An examination of the heterogeneity in the pattern and association between rates of change in grip strength and global cognition in late life. A multivariate growth mixture modelling approach

An examination of the heterogeneity in the pattern and association between rates of change in... Abstract Background previous research has demonstrated how older adults exhibit different patterns of change in cognitive and physical functioning, suggesting differences in the underlying causal processes. Objective to (i) identify subgroups of older adults that best account for different patterns of longitudinal change in performance on global cognition and grip strength, (ii) examine the interrelationship between global cognition and grip strength trajectories within these subgroups and (iii) identify demographic and health-related markers of class membership. Methods multivariate growth mixture models (GMM) were used to identify groups of individuals with similar developmental trajectories of muscle strength measured by grip strength, and global cognition measured by Mini Mental State Examination (MMSE). Results GMM analyses indicated high, moderate and low functioning groups. Individuals in the high and moderate classes demonstrated better cognitive and physical functioning at the start of the study and less decline than those in the low functioning group. Notably, cognitive performance was related to physical functioning at study entry only among individuals in the low functioning group. Conclusion the study demonstrates the applicability of the multivariate GMM to achieve a better understanding of the heterogeneity of various aging related processes. longitudinal analysis, trajectories, growth mixture model, cognition, grip strength, older people Introduction Current aging research is largely focused on better understanding underlying aging processes with the aim to promote healthy aging. Over the past few decades, the aging of physical and cognitive processes has received extensive attention and has been found to be associated with overall health and increased risk of morbidity, mortality and dementia [1, 2]. Still, a greater focus on the individual differences in patterns and causes of changes in physical and cognitive processes over the life course is needed for predicting late life outcomes and for developing interventions to reduce the risk of negative health outcomes. Although there is consensus that cognitive performance generally declines with age, some evidence suggests that this decline is not normative [3–5]. Similarly, longitudinal studies demonstrate physical decline with age [6]. Despite these similar findings, few, if any, studies have examined the interindividual heterogeneity and the interrelationships in these trajectories. Developments in the analysis of longitudinal data have resulted in recent studies focused on better understanding heterogeneity in the trajectories of cognitive functioning. One such development, referred to as growth mixture models (GMM), allows for the identification of different classes of individuals who cluster together, thus identifying more than one trajectory within the population [7–9]. GMM provides information about the optimal number of classes and the characteristics of each class, including the mean intercept and slope, proportion of membership, and significant predictors of class membership. When GMM [e.g. 7, 8] have been used to explore various patterns of cognitive aging [3–5, 10, 11], they have typically suggested two (slow and rapid decliners) [10, 11] or three (low, moderate and rapid decliners) [3, 4] classes that best represent the underlying heterogeneity of cognitive trajectories. Most analyses of the association between cognitive and physical functioning have been based on modelling approaches (e.g. multivariate growth curve model) not taking into account the possibility for different clusters of individuals exhibiting distinct patterns of change and cross-domain interrelationships. It is quite strongly supported that cognitive and physical functioning are related in cross-sectional studies, and more weakly in longitudinal analyses [12]. How these two processes are interrelated, however, is largely unknown. More studies are needed that explore the relationship between physical and cognitive functioning within different classes of trajectories. Using a GMM with multiple outcomes (i.e. multivariate GMM) allows the exploration of not only the different classes of cognitive and physical functioning but also the association between the two outcomes within each of the identified classes. The purpose of this study was to (i) identify subgroups of older adults best describing longitudinal change in performance on the Mini Mental State Examination (MMSE) and grip strength, (ii) examine the associations between MMSE and grip strength trajectories within each subgroup