Neurol Ther (2018) 7:87–101 https://doi.org/10.1007/s40120-018-0097-9 ORIGINAL RESEARCH Positive Effectiveness of Tafamidis in Delaying Disease Progression in Transthyretin Familial Amyloid Polyneuropathy up to 2 Years: An Analysis from the Transthyretin Amyloidosis Outcomes Survey (THAOS) . . . . Rajiv Mundayat Michelle Stewart Jose Alvir Sarah Short . . . Moh-Lim Ong Denis Keohane Denise Rill Marla B. Sultan Received: November 29, 2017 / Published online: April 9, 2018 The Author(s) 2018 Limbs (p \ 0.001) and its subscales (p \ 0.023) ABSTRACT compared with untreated subjects. There was signiﬁcantly less deterioration among tafamidis- Introduction: The effectiveness of tafamidis for treated subjects compared with untreated sub- the treatment of transthyretin familial amyloid jects on the Norfolk Quality of Life scale polyneuropathy (TTR-FAP) was evaluated using (p \ 0.001). There were no signiﬁcant differences data from the Transthyretin Amyloidosis Out- observed in functional (assessed by Karnofsky comes Survey (THAOS) registry. Performance Status Scale score) or nutritional Methods: Subjects receiving tafamidis (n = 252) (assessed by modiﬁed body mass index) status were compared with untreated subjects in a between the treated and untreated groups. The non-randomized, matched cohort analysis. primary model which examined survival from Subjects were matched with up to four baseline using the matched cohort was not able untreated controls by genetic mutation, region to yield estimates of the hazard ratio, as there of birth, and mean treatment propensity score. were no deaths in the tafamidis-treated subjects. Results: The matched, treated sample consisted Conclusion: These ﬁndings support the results predominantly of subjects with the Val30Met from clinical trials and strengthen evidence of genotype (92.5%), from Portugal, and with a the effectiveness of tafamidis beyond conven- mean age of 40.4 years. Over the course of the tional clinical trials. 2-year follow-up period, subjects treated with Trial Registration: ClinicalTrials.gov: NCT006 tafamidis showed signiﬁcantly less deterioration on the Neuropathy Impairment Score for Lower Funding: Pﬁzer. Enhanced content To view enhanced content for this Keywords: Familial amyloid polyneuropathy; article go to https://doi.org/10.6084/m9.ﬁgshare. 6015644. Matched cohort analysis; Transthyretin; Tafamidis R. Mundayat (&) J. Alvir M.-L. Ong D. Keohane D. Rill M. B. Sultan Pﬁzer, New York, NY, USA INTRODUCTION e-mail: Rajiv.Mundayat@pﬁzer.com Transthyretin familial amyloid polyneuropathy M. Stewart Pﬁzer, Groton, CT, USA (TTR-FAP) is a rare, life-threatening, autosomal- dominant illness caused by TTR gene mutations S. Short that result in the deposition of amyloid ﬁbrils in ICON plc, San Francisco, CA, USA 88 Neurol Ther (2018) 7:87–101 peripheral nerves and organs [1–3]. The disease using matching techniques to allow meaningful is predominantly characterized by debilitating comparison with an untreated control group. sensorimotor and autonomic polyneuropathy, but complications due to organ involvement MATERIALS AND METHODS are common . Of the more than 100 TTR genotypes identiﬁed, the most common is Study Design Val30Met [5–8]. Survival time from disease onset to death is The present analysis is a non-randomized, approximately 10 years on average [2, 3, 9], but this may vary by genotype and phenotype. For matched cohort analysis of longitudinal data, up to 2 years from baseline, among patients example, in a study in France, Val30Met patients with a later onset of disease (median enrolled in THAOS. Details regarding the basic design and methodology of THAOS (Clini- age 64 years), together with patients with the Ile107Val mutation, showed a more rapid and calTrials.gov: NCT 00628745) are described elsewhere . THAOS is an observational reg- severe disease progression than did more slowly istry for which any individual with a conﬁrmed progressing Portuguese Val30Met patients with an earlier onset of disease (median age 32 years) TTR mutation or wild-type ATTR amyloidosis is eligible, and includes some patients participat- and patients with the Ser77Tyr mutation . Liver transplantation is a recognized treatment ing in ongoing clinical trials . All study sites received ethical or institutional review board option for selected Val30Met patients in the early stages of TTR-FAP. However, transplanta- approval prior to subject enrollment, and each subject provided written informed consent. The tion is associated with the risk of mortality and the need for lifelong immunosuppressant ther- study followed the International Conference on Harmonisation Good Clinical Practice guideli- apy, and recent evidence suggests that survival rates are lower in patients with mutations other nes and the principles of the Declaration of Helsinki. than Val30Met [11, 12]. Tafamidis, a highly speciﬁc TTR-stabilizer, is the only approved medicine to delay progres- Study Population sion of TTR-FAP and has emerged as the new standard of care . While the efﬁcacy and This analysis included subjects with conﬁrmed safety of tafamidis in the treatment of TTR-FAP TTR mutations (subjects with wild-type ATTR have been demonstrated in clinical trials, little were excluded) who were symptomatic at is known about its real-world effectiveness enrollment or became symptomatic post [14–18]. In one recent observational study of enrollment, and who had baseline data with at approximately 3000 Val30Met patients from least one follow-up visit. Data were extracted Portuguese referral centers, both liver trans- from the THAOS database on January 6, 2015. plantation and treatment with tafamidis were Subjects with TTR mutations known to have a associated with a survival beneﬁt when com- predominantly cardiac phenotype (i.e., pared with untreated patients . Val122Ile, Ile68Leu, Leu111Met, and