest com- ponent was inpatient admission costs [22%; cost A large proportion of patients with diabetes difference US$807]). Similarly, pharmaceutical develop DN, leading to progressive increases in cost differences (US$3309) accounted for approx- albuminuria, declining GFR, and risk of ESRD imately50% of thedifferencebetween themod- [1]. Thus, an important step towards improving erate comorbidity/high insulin use and low the management, treatment, and clinical out- comorbidity/low treatment groups (the next lar- comes of patients with T2DM is to identify gest components were other medical services patients at high risk of DN progression. To the [25%; cost difference US$1809] and inpatient best of our knowledge, this is the first study that admission [17%; US$1221] costs). Conversely, the has utilized LCA to identify subgroups of T2DM majority of the cost differences between high patients based on their diabetes comorbidity comorbidity/moderate treatment and low comor- and treatment profile within a claims database, bidity/low treatment groups was due to differences and that has compared the clinical trajectories in inpatient admission costs (37%; cost difference of DN, HRU, and costs among the identified US$5316), followed by pharmaceutical (28%; subgroups. Patients with larger burdens of US$3965) and outpatient (19%; US$2772) costs. comorbidity had significantly increased risk of The incremental differences in total DN-re- progression to a more severe stage of DN, as well lated costs PPPY between the low comorbidity/ as higher all-cause and DN-related HRU and low treatment group and the low comorbidity/ healthcare costs. In particular, patients with high treatment, moderate comorbidity/high high comorbidity and moderate treatment use insulin use, and high comorbidity/moderate had the highest 5-year DN progression rate, treatment groups were $US97 (p\ 0.05), US$322, HRU, and costs compared to all of the other and US$1309 (p \ 0.001), respectively (Table 3). groups. The current results demonstrate the 1032 Diabetes Ther (2018) 9:1021–1036 feasibility of using LCA to identify clinical sub- therapy [33]. These prior findings may help groups using many aspects of diabetes severity, explain why group of patients in the current and the validity of the identified subgroups was study with moderate comorbidity/high insulin supported by the differential DN progression, use had lower rates of DN progression (5-year resource use, and costs observed across the rate 16.5%) than the group with low comor- subgroups. bidity/high treatment use (5-year rate 18.0%). Previous research has helped establish the A small number of studies have previously association between DN progression and dia- examined the relationship between T2DM betes complications or anti-diabetic treatments. severity and healthcare costs, reporting findings Hypertension, a common complication of dia- similar to the present results. A 2013 systematic betes, plays a major role in the onset and pro- literature review by Banerji et al. synthesized gression of DN, and anti-hypertensive studies reporting the impact of glycemic control treatment can reduce albuminuria and slow the and treatment adherence, which are important progression of DN [26, 27]. Remission of very contributors to DN and other diabetes compli- advanced DN has been observed among type 1 cations, on the healthcare costs of T2DM diabetes mellitus (T1DM) and T2DM patients patients [36]. These authors observed that the undergoing aggressive anti-hypertensive treat- healthcare resource utilization and costs asso- ment [28–31]. In the present study, the high ciated with T2DM management were reduced comorbidity/moderate treatment group had the when glycemic levels and comorbidity were highest prevalence of hypertension (94.7%) and better controlled, although comorbid condi- the highest rates of DN progression (5-year rate tions were still prevalent and anti-diabetic 27.7%) among all groups. The low comorbidity/ medication adherence was largely suboptimal high treatment group had the next highest [37–40]. A retrospective study in 2010 by Men- prevalence of hypertension (64%), as well as the zin et al. reported that the hospitalization rate next highest rate of DN progression (5-year rate and healthcare costs were significantly higher 18.0%). A previous retrospective analysis for T1DM and T2DM patients with poorly con- reported that the concomitant presence of trolled versus well-controlled blood glucose retinopathy was significantly associated with ([ 10 vs. \ 7% HbA1c) [41]. Furthermore, DN progression