and (iii) evaluate the role of cognitive and physical activity and demographic variables on differences in class membership. Grip strength was chosen as a biomarker of physical functioning given its sensitivity to physiological change [13] and its use as a valid marker of frailty [14]. Its simplicity to administer, portability and affordability make it easy to apply in numerous settings [13]. The MMSE was selected as the cognitive measure given that a systematic review by Clouston and colleagues (12) found that the MMSE was more strongly related with grip strength than domain-specific measures of cognition in older adults. Method Data Origins of Variance in the Oldest-Old (OCTO-Twin) The OCTO-Twin study included dizygotic and monozygotic twin pairs aged 80 years of age and older [15, 16]. The sample was selected from older adults in the population-based Swedish Twin Registry [17]. Five cycles of longitudinal data were collected at 2-year intervals. The full sample consisted of 702 individuals (351 same-sex pairs). Individuals were excluded due to missing data on the covariates. The sample used in the analysis consisted of 369 individuals ranging from 79.4 to 97.9 years of age. The rate of attrition ranged from 10.5% to 28.8% every two years, primarily due to death (see Table 1). Ethical approval for OCTO-Twin was obtained from the Ethics Committee at the Karolinska Institute in Stockholm and the Swedish Data Inspection Board. Informed consent was obtained from all participants. Descriptive statistics at each occasion for all variables are provided in Table 1. Table 1. Descriptive statistics and attrition rate for MMSE and grip strength at each wave of data collection Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Note. N = 369. *Sample size at each data wave and attrition rate every 2 years. Edu = years of education. Height in cm. Table 1. Descriptive statistics and attrition rate for MMSE and grip strength at each wave of data collection Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Note. N = 369. *Sample size at each data wave and attrition rate every 2 years. Edu = years of education. Height in cm. Measures MMSE The MMSE, administered in Swedish, was used to asses global cognitive functioning at each measurement occasion [18]. Grip strength A Martin vigorimeter (Elmed Inc., Addison, IL, USA; medium size bulb) to measure maximum force in pounds per square inch was used [19]. Participants performed the task three times per hand, with the final score being the maximum force exerted. Covariates Age, education and sex were included as covariates on the intercept and slope, and as markers of class membership. Height, level of physical activity, and engagement in cognitive activities were included as markers of class membership. Age, education and height were centred at their mean value. Sex was coded as 0 for males and 1 for females. Level of physical activity was measured by asking respondents: ‘Are you presently doing or have you previously done anything special to train your body or ‘keep your body fit’?’. Response options were ‘no’, ‘yes, to some extent’ or ‘yes, to a great extent’. These were used to derive an indicator as sedentary (0) and active (1). The cognitive activity measure was created by summing six items (e.g. playing games) dichotomously as ‘no’ (0) or ‘yes’ (1) and one item about training the memory and keeping the mind active as ‘no’ (0), ‘yes, to a certain degree’ (1), or ‘yes, definitely’ (2). Total scores ranged from 0 to 8 with higher scores signifying higher engagement in cognitive activity. Analysis Multivariate GMM [20] were employed to identify unobserved groups of individuals with similar linear developmental bivariate trajectories of grip strength and MMSE scores. Class membership was estimated using a multinomial logistic regression adjusted for the covariates. GMM provide estimates of the average MMSE and grip strength trajectories (intercept and slope) and variation for older adults in each latent class [8, 9, 21]. The number of classes is unknown a priori, therefore, we followed recommended procedures to select the best fitting model. First, models were fitted with increasing number of classes. To determine the optimal number of classes, Bayesian Information Criterion (BIC) values were compared from each model. Models with lower BIC values were considered better fitting [22]. Classification quality using entropy values (range from 0 to 1 with higher values meaning that individuals are better discriminated between classes) [23] and interpretability of classes was further considered. Mplus version 7.2 was used to run the GMMs [24]. The model was estimated by maximum likelihood with robust standard errors. Missing data were assumed to be missing at random. Results Number of classes A three-class model of cognitive and physical functioning was selected, given its