Thr60Ala) The Transthyretin Amyloidosis Outcomes or who received any treatment (other than Survey (THAOS) is the largest ongoing interna- tafamidis) intended to target the course of the tional disease registry collecting longitudinal disease (i.e., liver transplant or diﬂunisal) pre- data in patients with amyloid TTR (ATTR) enrollment were excluded. The analysis was amyloidosis . Comprising data from symp- limited to subjects in Coutinho stage 1 (as tomatic and asymptomatic patients who may be deﬁned by the modiﬁed Polyneuropathy Dis- either treated or untreated, THAOS offers a ability [mPND]) score, with subjects either valuable opportunity to study the real-world having sensory disturbances in their feet but effectiveness of treatments. The objective of the able to walk without difﬁculty [mPND score I], current study was to examine the effectiveness of tafamidis among patients enrolled in THAOS Neurol Ther (2018) 7:87–101 89 or having some difﬁculties walking but able to rated from 0 = no contractions to 5 = full range walk without aid [mPND score II] . of motion and maximum resistance], NCS-Sen- For patients receiving tafamidis, dosage rec- sory [range 0–124, with items rated as normal, ommendations (including dosage strength and decreased, or absent]) were calculated as the timing of the dosing) were based on the pre- sum of quantitative ratings for subjects’ neuro- scribing physician’s clinical judgment and the logic assessments of both upper and lower local approved product label, or the open-label limbs. The mapped Neuropathy Impairment study protocol for those patients receiving Score-Lower Limbs (NIS-LL, range 0 [normal] to tafamidis as part of their participation in a 88 [total impairment]) was calculated from a clinical study (the indicated dose of tafamidis is subset of neurologic assessment items for lower 20 mg once daily) . limbs and assesses function of the extremities Disease duration was calculated from the most affected early in disease progression . earliest onset date of any symptoms commonly Coutinho stage was derived using the mPND associated with TTR-FAP that the treating (mPND score I or II equating to Coutinho stage physician had ﬂagged as deﬁnitively or possibly 1; mPND score IIIa or IIIb equating to Coutinho attributable to the disease. An examination of stage 2), and time to progression from Coutinho disease duration at enrollment in all THAOS stage 1 to stage 2 (or higher) was determined subjects identiﬁed a subgroup with unusually . Total Quality of Life (TQoL) score (range long disease duration, raising the possibility - 4 to 136) was derived from the patient-re- that these unusually long disease duration val- ported Norfolk Quality of Life-Diabetic Neu- ues could distort the multiple imputation and ropathy questionnaire, with higher scores treatment propensity score models. Because reﬂecting poorer quality of life [23, 24]. The disease duration was an important factor in the Karnofsky Performance Status Scale score was matching process, subjects in the highest quar- used to quantify the subjects’ ability to perform tile (C 8.6 years) were excluded from the anal- normal daily life activities and their need for ysis due to these concerns and concerns that assistance (ranging from 0 [dead] to 100 [nor- there would be an insufﬁcient number of mal; no complaints]). appropriate matches for subjects who had been symptomatic for many years prior to enroll- Statistics ment in THAOS. Matching Methodology Outcome Measures The matching methodology was designed to ensure that the heterogeneity among TTR-FAP Subject data from the participating THAOS cases worldwide, including variability in muta- registry sites included medical, laboratory, tion, age of onset, phenotype, and course of neurologic, and quality-of-life assessments col- disease, was well balanced between the treated lected as part of the local standard of care . and untreated analysis populations. Matching The focus of the present analysis was disease was accomplished in a three-step process by progression, based on neurologic and quality- matching on mutation group, country of origin, of-life assessments, and mortality. Outcome and by mean treatment propensity score to assessments performed at regular clinic visits account for clinical status. were examined based on approximately Subjects were matched within mutation 6-month intervals (± 3 months). Using data groups, deﬁned as Val30Met and non-Val30- from the clinical neurologic examination, the Met. Thus, treated Val30Met subjects were Neurologic Composite Score (NCS; range 0–294, matched with untreated Val30Met subjects. with higher scores reﬂecting worsening neuro- Within the group of subjects with non-Val30- logic functioning) and subscale scores (NCS- Met mutations, it was not possible to match on Reﬂex [range 0–10, with items rated as present the exact mutation owing to low numbers; or absent], NCS-Motor [range 0–160, with items therefore, treated subjects were matched to 90 Neurol Ther (2018) 7:87–101 untreated subjects within the non-Val30Met subjects with incomplete data. Missing data mutation group. Although country of ancestry were imputed via multiple imputation using a may be the closest approximation of genetic Markov chain approach. The propensity score background, this information was not captured models included eight selected baseline clinical consistently; thus, data on country of birth, variables: TQoL; disease duration; Karnofsky which were available for [ 99% of subjects, score; the three NCS sub-scores (Reﬂex, Sensory, were used as the indicator of genetic and Motor); blood urea nitrogen (BUN); and background. modiﬁed body mass index (mBMI). Up to 101 Finally, propensity score estimation was imputations were performed, separately for used to balance the treated and untreated Val30Met and non-Val30Met