in patients with T2DM [32], patients with severe complications or comor- perhaps due to similar underlying contributing bidities related to T2DM have been shown to pathology. In the present study, the low incur healthcare costs that are threefold higher comorbidity/low treatment group had a much than those of matched patients without lower prevalence of retinopathy (2.6%) than did comorbidities [42], similar to the present the other groups (range 13.1–26.8%), as well as results. Our results emphasize the magnitude of the lowest rate of DN progression (5-year rate this cost difference and may present opportu- 11.8% vs. [range] 16.5–27.7%). nities to identify patients at risk of higher In addition, several studies have shown that healthcare expenditures. DN progression can be slowed with intensive The results of this study have potential glycemic control [33–35]. The randomized trial implications for the management of T2DM and ACCORD reported that therapy targeting DN as well as for the further application of this HbA1c to levels of \ 6.0 vs. 7.0–7.9% delayed method to identify high-risk patients based on the onset of severely increased albuminuria and claims data. The heterogeneous nature of T2DM some microvascular complications, but complicates clinical management, and the cur- increased risk mortality and other complica- rent results indicate that some patients (e.g., the tions (e.g., hypoglycemia) [34]. A similar, high comorbidity/moderate treatment group) prospective study reported that anti-diabetic may have received suboptimal anti-diabetic therapy targeting HbA1c to levels of \ 6.5% treatment or suboptimal blood pressure control, resulted in delayed onset of DN, retinopathy, which led to a high risk of DN progression. and neuropathy in comparison to conventional However, although patients with the highest Diabetes Ther (2018) 9:1021–1036 1033 comorbidity were at the highest risk of DN Diabetes Complication Severity Index (DCSI), a progression, the 5-year rates of DN reversal 13-point scale for scoring patients’ diagnostic, among both the low comorbidity/high treat- pharmacy, and laboratory data, has also been ment group (10.5%) and the high comorbid- used to assess healthcare costs and comorbidity ity/moderate treatment group (13.9%) were management [45–47]. However, deriving the higher than those of the low comorbidity/low DSCI from claims data is again restrained by the treatment and moderate comorbidity/high availability of laboratory information in claims insulin use groups (both 8.9%). This result databases. In the present study, we derived dif- suggests that anti-diabetic treatments can be ferent indicators of T2DM disease severity from effective in slowing or reversing DN even in the claims database, which is also subject to the patients with more severe T2DM. Additionally, limitations introduced by the data source. First, the LCA results highlight that there may be DN severity may be misclassified due to the use interactions between comorbid conditions, of the results of a single urine albumin test to T2DM treatment, and patient characteristics as define moderately increased and severely there are no clear indicators of group member- increased albuminuria. To maximize availability ship, i.e., group membership as a proxy for data to assess DN progression or reversal and diabetes severity is not clearly defined by num- because the urine albumin test data were lim- ber or type of T2DM medications or number of ited, we did not require a confirmatory urine co-morbidities, as would be the case in tradi- albumin test. Second, patients with T2DM were tional claims database studies. Future research is identified by ICD-9-CM codes in medical ser- warranted to identify any unmet treatment vices claims; diagnoses or procedures related to needs in this patient population, to characterize ESRD, transplantation, or dialysis were also patients with high risk of DN progression, and identified using diagnostic or procedural codes to identify groups with the largest economic in the claims database. However, such diagnoses burden related to T2DM and DN. In addition, and procedures may be over- or under-reported the current study demonstrates the potential for in the data. Similarly, rates of DN progression LCA to be used to identify classes of patients may have been underestimated, as the data do with clinically distinct T2DM profiles using not contain complete information regarding information available in commercial claims laboratory values. Finally, only patients with a data. This is an important