lower BIC values (2 class = 18,411.00; 3 class = 18,254.08) and good classification of individuals in each class (entropy 2 class = 0.85; entropy 3 class = 0.81). Parameter estimates and standard errors for the three-class model are presented in Table 2. Table 2. Estimates and standard errors for the three-class growth mixture model of cognitive and physical functioning Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Note. N = 369. SE = standard errors. Class 1 = high functioning group. Class 2 = moderate functioning group. Class 3 = low functioning group. Table 2. Estimates and standard errors for the three-class growth mixture model of cognitive and physical functioning Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Note. N = 369. SE = standard errors. Class 1 = high functioning group. Class 2 = moderate functioning group. Class 3 = low functioning group. Grip strength and MMSE trajectories Class 1: The largest class was composed of 147 individuals (40% of sample) with good physical and cognitive functioning and slow decline. In this high functioning group, individuals had an average baseline MMSE score of 28.62 (SE = 0.17) and declined at a rate of 0.12 (SE = 0.048). See upper solid line in Figure 1. The average baseline grip strength score was 11.49 kPa (SE = 0.48) and rate of decline was 0.37 kPa (SE = 0.077) per year. See lower solid line in Figure 1. No associations were found between grip strength and MMSE. See intercept–intercept and slope–slope covariances in Table 2. Figure 1. View largeDownload slide Average trajectories of physical and cognitive functioning for the three classes. The three upper lines represent cognitive functioning and the three lower lines physical functioning. The solid lines represent high functioning individuals, the dotted lines represent moderate functioning individuals, and the dotdash lines represent low functioning individuals. Figure 1. View largeDownload slide Average trajectories of physical and cognitive functioning for the three classes. The three upper lines represent cognitive functioning and the three lower lines physical functioning. The solid lines represent high functioning individuals, the dotted lines represent moderate functioning individuals, and the dotdash lines represent low functioning individuals. Class 2: A class of 114 individuals (31% of sample) with slightly lower average baseline scores and steeper declining MMSE scores and grip strength was identified. In this moderate functioning group, the average baseline MMSE score was 26.84 (SE = 1.08) and average rate of decline was 0.66 (SE = 0.19). See upper dotted line in Figure 1. Average baseline grip strength was 10.54 kPa (SE = 0.49) and decline rate was 0.57 kPa (SD = 0.064). See lower dotted line in Figure 1. No associations were found between grip strength and MMSE. See intercept–intercept and slope–slope covariances in Table 2. Class 3: A low functioning group, composed of 108 individuals (29% of sample), demonstrated lower MMSE scores and weaker grip strength at the start of the study and steeper rate of decline compared to those in the other two classes. These individuals had an average baseline MMSE score of 23.71 (SE = 1.24) and decline rate of 2.10 kPa (SE = 0.18) per year (see upper dotdash line in Figure 1); and a baseline grip strength score of 8.46 kPa (SE = 0.82) and decline rate of 0.84 kPa (SE = 0.12) per year (see lower dotdash line in Figure 1). Individuals with stronger grip strength at the start of the study also demonstrated higher MMSE scores. No relationship was found between the rate of change in grip strength and the rate of change in MMSE See intercept–intercept and slope–slope covariances in Table 2. Covariates In the high functioning group (class 1), none of the covariates were associated with MMSE at study entry. Sex was associated with baseline grip strength, with females having lower grip strength scores. In the moderate functioning group (class 2), individuals who were older in age at baseline had lower MMSE scores. Female participants and individuals who were older had lower grip strength. In the low functioning group (class 3), individuals with higher educational attainment had higher baseline MMSE scores. Markers of class membership Participation in physical and cognitive activities was markers of class membership. Membership in class 2 was associated with engagement in physical activities (Estimate = 1.31*; SE = 0.53). Individuals who were engaged in physical activities had higher probability of being in the moderate functioning group compared to the low functioning group (class 3) than individuals who did not participate in physical activity. Membership in class 1 was associated with engagement in cognitive activities (Estimate = 0.68***; SE = 0.18). Individuals who reported higher levels of engagement in cognitive activities had a higher probability