populations, groups on disease characteristics (e.g., severity resulting in a corresponding number of fully and duration). Propensity scores are frequently imputed datasets. Each imputed dataset was employed to reduce indication bias associated entered into a logistic regression model of with treatment status in observational studies, treatment probability, as predicted by the same thus facilitating causal inference . Propen- set of eight baseline clinical variables. sity score matching is widely used in observa- The mean propensity score for each subject tional studies and, when appropriately applied, at baseline was used for matching in a nearest- has been shown to be an effective method neighbor approach . To enhance the quality [26, 27]. Indication bias occurs in real-world of matched sets of patients, propensity scores (i.e., non-randomized) scenarios, in which for the treated and untreated subjects were treatment decisions are made by clinicians required to be within 0.25 standard deviations based on disease severity and any number of (SD) on the logit scale (i.e., logit of the other important demographic and clinical propensity score) . Furthermore, multiple factors. The estimated treatment propensity baseline dates were considered for the untreated scores reﬂect the probability of receiving population (i.e., any THAOS visit occurring treatment based on observed baseline covari- after disease onset and otherwise meeting eli- ates and allow imbalances observed in the gibility criteria). This allowed for the matching study population to be addressed in the anal- of treated and untreated at a point in time ysis . where clinical status was most similar rather Baseline for treated subjects was deﬁned as than at an arbitrary point in time, such as the enrollment date (if tafamidis treatment enrollment. Finally, the quality of the matches started before enrollment) or the treatment start was maintained through the use of matching date (if tafamidis treatment started on or after ‘‘with replacement’’ of untreated subjects. enrollment). Baseline for untreated subjects was Untreated subjects were allowed to be used in deﬁned as any visit occurring after symptom more than one matched set, but not more than onset and not more than 3 months prior to once with a given treated subject. enrollment (the THAOS registry permits collec- tion of retrospective data). Owing to the obser- Analyses vational nature of the THAOS registry, clinical Descriptive statistics were calculated for demo- data are collected and reported as part of rou- graphic and baseline clinical characteristics. tine care. Given variability in practice patterns Treated subjects are presented at a given time and the inﬂuence of disease status on care point only if one or more of their matched decisions, data are assumed not to be missing at controls also had data available. Untreated random, that is, the probability of a data point subjects are presented at a given time point only being missing is related to other factors of if their matched treated subject also had data interest at both the site and patient levels. available. Values for matched controls within Missing baseline data were imputed for the each matched set are weighted. Treatment purpose of propensity score modeling to avoid effects were evaluated by repeated measures possible selection bias associated with excluding analyses with appropriate covariates (age at Neurol Ther (2018) 7:87–101 91 baseline, gender, duration of symptoms, treat- RESULTS ment propensity score, time and treatment-by- time interaction, and baseline values). Survival Population and time to progression from Coutinho stage 1 were analyzed using Cox proportional hazards As of January 6, 2015 (data extraction date), regression models. Subjects who received alter- 2535 subjects were enrolled in the THAOS reg- native disease-modifying treatment (i.e., liver istry; of these, 274 tafamidis-treated subjects and transplant or diﬂunisal) were censored at the 255 untreated controls met the inclusion criteria earlier of liver transplant date or the start of for the current analysis (Fig. 1). A total of 22 diﬂunisal treatment. For time-to-event analyses, treated and 88 untreated subjects were unable to age at baseline and gender were used as covari- be matched, per the pre-speciﬁed matching cri- ates for the primary analysis. teria. Prior to the matching, treated subjects had Secondary and tertiary models were used to higher disease burden at baseline than untreated more fully characterize the survival and time-to- subjects, as evidenced by higher NIS-LL and progression results. The secondary analysis TQoL scores. Box plots of treated and untreated included all subjects (unmatched) who were treatment propensity scores on a logit scale ever at Coutinho stage 1 during the THAOS before matching indicated overlap between the registry. As the start time was unrelated to the groups, and thus there were potential matches treatment start date, the treatment effect was for all treated subjects (Fig. 2a). estimated through use of a time-varying Each of the 252 treated subjects was matched covariate for tafamidis treatment. The tertiary with up to four untreated subjects, with a total analysis included the matched cohorts but was of 167 unique untreated subjects. In total, 208 limited to subjects with mPND score I (having (82.5%) treated subjects were each matched to sensory disturbances in their feet but able to four untreated subjects, 13 (5.2%) to three walk without difﬁculty) at baseline (i.e., untreated subjects, eight (3.2%) to two excluding subjects with mPND score II; some untreated subjects, and 23 (9.1%) to one difﬁculties walking but could walk without aid) untreated subject. This resulted in 910 unique to account for the truncated time to progression matched treated–untreated