contribution, as urine albumin test were included in the study, claims databases represent large and readily which may not capture all patients with DN accessible sources of healthcare information but disease status since alternative laboratory tests, may lack data typically used to indicate prog- such as proteinuria and estimated GFR decline, nosis (e.g., HbA1c). Future research using LCA may also indicate DN [48]. Although the data- could further contribute to understanding and base is geographically representative of the USA, reducing the risk of DN progression and the data include only commercially insured improving clinical outcomes for patients with patients and those who have commercial T2DM. insurance in supplement to their Medicare insurance. Thus, the current results may not be generalizable across different study populations Limitations that may be relevant, such as various socioeco- nomic groups or the uninsured. Estimating T2DM disease severity using data from claims databases can be challenging due to the limited amount and types of clinical infor- CONCLUSIONS mation they contain. Laboratory measures, such as HbA1c levels, are useful indicators of This exploratory study demonstrated the feasi- diabetes severity [43, 44], although claims bility of using LCA to identify patient groups databases contain limited proportions of with distinct clinical profiles of T2DM severity patients with HbA1c lab values available. The and explored the association between T2DM 1034 Diabetes Ther (2018) 9:1021–1036 disease severity, DN progression or reversal, and consulting fees from Takeda Pharmaceuticals. economic outcomes. Increasing levels of At the time of analysis, Raafat Seifeld was an comorbidity were generally associated with employee of Takeda, and held Takeda stock or higher HRU, healthcare costs, and risk of DN stock options. progression, while anti-diabetic treatment Compliance with Ethics Guidelines. All appeared to slow DN progression. procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and ACKNOWLEDGEMENTS national) and with the Helsinki Declaration of 1964, as revised in 2013. Only de-identified data was used in this study, thus no institutional Funding. Sponsorship for this study and review was required. This article is based on article processing charges were funded by previously conducted studies and does not con- Takeda Development Center Americas, Inc. All tain any studies with human participants or authors had full access to all of the data in this animals performed by any of the authors. study and take complete responsibility for the integrity of the data and accuracy of the data Data Availability. The datasets generated analysis. during and/or analyzed during the current study are not publicly available due to confi- Medical Writing and Editorial Assis- dentiality agreement. tance. Editorial assistance in the preparation of this manuscript was provided by Dr. Shelley Open Access. This article is distributed Batts of Analysis Group, Inc. The authors would under the terms of the Creative Commons like to thank Jing Zhao from Analysis Group for Attribution-NonCommercial 4.0 International significant contribution towards medical writ- License (http://creativecommons.org/licenses/ ing and analytical support. The authors would by-nc/4.0/), which permits any non- also like to thank Melvin Munsaka for his sup- commercial use, distribution, and reproduction port in data analysis. This support was funded in any medium, provided you give appropriate by Takeda Development Center Americans, Inc. credit to the original author(s) and the source, provide a link to the Creative Commons license, Authorship. All named authors meet the and indicate if changes were made. International Committee of Medical Journal Editors (ICMJE) criteria for authorship for this manuscript, take responsibility for the integrity REFERENCES of the work as a whole, and have given final approval to the version to be published. 