of being in the high functioning group compared to the low functioning group than individuals who engaged in fewer cognitive activities. Class membership was not influenced by age, height, education or sex. Discussion Our findings demonstrate the use of GMM to examine clusters of trajectories based on MMSE and grip strength scores and for profiling the characteristics of individuals within clusters. This is very important as many research questions focus on the dynamic relationship between selected developmental processes and how they may change over time. Our results show that age-related decline in physical and cognitive function was heterogeneous and not universal. We identified three distinct classes of trajectories for MMSE and grip strength in the oldest-old segment of the population. The high functioning group with good physical and cognitive performance showed minor decline and represents successful agers. The moderate functioning group who had lower cognitive and physical performance and steeper rate of decline than those in the high functioning group might be at greater risk of more substantial decline. On average, they entered the study with an MMSE score of 26.8 which is close to the cut-off score of 26 used to identify individuals at risk of mild cognitive impairment (MCI) and dementia [25]. The low functioning group, with the lowest cognitive and physical performance scores and steepest rate of decline, is likely those who are at greatest risk of also developing clinical dementia. Their average baseline score of 23.7 meets MCI criteria, [25] and almost meets the cut-off of 23 often used for clinical dementia [18]. Interestingly, we found no evidence for differences between the sexes in the low functioning group, suggesting that, for those individuals who are most frail, there are only minor or no differences between men and women. Our finding about the role of education suggests that less educated older adults in the low functioning group are at an increased risk of dementia and highlights the protective role of educational attainment in delaying the onset of the disease which is in line with the cognitive reserve hypothesis [26]. The importance of specific covariates in different classes would have been masked had we examined only one trajectory class, highlighting the importance of GMM. Participation in cognitive and physical activities was associated with an increased likelihood of being in the highly or moderately functioning group compared to the lower functioning group. This aligns with previous longitudinal research on the importance of physical and cognitive activity [27]. Unlike previous studies which focused on the importance of lifestyle activities on the intercept and rate of change (i.e. whether participation in activities can delay decline) of physical and cognitive functioning, the current study demonstrated that these activities increase the likelihood that an individual will follow the highly or moderately functioning trajectory over the course of their later life rather than following the poorly functioning trajectory. This finding suggests that it would be effective to implement interventions which entail increased participation in physical and cognitive activities. By identifying subgroups of individuals with similar changes in cognitive and physical functioning, we were able to better understand the relationship between these two dimensions of the aging process. Our finding that individuals with higher cognitive scores at baseline also had higher grip strength in only one class (low functioning group) suggests that the relationship between the two processes may be most relevant for those individuals who are especially frail and cognitively impaired. The fact that we only found a relationship in one class may help explain previous mixed results about the association between cognitive and physical functioning as other modelling approaches tend to assume that individual trajectories follow a single pattern of change. One limitation of this study is the use of the MMSE with its well-known ceiling effect and modest accuracy in identifying individuals with dementia [28]. Still, the MMSE is suitable for this study since it was not used as a diagnostic tool. Another limitation is that the sample size was too small to model the GMM separately for male and female participants knowing that older adults differ in grip strength norms [29]. Still, a recent bivariate study modelled both sexes together and found a relationship between grip strength and cognition [30]. To the best of our knowledge, this is the first study to examine the longitudinal relationship of subpopulations characterised by their physical and cognitive functioning trajectories. Our research demonstrates that cognitive and physical functioning is best described by more than one trajectory and that studies assuming a