pairs. The majority in subjects with a higher mPND score. Covari- of pairs had an exact match on country of birth ates for the secondary analysis also included age (n = 762; 83.7%). Approximately half the pairs at baseline, genetic mutation, and birth region, matched exactly on gender (n = 451; 49.6%). while covariates for the tertiary analysis inclu- Box plots of treated and untreated treatment ded age at the start of Coutinho stage 1 and propensity scores on a logit scale after matching gender. are shown in Fig. 2b. The mean (SD) difference between propensity scores for the 910 matched Sensitivity Analysis pairs was 0.012 (0.013) logits. Assessment data for the period from the start of treatment to enrollment in the registry gener- Demographic and Baseline Characteristics ally were not available for subjects treated with tafamidis prior to enrolling in THAOS. Further- Baseline demographic characteristics were more, these subjects typically had a longer broadly similar between tafamidis-treated sub- overall duration of treatment than those who jects and matched controls, with a mean (SD) began receiving treatment at or after enroll- duration since disease onset of 3.18 (1.9) and ment. To evaluate the robustness of the model, 3.27 (1.3) years, respectively (Table 1). The given the uncertainty of the effect of these two median post-baseline follow-up time was factors, a sensitivity analysis was performed, 23.6 months for treated subjects and analyzing matched cohorts that excluded sub- 21.4 months for untreated subjects. In the jects who received tafamidis treatment prior to treated cohort, Val30Met (n = 233; 92.5%) was THAOS enrollment. the predominant TTR-FAP genotype, the gender 92 Neurol Ther (2018) 7:87–101 Fig. 1 Eligible study population. THAOS Transthyretin Amyloidosis Outcomes Survey, TTR transthyretin ratio was approximately equal (n = 131 male; At baseline, treated subjects had higher 52.0%), and the majority of subjects (n = 202; scores across all neurologic endpoints, indicat- 80.2%) were born in Portugal. Other reported ing greater impairment (Table 1). However, a countries of birth were Angola, Argentina, comparison of baseline values in the matched Austria, Brazil, France, Germany, Italy, Japan, sample analysis, which adjusted for baseline Sweden, Turkey, United States, and Venezuela. treatment propensity score, gender, age at Neurol Ther (2018) 7:87–101 93 Fig. 2 Treatment propensity scores for tafamidis-treated and matched control subjects, a before matching and b after matching Table 1 Demographic and clinical characteristics of treated patients and untreated matched controls at baseline Characteristic, mean (SD) Tafamidis n = 252 Matched controls n = 252 Age, years 40.36 (11.90) 40.85 (9.36) TTR disease duration, years 3.18 (1.87) 3.27 (1.28) Total time on tafamidis, months 29.11 (22.97) – Mapped NIS-LL 8.92 (9.66) 7.57 (7.29) NCS 21.34 (22.84) 19.18 (19.20) Reﬂex score 0.78 (1.50) 0.55 (0.87) Motor score 1.70 (6.56) 1.10 (2.07) Sensory score 19.33 (18.46) 17.45 (16.48) Norfolk TQoL score 23.44 (23.71) 22.26 (15.26) Karnofsky index 86.39 (6.92) 85.43 (7.77) mBMI, kg/m 9 g/L 1068.09 (248.44) 1049.71 (184.25) BUN, mg/dL 30.61 (9.71) 30.75 (6.30) The predominant mutation was Val30Met (92.5% of treated subjects; n = 233), and the majority of subjects were born in Portugal (80.2% of treated subjects; n = 202) BUN blood urea nitrogen, mBMI modiﬁed body mass index, NCS Neurologic Composite Score, NIS-LL Neuropathy Impairment Score in the Lower Limbs, SD standard deviation, TQoL Total Quality of Life, TTR transthyretin For matched controls, descriptive statistics were calculated using means within matched sets. The unit of analysis (n) is the matched set rather than the individual 94 Neurol Ther (2018) 7:87–101 baseline, and duration of TTR-FAP–related Lower rates of progression in tafamidis-treated symptoms at baseline, found no statistically subjects compared with untreated subjects were signiﬁcant differences for the mapped NIS-LL, statistically signiﬁcant for the NIS-LL, NCS, and NCS-Motor and NCS-Sensory scores, TQoL, each of the NCS subscale scores (p B 0.023 for mBMI, and BUN. There were statistically sig- each; Table 2).There was also a signiﬁcantly niﬁcant differences between the treated and lower rate of worsening in quality of life, as untreated groups for NCS-Reﬂex (p = 0.018) and assessed by TQoL (p \ 0.001). the Karnofsky score (p = 0.040), with scores The trends in the estimated rate of change indicating that treated subjects had diminished between tafamidis-treated and untreated sub- reﬂexes, but better daily functioning, compared jects for daily functioning (Karnofsky score) and with untreated subjects at baseline. nutritional status (assessed using mBMI) were not signiﬁcant (Table 2). The estimated rate of change per year in BUN demonstrated a statis- Change in Neurologic, Quality of Life, tically signiﬁcant lower rate of worsening in Functional, and Nutritional Status tafamidis-treated subjects (p = 0.004). While BUN is not considered a direct measure of ATTR Over the course of 2 years, tafamidis-treated amyloidosis disease, it was included in this subjects had a signiﬁcantly lower rate of disease analysis, as it correlated with disease severity at progression than untreated subjects, as baseline. demonstrated by measures of neurologic func- tion and quality of life (Table 2 and Fig. 3). Table 2 Comparison of estimated rate of change per year of clinical characteristics by treatment status Clinical outcome Tafamidis Matched controls Tafamidis—matched controls p value (Est – SE) (Est – SE) (Est – SE) Mapped NIS-LL 0.32 ± 0.49 2.94 ± 0.37 - 2.61 ± 0.46 \0.001 NCS 0.75 ± 1.03 5.82 ± 0.75 - 5.07 ± 1.10 \0.001 Reﬂex score 0.17 ± 0.06 0.33 ± 0.04 - 0.16 ± 0.07 0.023 Motor score 0.32 ± 0.42 2.65 ± 0.30 - 2.33 ± 0.41 \0.001 Sensory score 0.53 ± 0.56 