1. Tuttle KR, Bakris GL, Bilous RW, et al. Diabetic kidney disease: a report from an ADA consensus Disclosures. Ruixuan Jiang was supported conference. Diabetes Care. 2014;37:2864–83. by the University of Illinois/Takeda Health 2. Chen J. Diabetic nephropathy: scope of the prob- Economics and Outcomes Research Fellowship. lem. In: Lerma EV, Batuman V, eds. Diabetes and Ernest Law was also supported by the University Kidney Disease. Heidelberg: Springer; 2014. p. 1–14. of Illinois/Takeda Health Economics and Out- 3. Lim A. Diabetic nephropathy—complications and comes Research Fellowship. Zhou Zhou is an treatment. Int J Nephrol Renovasc Dis. employee of Analysis Group, Inc., which has 2014;7:361–81. received consulting fees from Takeda Pharma- ceuticals. Hongbo Yang is also an employee of 4. Valmadrid CT, Klein R, Moss SE, Klein BE. The risk of cardiovascular disease mortality associated with Analysis Group, Inc., which has received con- microalbuminuria and gross proteinuria in persons sulting fees from Takeda Pharmaceuticals. with older-onset diabetes mellitus. Arch Intern Additionally, Eric Q. Wu is an employee of Med. 2000;160:1093–100. Analysis Group, Inc., which has received Diabetes Ther (2018) 9:1021–1036 1035 5. Kim JJ, Hwang BH, Choi IJ, et al. A prospective two- in U.S. integrated health care delivery systems: center study on the associations between microalbu- 2005–2011. Diabetes care. 2015;39(3):363–70. minuria, coronary atherosclerosis and long-term clin- ical outcome in asymptomatic patients with type 2 17. Lanza ST, Rhoades BL. Latent class analysis: an diabetes mellitus: evaluation by coronary CT angiog- alternative perspective on subgroup analysis in pre- raphy. Int J Cardiovasc Imaging. 2015;31:193–203. vention and treatment. Prev Sci. 2013;14:157–68. 6. Ahmad J. Management of diabetic nephropathy: 18. Virtanen M, Vahtera J, Head J, et al. Work disability recent progress and future perspective. Diabetes among employees with diabetes: latent class anal- Metab Syndr. 2015;9:343–58. ysis of risk factors in three prospective cohort studies. PLoS One. 2015;10:e0143184. 7. de Boer IH, Rue TC, Hall YN, Heagerty PJ, Weiss NS, Himmelfarb J. Temporal trends in the prevalence of 19. Fitzpatrick SL, Coughlin JW, Appel LJ, et al. Appli- diabetic kidney disease in the United States. JAMA. cation of latent class analysis to identify behavioral 2011;305:2532–9. patterns of response to behavioral lifestyle inter- ventions in overweight and obese adults. Int J 8. Afkarian M, Zelnick LR, Hall YN, et al. Clinical Behav Med. 2015;22:471–80. manifestations of kidney disease among US adults with diabetes, 1988–2014. JAMA. 2016;316:602–10. 20. Dey A, Chakraborty A, Majumdar K, Mandel A. Application of latent class analysis to estimate sus- 9. Haneda M, Utsunomiya K, Koya D, et al. A new ceptibility to adverse health outcomes based on classification of diabetic nephropathy 2014: a several risk factors. Int J Com Med Pub Health. report from joint committee on diabetic 2016;3:3423–9. nephropathy. J Diabetes Investig. 2015;6:242–6. 21. Jiang L, Beals J, Zhang L, et al. Latent class analysis 10. de Boer IH, Rue TC, Cleary PA, et al. Long-term of stages of change for multiple health behaviors: renal outcomes of patients with type 1 diabetes results from the special diabetes program for indi- mellitus and microalbuminuria: an analysis of the ans diabetes prevention program. Prev Sci. diabetes control and complications trial/epidemi- 2012;13:449–61. ology of diabetes interventions and complications cohort. Arch Intern Med. 2011;171:412–20. 22. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM 11. Karalliedde J, Gnudi L. Diabetes mellitus, a complex and ICD-10 administrative data. Med Care. and heterogeneous disease, and the role of insulin 2005;43:1130–9. resistance as a determinant of diabetic kidney dis- ease. Nephrol Dial Transplant. 2016;31:206–13. 23. Gross JL, de Azevedo MJ, Silveiro SP, Canani LH, Caramori ML, Zelmanovitz T. Diabetic nephropa- 12. Adler AI, Stevens RJ, Manley SE, et al. Development thy: diagnosis, prevention, and treatment. Diabetes and progression of nephropathy in type 2 diabetes: Care. 2005;28:164–76. the United Kingdom Prospective Diabetes Study (UKPDS 64). Kidney Int. 2003;63:225–32. 24. Bureau of Labor Statistics (U.S. Department of Labor). Consumer price index 2016 [Jan 1, 2017]. 13. Lizicarova D, Krahulec B, Hirnerova E, Gaspar L, Available from: https://www.bls.gov/cpi/. Celecova Z. Risk factors in diabetic nephropathy progression at present. Bratisl Lek Listy. 25. Lanza ST, Collins LM, Lemmon DR, Schafer JL. 2014;115:517–21. PROC LCA: a SAS procedure for latent class analysis. Struct Equ Model. 2007;14:671–94. 14. Duran-Perez EG, Almeda-Valdes P, Cuevas-Ramos D, Campos-Barrera E, Munoz-Hernandez L, Gomez- 26. Rossing K. Progression and remission of nephropa- Perez FJ. Treatment of metabolic syndrome slows thy in type 2 diabetes: new strategies of treatment progression of diabetic nephropathy. Metab Syndr and monitoring. Dan Med Bull. 2007;54:79–98. Relat Disord. 2011;9:483–9. 