single population with an average developmental trajectory tend to miss important information about the class-specific relation between physical and cognitive functioning and class-specific markers. More research is needed to further identify risk factors to being in the low functioning group as these individuals might benefit the most from targeted interventions. Further research is also needed to examine whether similar subpopulations are found when including younger older adults and whether subpopulations of trajectories also exist for other biopsychosocial aging processes. Rather than assuming that aging individuals all follow a similar pattern of change, GMM allows for the modelling of subpopulations of trajectories which can have important implications for better tailored interventions. Key points Cognitive and physical functioning is best described by three distinct classes of trajectories. This study demonstrates the applicability of the multivariate growth mixture model (GMM), decline in physical and cognitive function is heterogeneous and not normative. Modelling subpopulations of trajectories is needed for better tailored interventions. Conflict of interest None. Funding Research reported in this publication was supported by the National Institute on Aging of the National Institutes of Health under Award Number P01AG043362, Integrative Analysis of Longitudinal Studies of Aging and Dementia. The OCTO-Twin study was funded by the National Institute on Aging of the National Institutes of Health (Grant number AG08861) and the Swedish Research Council for Health, Working Life and Welfare (Forte), and the Swedish Brain Power Consortium. The funding source played no role in the design, execution, analysis and interpretation of data, or writing of the study. References 1 Cooper R , Hardy R , Aihie Sayer A et al. . Age and gender differences in physical capability levels from mid-life onwards: the harmonisation and meta-analysis of data from eight UK cohort studies . PLoS One 2011 ; 6 : e27899 . doi:10.1371/journal.pone.0027899. Google Scholar CrossRef Search ADS PubMed 2 Brown JC , Harhay MO , Harhay MN . Walking cadence and mortality among community-dwelling older adults . J Gen Intern Med 2014 ; 29 : 1263 – 9 . Google Scholar CrossRef Search ADS PubMed 3 Terrera GM , Brayne C , Matthews F . One size fits all? Why we need more sophisticated analytical methods in the explanation of trajectories of cognition in older age and their potential risk factors . Int Psychogeriatr 2010 ; 22 : 291 – 9 . doi: http://dx.doi.org/10.1017/S1041610209990937; PubMed PMID: 223742514; 19906326. Google Scholar CrossRef Search ADS PubMed 4 Hayden KM , Reed BR , Manly JJ et al. . Cognitive decline in the elderly: an analysis of population heterogeneity . Age Ageing 2011 ; 40 : 684 – 9 . Google Scholar CrossRef Search ADS PubMed 5 Xie H , Mayo N , Koski L . Identifying and characterizing trajectories of cognitive change in older persons with mild cognitive impairment . Dement Geriatr Cogn Disord 2011 ; 31 : 165 – 72 . doi: http://dx.doi.org/10.1159/000323568; PubMed PMID: 859334228; 21346357. Google Scholar CrossRef Search ADS PubMed 6 Jerome GJ , Ko S-U , Kauffman D , Studenski SA , Ferrucci L , Simonsick EM . Gait characteristics associated with walking speed decline in older adults: results from the Baltimore longitudinal study of aging. Arch Gerontol Geriatr 2015 ; 60 : 239 – 43 . doi:10.1016/j.archger.2015.01.007. Google Scholar CrossRef Search ADS PubMed 7 Liu M , Hancock G . Unrestricted mixture models for class identification in growth mixture modeling. Educ Psychol Meas 2014 ; 74 : 557 – 84 . doi:10.1177/0013164413519798. Google Scholar CrossRef Search ADS 8 Muthén B . Latent Variable Analysis: Growth Mixture Modeling and Related Techniques for Longitudinal Data. The SAGE Handbook of Quantitative Methodology for the Social Sciences . Thousand Oaks, CA : SAGE Publications, Inc , 2004 ; pp. 346 – 70 . 9 Muthén B , Shedden K . Finite mixture modeling with mixture outcomes using the EM algorithm . Biometrics 1999 ; 55 : 463 – 9 . doi:10.1111/j.0006-341×.1999.00463.x. Google Scholar CrossRef Search ADS PubMed 10 Proust C , Jacqmin-Gadda H . Estimation of linear mixed models with a mixture of distribution for the random effects . Comput Methods Programs Biomed 2005 ; 78 : 165 – 73 . doi:http://dx.doi.org/10.1016/j.cmpb.2004.12.004. Google Scholar CrossRef Search ADS PubMed 11 Small BJ , Backman L . Longitudinal trajectories of cognitive change in preclinical Alzheimer’s disease: a growth mixture modeling analysis. Cortex 2007 ; 43 : 826 – 34 . doi:10.1016/S0010-9452(08)70682-8. Google Scholar CrossRef Search ADS PubMed 12 Clouston SA , Brewster P , Kuh D et al. . The dynamic relationship between physical function and cognition in longitudinal aging cohorts . Epidemiol Rev 2013 ; 35 : 33 – 50 . Epub 2013/01/26. doi: 