2.37 ± 0.38 - 1.84 ± 0.64 0.004 Norfolk TQoL - 1.56 ± 0.78 1.67 ± 0.46 - 3.23 ± 0.91 \0.001 score Karnofsky index - 0.43 ± 0.31 0.09 ± 0.17 - 0.53 ± 0.35 0.13 mBMI, kg/ 18.21 ± 4.81 18.57 ± 3.02 - 0.35 ± 5.68 0.95 m 9 g/L BUN, mg/dL 0.94 ± 0.91 4.03 ± 0.54 - 3.10 ± 1.06 0.004 Estimates were calculated from models of tafamidis treatment effect on outcome, adjusted for the following covariates: baseline value, follow-up time, baseline treatment propensity score, gender, age at baseline, duration of ATTR-related symptoms at baseline, and a treatment-by-time interaction. The number of treated subjects (numbers were matched with controls) at baseline, 12 months, and 24 months, respectively, for each measure were as follows: Mapped NIS-LL 203, 173, 51; NCS 175, 146, 43; Reﬂex score 238, 200, 73; Motor score 227, 192, 68; Sensory score 178, 150, 44; Norfolk TQoL score 189, 176, 58; Karnofsky index 243, 206, 74; mBMI 170, 130, 47; BUN 210, 171, 56 BUN blood urea nitrogen, Est estimate, mBMI modiﬁed body mass index, NCS Neurologic Composite Score, NIS-LL Neuropathy Impairment Score in the Lower Limbs, SE standard error, TQoL Total Quality of Life Neurol Ther (2018) 7:87–101 95 Fig. 3 Slope estimates from repeated measures analysis model. mBMI modiﬁed body mass index, NCS Neurologic Composite Score, NIS-LL Neuropathy Impairment Score-Lower Limbs, TQoL Total Quality of Life Sensitivity Analyses NCS-Motor subscale, NCS-Sensory subscale, TQoL, and BUN (p B 0.029 for each), but there was no signiﬁcant difference in the Karnofsky The sensitivity analyses excluded subjects who score or mBMI. In contrast to the full cohort, initiated tafamidis treatment prior to enroll- there was no signiﬁcant difference in the NCS- ment in THAOS (n = 53). Of the remaining 199 Reﬂex subscale in the sensitivity analysis. A subjects, matches were found for 196 subjects. total of 44 subjects in the full population had Demographics and clinical characteristics at also participated in a tafamidis clinical trial and baseline were similar to those of the overall were among those excluded in the sensitivity population, with the exception that the mean analysis. total time (SD) on tafamidis (n = 196) was 19.01 (7.05) months, compared with 29.11 (22.97) for the overall population. The results of the sen- Time to Progression and Survival sitivity analyses were generally similar to those for the full population (Table 3). There were None of the primary, secondary, or tertiary statistically signiﬁcant differences between the statistical models examining time to disease treated and untreated groups for NIS-LL, NCS, 96 Neurol Ther (2018) 7:87–101 Table 3 Comparison of estimated rate of change per year in clinical characteristics by treatment status, excluding subjects who initiated tafamidis treatment prior to enrollment in THAOS Clinical outcome Tafamidis Matched controls Tafamidis—matched controls p value (Est – SE) (Est – SE) (Est – SE) Mapped NIS-LL 0.25 ± 0.56 2.80 ± 0.44 - 2.55 ± 0.50 \0.001 NCS 0.51 ± 1.14 4.88 ± 0.84 - 4.37 ± 1.16 \0.001 Reﬂex score 0.21 ± 0.06 0.19 ± 0.03 0.03 ± 0.07 0.71 Motor score 0.40 ± 0.46 2.47 ± 0.33 - 2.07 ± 0.44 \0.001 Sensory score 0.20 ± 0.69 1.88 ± 0.46 - 1.67 ± 0.77 0.029 Norfolk TQoL - 2.75 ± 0.92 1.60 ± 0.54 - 4.35 ± 1.05 \0.001 score Karnofsky index - 0.47 ± 0.34 0.18 ± 0.19 - 0.65 ± 0.38 0.085 mBMI, kg/ 20.54 ± 5.18 20.57 ± 3.13 - 0.03 ± 6.05 [0.99 m 9 g/L BUN, mg/dL 0.44 ± 0.85 3.27 ± 0.49 - 2.83 ± 0.98 0.004 Estimates were calculated from models of tafamidis treatment effect on outcome, adjusted for the baseline covariate BUN blood urea nitrogen, Est estimate, mBMI modiﬁed body mass index, NCS Neurologic Composite Score, NIS-LL Neuropathy Impairment Score in the Lower Limbs, SE standard error, TQoL Total Quality of Life progression from Coutinho stage 1 to 2 over the treated subjects, the primary model that exam- course of 2 years demonstrated a statistically ined survival from baseline using the matched signiﬁcant difference between tafamidis-treated cohort was unable to yield estimates of the and untreated subjects. However, it should be hazard ratio. A secondary analysis of all subjects noted that only a small number of subjects (unmatched) examining survival from symp- progressed from Coutinho stage 1 (sensory dis- tom onset, using a time-varying treatment turbances in feet but able to walk without dif- covariate and a delayed entry model without ﬁculty, or having difﬁculties walking but able to regard to matching, yielded a statistically sig- walk without aid) to stage 2 (able to walk with niﬁcant hazard ratio of 3.95 (95% conﬁdence cane[s] or crutch[es]) during the observation interval: 1.54 to 10.14) for untreated versus period. All subjects were at stage 1 at baseline, treated subjects (p = 0.0042). Results of the and 98.6% of the treated and 97.3% of the sensitivity analyses for survival, excluding sub- untreated subjects remained in stage 1 through jects treated with tafamidis prior to baseline, the 24 months. showed a similar trend to that seen for the full Seven deaths were observed; all were in the cohort. untreated group. The primary causes of death were as follows: sudden death due to cardiac DISCUSSION arrhythmia (two subjects), cardiac arrest (one subject), pneumonia (one subject), unknown In this non-randomized, matched cohort anal- cause (one subject), and post-organ transplant ysis of data from THAOS, the largest observa- complications (two subjects; these subjects were tional disease registry of TTR-FAP patients, censored from the analysis at the date they tafamidis treatment resulted in signiﬁcantly less received their liver transplantation). As there neurologic disease progression (based on NCS were no deaths reported for the tafamidis- and NIS-LL) and less deterioration in quality of Neurol Ther (2018) 7:87–101 97 life (TQoL) compared with