27. Parving HH, Hovind P, Rossing K, Andersen S. 15. Araki S, Haneda M, Sugimoto T, et al. Factors asso- Evolving strategies for renoprotection: diabetic ciated with frequent remission of microalbumin- nephropathy. Curr Opin Nephrol Hypertens. uria in patients with type 2 diabetes. Diabetes. 2001;10:515–22. 2005;54:2983–7. 28. Hebert LA, Bain RP, Verme D, et al. Remission of 16. Pathak RD, Schroeder EB, Seaquist ER, et al. Severe nephrotic range proteinuria in type I diabetes. hypoglycemia requiring medical intervention in a Collaborative study group. Kidney Int. large cohort of adults with diabetes receiving care 1994;46:1688–93. 1036 Diabetes Ther (2018) 9:1021–1036 29. Hovind P, Rossing P, Tarnow L, Toft H, Parving J, 39. Ho PM, Rumsfeld JS, Masoudi FA, McClure DL, Parving HH. Remission of nephrotic-range albu- Plomondon ME, Steiner JF, et al. Effect of medica- minuria in type 1 diabetic patients. Diabetes Care. tion nonadherence on hospitalization and mortal- 2001;24:1972–7. ity among patients with diabetes mellitus. Arch Intern Med. 2006;166:1836–41. 30. Hovind P, Tarnow L, Rossing P, Carstensen B, Parving HH. Improved survival in patients obtain- 40. Wild H. The economic rationale for adherence in ing remission of nephrotic range albuminuria in the treatment of type 2 diabetes mellitus. Am J diabetic nephropathy. Kidney Int. 2004;66:1180–6. Manag Care. 2012;18:S43–8. 31. Rossing K, Christensen PK, Hovind P, Parving HH. 41. Menzin J, Korn JR, Cohen J, et al. Relationship Remission of nephrotic-range albuminuria reduces between glycemic control and diabetes-related risk of end-stage renal disease and improves survival hospital costs in patients with type 1 or type 2 in type 2 diabetic patients. Diabetologia. diabetes mellitus. JMCP. 2010;16:264–75. 2005;48:2241–7. 42. Gandra SR, Lawrence LW, Parasuraman BM, Darin 32. Alwakeel JS, Isnani AC, Alsuwaida A, et al. Factors RM, Sherman JJ, Wall JL. Total and component affecting the progression of diabetic nephropathy health care costs in a non-Medicare HMO popula- and its complications: a single-center experience in tion of patients with and without type 2 diabetes Saudi Arabia. Ann Saudi Med. 2011;31:236–42. and with and without macrovascular disease. JMCP. 2006;12:546–54. 33. Ohkubo Y, Kishikawa H, Araki E, et al. Intensive insulin therapy prevents the progression of diabetic 43. Lipska KJ, Warton EM, Huang ES, et al. HbA1c and microvascular complications in Japanese patients risk of severe hypoglycemia in type 2 diabetes: the with non-insulin-dependent diabetes mellitus: a diabetes and aging study. Diabetes Care. randomized prospective 6-year study. Diabetes Res 2013;36:3535–42. Clin Pract. 1995;28:103–17. 44. Zhao W, Katzmarzyk PT, Horswell R, Wang Y, 34. Ismail-Beigi F, Craven T, Banerji MA, et al. Effect of Johnson J, Hu G. HbA1c and coronary heart disease intensive treatment of hyperglycaemia on risk among diabetic patients. Diabetes Care. microvascular outcomes in type 2 diabetes: an 2014;37:428–35. analysis of the ACCORD randomised trial. Lancet. 2010;376:419–30. 45. Selby JV, Karter AJ, Ackerson LM, Ferrara A, Liu J. Developing a prediction rule from automated clin- 35. Callaghan BC, Little AA, Feldman EL, Hughes RA. ical databases to identify high-risk patients in a Enhanced glucose control for preventing and large population with diabetes. Diabetes Care. treating diabetic neuropathy. Cochrane Database 2001;24:1547–55. Syst Rev. 2012;6:CD007543. 46. Rosenzweig JL, Weinger K, Poirier-Solomon L, 36. Banerji MA, Dunn JD. Impact of glycemic control Rushton M. Use of a disease severity index for on healthcare resource utilization and costs of type evaluation of healthcare costs and management of 2 diabetes: current and future pharmacologic comorbidities of patients with diabetes mellitus. approaches to improving outcomes. Am Health Am J Manag Care. 2002;8:950–8. Drug Benefits. 2013;6:382–92. 47. Young BA, Lin E, Von Korff M, et al. Diabetes 37. Stark Casagrande S, Fradkin JE, Saydah SH, Rust KF, complications severity index and risk of mortality, Cowie CC. The prevalence of meeting A1C, blood hospitalization, and healthcare utilization. Am J pressure, and LDL goals among people with dia- Manag Care. 2008;14:15–23. betes, 1988–2010. Diabetes Care. 2013;36:2271–9. 48. Porrini E, Ruggenenti P, Mogensen CE, et al. Non- 38. Cramer JA. A systematic review of adherence with proteinuric pathways in loss of renal function in medications for diabetes. Diabetes Care. patients with type 2 diabetes. Lancet Diabetes 2004;27:1218–24. Endocrinol. 2015;3:382–91.

Journal

Diabetes TherapySpringer Journals

Published: Mar 29, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off