10.1093/epirev/mxs004; PubMed PMID: 23349427; PubMed Central PMCID: PMCPMC3578448. Google Scholar CrossRef Search ADS PubMed 13 Bohannon RW . Hand-grip dynamometry predicts future outcomes in aging adults . J Geriatr Phys Ther 2008 ; 31 : 3 – 10 . Epub 2008/05/21. PubMed PMID: 18489802. Google Scholar CrossRef Search ADS PubMed 14 Syddall H , Cooper C , Martin F , Briggs R , Aihie Sayer A . Is grip strength a useful single marker of frailty? Age Ageing 2003 ; 32 : 650 – 6 . Epub 2003/11/06. PubMed PMID: 14600007. Google Scholar CrossRef Search ADS PubMed 15 Johansson B , Hofer SM , Allaire JC et al. . Change in memory and cognitive functioning in the oldest-old: The effects of proximity to death in genetically related individuals over a six-year period . Psychol Aging 2004 ; 19 : 145 – 56 . Google Scholar CrossRef Search ADS PubMed 16 McClearn GE , Johansson B , Berg S et al. . Substantial genetic influence on cognitive abilities in twins 80 or more years old . Science 1997 ; 276 : 1560 – 3 . Google Scholar CrossRef Search ADS PubMed 17 Cederlöf R , Lorich U . The Swedish Twin Registry. In: Nance WE , Allen G , Parisi P , eds . Twin research: Biology and epidemiology . New York, NY : Alan R. Riss , 1978 ; pp. 189 – 95 . 18 Folstein MF , Folstein SE , McHugh PR . Mini-mental state’: a practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res 1975 ; 12 : 189 – 98 . Google Scholar CrossRef Search ADS PubMed 19 Sipers WM , Verdijk LB , Sipers SJ , Schols JM , van Loon LJ . The Martin vigorimeter represents a reliable and more practical tool than the Jamar dynamometer to assess handgrip strength in the geriatric patient . J Am Med Dir Assoc 2016 ; 17 : 466.e1 – 7 . Epub 2016/04/24. doi:10.1016/j.jamda.2016.02.026; PubMed PMID: 27107163. Google Scholar CrossRef Search ADS 20 Wu J , Witkiewitz K , McMahon RJ , Dodge KA . A parallel process growth mixture model of conduct problems and substance use with risky sexual behavior . Drug Alcohol Depend 2010 ; 111 : 207 – 14 . doi: http://dx.doi.org/10.1016/j.drugalcdep.2010.04.013; PubMed PMID: 2010-20916-004. Google Scholar CrossRef Search ADS PubMed 21 Berlin KS , Parra GR , Williams NA . An introduction to latent variable mixture modeling (part 2): longitudinal latent class growth analysis and growth mixture models. J Pediatr Psychol 2014 ; 39 : 188 – 203 . doi:10.1093/jpepsy/jst085. Google Scholar CrossRef Search ADS PubMed 22 Nylund KL , Asparouhov T , Muthén BO . Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study . Struct Equ Modeling 2007 ; 14 : 535 – 69 . doi:10.1080/10705510701575396. Google Scholar CrossRef Search ADS 23 Celeux G , Soromenho G . An entropy criterion for assessing the number of clusters in a mixture model . J Classif 1996 ; 13 : 195 – 212 . doi:10.1007/BF01246098. Google Scholar CrossRef Search ADS 24 Muthén LK , Muthén BO . Mplus User’s Guide , 7th ed . Los Angeles, CA : Muthén & Muthén , 1998 -2015. 25 Nasreddine ZS , Phillips NA , Bédirian V et al. . The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment . J Am Geriatr Soc 2005 ; 53 : 695 – 9 . doi:10.1111/j.1532-5415.2005.53221.x. Google Scholar CrossRef Search ADS PubMed 26 Terrera GM , Minett T , Brayne C , Matthews FE . Education associated with a delayed onset of terminal decline . Age Ageing 2014 ; 43 : 26 – 31 . doi: 10.1093/ageing/aft150; PubMed PMID: PMC3861340. Google Scholar CrossRef Search ADS PubMed 27 Lindwall M , Cimino CR , Gibbons LE et al. . Dynamic associations of change in physical activity and change in cognitive function: coordinated analyses of four longitudinal studies . J Aging Res 2012 ; 2012 : 493598 . doi:10.1155/2012/493598. Google Scholar CrossRef Search ADS PubMed 28 Mitchell AJ . A meta-analysis of the accuracy of the mini-mental state examination in the detection of dementia and mild cognitive impairment . J Psychiatr Res 2009 ; 43 : 411 – 31 . doi: http://dx.doi.org/10.1016/j.jpsychires.2008.04.014. Google Scholar CrossRef Search ADS PubMed 29 Proctor DN , Fauth EB , Hoffman L et al. . Longitudinal changes in physical functional performance among the oldest old: insight from a study of Swedish twins . Aging Clin Exp Res 2006 ; 18 : 517 – 30 . Epub 2007/01/27. PubMed PMID: 17255642. Google Scholar CrossRef Search ADS PubMed 30 Björk MP , Johansson B , Hassing LB . I forgot when I lost my grip—strong associations between cognition and grip strength in level of performance and change across time in relation to impending death . Neurobiol Aging 2016 ; 38 : 68 – 72 . doi:10.1016/j.neurobiolaging.2015.11.010. Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Age and Ageing Oxford University Press

An examination of the heterogeneity in the pattern and association between rates of change in grip strength and global cognition in late life. A multivariate growth mixture modelling approach

Age and Ageing , Volume 47 (5) – Sep 1, 2018

Loading next page...