untreated matched notably late age at disease onset (59 years) and controls. No effect on nutritional status or the majority with a non-Val30Met mutation, functional impairment was observed. The only one-third of tafamidis-treated patients did results from this large international study of not show signiﬁcant disease progression (de- predominantly Val30Met subjects, but includ- ﬁned as an increase in NIS-LL of\ 2 points from ing non-Val30Met subjects, extend the efﬁcacy baseline), although this appeared to be inde- of tafamidis observed in clinical trials [14–18]to pendent of mutation type and disease stage real-world clinical settings. . The authors also noted that disease pro- No difference was observed between treated gression was observed primarily in the ﬁrst and untreated subjects in the time to disease 6 months of treatment . Another study in 46 progression (beyond Coutinho stage 1–2 or French patients, also with a late age at disease higher) in the present study. This is most likely onset (59 years) and majority non-Val30Met, due to the low number of subjects who pro- suggested that tafamidis slowed disease pro- gressed beyond Coutinho stage 1 during the gression in only some patients (approximately course of available follow-up. The median fol- one-third) after 3 years of follow-up . The low-up time was 23.6 months for treated sub- change in NIS-LL in both the tafamidis and jects and 21.4 months for untreated subjects; matched control groups in this analysis was less stage 1 is estimated to last 5.6 (± 2.8) years . pronounced than in these past studies in which The primary survival analysis did not yield the change in NIS-LL with tafamidis at 1 year hazard ratios due to a lack of deaths among was 4  and 5.9 . This difference may be treated subjects, but a secondary analysis that related to the differences in patient popula- included all subjects without regard to match- tions, especially disease severity at baseline and ing found that untreated subjects had a greater the response to tafamidis, with higher neuro- risk of death than did tafamidis-treated subjects. logical severity at baseline having been shown The results of the analysis in this study to be predictive of greater neurological pro- population showing that tafamidis delayed dis- gression over time regardless of treatment . ease progression in real-world clinical settings The difference may also be related to differences are consistent with data from tafamidis clinical in disease course with some genotypes and trials [14–16, 18, 30]. Together, these data pro- phenotypes . Past clinical trials have vide further support for early intervention with demonstrated signiﬁcant improvements in BMI tafamidis when treating patients with TTR-FAP. with tafamidis [15, 16]. This improvement was These results are also consistent with a phase 2 not seen at most time points in these recent open-label trial in 21 patients with non-Val30- uncontrolled studies, although a higher BMI or Met mutations that showed stabilization of TTR mBMI at baseline was associated with less dis- after 12 months of treatment with tafamidis ease progression with treatment [31, 32]. There , and with a real-world study using registry was no signiﬁcant improvement in mBMI in data from Portuguese TTR-FAP referral centers this analysis, which may have been related to that demonstrated the effectiveness of tafamidis the large amount of missing mBMI data at later for long-term survival in 373 Val30Met patients time points (see Table 2). . In the real-world study, estimated median Rare disease registries offer an important survival time from disease onset was about opportunity to study the effectiveness of treat- 12 years for the untreated group and about ments in real-world clinical settings and provide 25 years in the liver transplantation group, a practical approach for ensuring adequate sta- while median survival time was not reached in tistical power and long-term follow-up . the tafamidis group . The tafamidis-treated Careful matching between treated and patients had 5- and 10-year survival rates of 99.7 untreated cohorts is essential to facilitate causal and 94.6%, respectively . inference. The propensity scoring technique Other recent uncontrolled studies in differ- used in the present analysis to address variabil- ent patient groups have demonstrated mixed ity in clinical status has been successfully results. In one study in 61 Italian patients with a implemented to evaluate the effectiveness of 98 Neurol Ther (2018) 7:87–101 treatments in a variety of disease states [35–40], that tafamidis both delayed neurologic disease including rare diseases . In the current progression and maintained quality of life study, there was substantial overlap between compared with untreated matched controls. the two cohorts on treatment propensity scores, These ﬁndings conﬁrm previous results and indicating comparability between the treated provide evidence for the effectiveness of tafa- and untreated groups. Additionally, all resulting midis outside conventional clinical trials. pairs matched exactly on genotype group (Val30Met or non-Val30Met), and the majority were matched exactly on region of birth. This ACKNOWLEDGEMENTS close matching is fundamentally important in analyses using data from observational studies We thank all THAOS patients and investigators and strengthens the interpretation of the for their important contributions to this study. delayed disease progression observed in the We thank the THAOS Scientiﬁc Board (thaos.- tafamidis group compared with untreated net/THAOS/ScientiﬁcBoard.cfm), including Dr. controls. Claudio Rapezzi and Dr. Marcia Waddington- As with any observational study, there are Cruz as Scientiﬁc Board Chairs, for their part- potential sources of bias associated with subject nership and guidance on this study. selection; in this study, efforts were taken to minimize this bias by employing established, Funding. The Transthyretin Amyloidosis robust, multi-step matching techniques. Speci- Outcomes Survey (THAOS) and the analyses ﬁc limitations associated with THAOS as a dis- and the Article Processing Charges were spon- ease registry are described elsewhere . The sored by Pﬁzer. All authors had full access to all majority of treated subjects in this analysis were of the data in this study and take complete from Portugal (80.2%), were relatively young responsibility for the integrity of the data and (with a mean age of 40.4 years), and had the accuracy of the data analysis. Val30Met mutation (92.5%). This broadly reﬂects the population enrolled in THAOS who Medical Writing, Editorial, and Other have been treated with tafamidis. The matching Assistance. Medical writing support was pro- analysis used in this study requires a sufﬁciently vided by Mary Kunjappu, PhD, and Joshua Fink, large number of subjects for assessment, so it PhD, of Engage Scientiﬁc Solutions and was was not possible to perform comparisons funded by Pﬁzer. between different subject sub-groups to deter- mine whether there were any differences in Authorship. All named authors meet the their response to treatment. Therefore, these International Committee of Medical Journal ﬁndings may not be able to be extrapolated to Editors (ICMJE) criteria for authorship for this subjects with non-Val30Met mutations or late- manuscript, take responsibility for the integrity onset Val30Met. This may be particularly rele- of the work as a whole, and have given ﬁnal vant, as some past studies have suggested that approval for the version to be published. tafamidis may be less effective in older patients with more advanced disease [32, 42]. Disclosures. Rajiv Mundayat is an employee of Pﬁzer and hold stock and/or stock options. Michelle Stewart is an employee of Pﬁzer and CONCLUSION hold stock and/or stock options. Jose Alvir is an employee of Pﬁzer and hold stock and/or stock This is the largest international study to com- options. Moh-Lim Ong is an employee of Pﬁzer prehensively assess the effectiveness of tafami- and hold stock and/or stock options. Denis dis, using multiple clinical outcome measures, Keohane is an employee of Pﬁzer and hold stock under real-world conditions in a predominantly and/or stock options. Denise Rill is an employee Val30Met population which also included non- of Pﬁzer and hold stock and/or stock options. Val30Met subjects. These ﬁndings demonstrate Neurol Ther (2018) 7:87–101 99 7. Coelho T, Maurer MS, Suhr OB. THAOS—The Marla B. Sultan is an employee of Pﬁzer and transthyretin amyloidosis outcomes survey: initial hold stock and/or stock options. Sarah Short is report on clinical manifestations in patients with an employee of ICON plc, who were paid con- hereditary and wild-type transthyretin amyloidosis. sultants to Pﬁzer in connection with the devel- Curr Med Res Opin. 2013;29:63–76. opment of this work. Denis Keohane is a THAOS 8. Connors LH, Lim A, Prokaeva T, Roskens VA, Scientiﬁc Board member. Costello CE. Tabulation of human transthyretin (TTR) variants, 2003. Amyloid. 2003;10:160–84. Compliance with Ethics Guidelines. All procedures followed were in accordance with 9. Plante-Bordeneuve V, Lalu T, Misrahi M, et al. Genotypic-phenotypic variations in a series of 65 the ethical standards of the responsible com- patients with familial amyloid polyneuropathy. mittee on human experimentation (institu- Neurology. 1998;51:708–14. tional and national) and with the Helsinki Declaration of 1964, as revised in 2013. 10. Mariani LL, Lozeron P, Theaudin M, et al. Geno- type-phenotype correlation and course of trans- Informed consent was obtained from all thyretin familial amyloid polyneuropathies in patients for inclusion in the study. France. Ann Neurol. 2015;78:901–16. Open Access. This article is distributed 11. Ericzon BG, Wilczek HE, Larsson M, et al. Liver transplantation for hereditary transthyretin amy- under the terms of the Creative Commons loidosis: after 20 years still the best therapeutic Attribution-NonCommercial 4.0 International alternative? Transplantation. 2015;99:1847–54. License (http://creativecommons.org/licenses/ by-nc/4.0/), which permits any non- 12. Suhr OB, Larsson M, Ericzon BG, Wilczek HE. Sur- vival after transplantation in patients with muta- commercial use, distribution, and reproduction tions other than Val30Met: extracts from the FAP in any medium, provided you give appropriate world transplant registry. Transplantation. credit to the original author(s) and the source, 2016;100:373–81. provide a link to the Creative Commons license, 13. Adams D, Suhr OB, Hund E, et al. First European and indicate if changes were made. consensus for diagnosis, management, and treat- ment of transthyretin familial amyloid polyneu- ropathy. Curr Opin Neurol. 2016;29(Suppl 1):S14–26. REFERENCES 14. Ando Y, Yoshiki S, Konen O, et al. Effects of tafa- midis treatment on transthyretin (TTR) stabiliza- 1. Falk RH, Comenzo RL, Skinner M. The systemic tion, efﬁcacy, and safety in Japanese patients with amyloidoses. N Engl J Med. 1997;337:898–909. familial amyloid polyneuropathy (TTR-FAP) with Val30Met and non-Val30Met: a phase III, open-la- 2. Plante-Bordeneuve V, Said G. Familial amyloid bel study. J Neurol Sci. 2016;362:266–71. polyneuropathy. Lancet Neurol. 2011;10:1086–97. 15. Coelho T, Maia LF, da Silva AM, et al. Long-term 3. Ando Y, Coelho T, Berk JL, et al. Guideline of effects of tafamidis for the treatment of transthyr- transthyretin-related hereditary amyloidosis for etin familial amyloid polyneuropathy. J Neurol. clinicians. Orphanet J Rare Dis. 2013;8:31. 2013;260:2802–14. 4. Plante-Bordeneuve V. Update in the diagnosis and 16. Coelho T, Maia LF, Martins da Silva A, et al. Tafa- management of transthyretin familial amyloid midis for transthyretin familial amyloid polyneu- polyneuropathy. J Neurol. 2014;261:1227–33. ropathy: a randomized, controlled trial. Neurology. 2012;79:785–92. 5. Benson MD, Kincaid JC. The molecular biology and clinical features of amyloid neuropathy. Muscle 17. Coelho T, Merlini G, Bulawa CE, et al. Mechanism Nerve. 2007;36:411–23. of action and clinical application of tafamidis in hereditary transthyretin amyloidosis. Neurol Ther. 6. Rowczenio DM, Noor I, Gillmore JD, et al. Online 2016;5:1–25. registry for mutations in hereditary amyloidosis including nomenclature recommendations. Hum 18. Merlini G, Plante-Bordeneuve V, Judge DP, et al. Mutat. 2014;35:E2403–12. Effects of tafamidis on transthyretin stabilization and clinical outcomes in patients with non- 100 Neurol Ther (2018) 7:87–101 Val30Met transthyretin amyloidosis. J Cardiovasc with tafamidis provides long-term (5.5-year) delay Transl Res. 2013;6:1011–20. of neurologic progression in transthyretin heredi- tary amyloid polyneuropathy. Amyloid. 19. Coelho T, Ineˆs M, Conceic¸a˜o I, Saramago P, de 2016;23:178–83. Carvalho M, Costa J. Long-term survival of 3000 transthyretin familial amyloid polyneuropathy 31. Cortese A, Vita G, Luigetti M, et al. Monitoring patients over a century. Int J Clin Neurosci Mental effectiveness and safety of Tafamidis in transthyr- Health. 2016;3(Suppl. 1):P126. etin amyloidosis in Italy: a longitudinal multicenter study in a non-endemic area. J Neurol. 20. Plante-Bordeneuve V, Suhr OB, Maurer MS, White 2016;263:916–24. B, Grogan DR, Coelho T. The transthyretin amy- loidosis outcomes survey (THAOS) registry: design 32. Plante-Bordeneuve V, Gorram F, Salhi H, et al. and methodology. Curr Med Res Opin. Long-term treatment of transthyretin familial 2013;29:77–84. amyloid polyneuropathy with tafamidis: a clinical and neurophysiological study. J Neurol. 21. Coutinho P, de Martins SA, Lopes LJ, Resende BA. 2017;264:268–76. Forty years of experience with type I amyloid neu- ropathy. Review of 483 cases. In: Glenner G, Costa 33. Gundapaneni BK, Sultan MB, Keohane DJ, Schwartz P, de Freitas A, editors. Amyloid and Amyloidosis. JH. Tafamidis delays neurological progression Amsterdam: Execerpta Medica; 1980. p. 88–98. comparably across Val30Met and non-Val30Met genotypes in transthyretin familial amyloid 22. Pﬁzer Ltd. Vyndaqel Summary of Product Char- polyneuropathy. Eur J Neurol. 2018;25:464–8. acteristics. Sandwich, UK. http://www.ema.europa. eu/docs/en_GB/document_library/EPAR_-_Product_ 34. Gagne JJ, Thompson L, O’Keefe K, Kesselheim AS. Information/human/002294/WC500117862.pdf. Innovative research methods for studying treat- Accessed: March 10, 2017. ments for rare diseases: methodological review. BMJ. 2014;349:g6802. 23. Coelho T, Vinik A, Vinik EJ, Tripp T, Packman J, Grogan DR. Clinical measures in transthyretin 35. Lee J, Ko Y, Ahn S, et al. Comparison of US and CT familial amyloid polyneuropathy. Muscle Nerve. on the effect on negative appendectomy and 2017;55:323–32. appendiceal perforation in adolescents and adults: a post hoc analysis using propensity-score methods. 24. Vinik EJ, Vinik AI, Paulson JF, et al. Norfolk QOL-DN: J Clin Ultrasound. 2016;44:401–10. validation of a patient reported outcome measure in transthyretin familial amyloid polyneuropathy. 36. Lim C, Doussot A, Osseis M, et al. Primary tumor J Peripher Nerv Syst. 2014;19:104–14. versus liver-ﬁrst strategy in patients with stage IVA colorectal cancer: a propensity score analysis of 25. Rosenbaum PR, Rubin DB. The central role of the long-term outcomes and recurrence pattern. Ann propensity score in observational studies for causal Surg Oncol. 2016;23:3024–32. effects. Biometrika. 1983;70:41–55. 37. Lip GY, Keshishian A, Kamble S, et al. Real-world 26. Martens EP, Pestman WR, de Boer A, Belitser SV, comparison of major bleeding risk among non- Klungel OH. Systematic differences in treatment valvular atrial ﬁbrillation patients initiated on effect estimates between propensity score methods apixaban, dabigatran, rivaroxaban, or warfarin. A and logistic regression. Int J Epidemiol. propensity score matched analysis. Thromb Hae- 2008;37:1142–7. most. 2016;116:975–86. 27. Austin PC. A critical appraisal of propensity-score 38. Nduka CU, Stranges S, Bloomﬁeld GS, et al. A matching in the medical literature between 1996 plausible causal link between antiretroviral therapy and 2003. Stat Med. 2008;27:2037–49. and increased blood pressure in a sub-Saharan African setting: a propensity score-matched analy- 28. Rosenbaum PR, Rubin DB. Constructing a control sis. Int J Cardiol. 2016;220:400–7. group using multivariate matched sampling meth- ods that incorporate the Propensity Score. Am Stat. 39. Tian F, Hong XF, Wu WM, et al. Propensity score- 1985;39:33–8. matched analysis of robotic versus open surgical enucleation for small pancreatic neuroendocrine 29. Dehejia RH, Wahba S. Propensity Score-Matching tumours. Br J Surg. 2016;103:1358–64. Methods for Nonexperimental Causal Studies. Rev Econ Stat. 2002;84:151–61. 40. Wada T, Yasunaga H, Horiguchi H, et al. Ozagrel for patients with noncardioembolic ischemic stroke: a 30. Waddington Cruz M, Amass L, Keohane D, Propensity Score-matched analysis. J Stroke Cere- Schwartz J, Li H, Gundapaneni B. Early intervention brovasc Dis. 2016;25:2828–37. Neurol Ther (2018) 7:87–101 101 41. Grabowski GA, Kacena K, Cole JA, et al. Dose-re- 42. Lozeron P, Theaudin M, Mincheva Z, Ducot B, sponse relationships for enzyme replacement ther- Lacroix C, Adams D. Effect on disability and safety apy with imiglucerase/alglucerase in patients with of Tafamidis in late onset of Met30 transthyretin Gaucher disease type 1. Genet Med. familial amyloid polyneuropathy. Eur J Neurol. 2009;11:92–100. 2013;20:1539–45.
Neurology and Therapy – Springer Journals
Published: Apr 9, 2018