 
/lp/ou_press/an-examination-of-the-heterogeneity-in-the-pattern-and-association-IFQswWdd4N
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
0002-0729
eISSN
1468-2834
D.O.I.
10.1093/ageing/afy048
Publisher site
See Article on Publisher Site

Abstract

Abstract Background previous research has demonstrated how older adults exhibit different patterns of change in cognitive and physical functioning, suggesting differences in the underlying causal processes. Objective to (i) identify subgroups of older adults that best account for different patterns of longitudinal change in performance on global cognition and grip strength, (ii) examine the interrelationship between global cognition and grip strength trajectories within these subgroups and (iii) identify demographic and health-related markers of class membership. Methods multivariate growth mixture models (GMM) were used to identify groups of individuals with similar developmental trajectories of muscle strength measured by grip strength, and global cognition measured by Mini Mental State Examination (MMSE). Results GMM analyses indicated high, moderate and low functioning groups. Individuals in the high and moderate classes demonstrated better cognitive and physical functioning at the start of the study and less decline than those in the low functioning group. Notably, cognitive performance was related to physical functioning at study entry only among individuals in the low functioning group. Conclusion the study demonstrates the applicability of the multivariate GMM to achieve a better understanding of the heterogeneity of various aging related processes. longitudinal analysis, trajectories, growth mixture model, cognition, grip strength, older people Introduction Current aging research is largely focused on better understanding underlying aging processes with the aim to promote healthy aging. Over the past few decades, the aging of physical and cognitive processes has received extensive attention and has been found to be associated with overall health and increased risk of morbidity, mortality and dementia [1, 2]. Still, a greater focus on the individual differences in patterns and causes of changes in physical and cognitive processes over the life course is needed for predicting late life outcomes and for developing interventions to reduce the risk of negative health outcomes. Although there is consensus that cognitive performance generally declines with age, some evidence suggests that this decline is not normative [3–5]. Similarly, longitudinal studies demonstrate physical decline with age [6]. Despite these similar findings, few, if any, studies have examined the interindividual heterogeneity and the interrelationships in these trajectories. Developments in the analysis of longitudinal data have resulted in recent studies focused on better understanding heterogeneity in the trajectories of cognitive functioning. One such development, referred to as growth mixture models (GMM), allows for the identification of different classes of individuals who cluster together, thus identifying more than one trajectory within the population [7–9]. GMM provides information about the optimal number of classes and the characteristics of each class, including the mean intercept and slope, proportion of membership, and significant predictors of class membership. When GMM [e.g. 7, 8] have been used to explore various patterns of cognitive aging [3–5, 10, 11], they have typically suggested two (slow and rapid decliners) [10, 11] or three (low, moderate and rapid decliners) [3, 4] classes that best represent the underlying heterogeneity of cognitive trajectories. Most analyses of the association between cognitive and physical functioning have been based on modelling approaches (e.g. multivariate growth curve model) not taking into account the possibility for different clusters of individuals exhibiting distinct patterns of change and cross-domain interrelationships. It is quite strongly supported that cognitive and physical functioning are related in cross-sectional studies, and more weakly in longitudinal analyses [12]. How these two processes are interrelated, however, is largely unknown. More studies are needed that explore the relationship between physical and cognitive functioning within different classes of trajectories. Using a GMM with multiple outcomes (i.e. multivariate GMM) allows the exploration of not only the different classes of cognitive and physical functioning but also the association between the two outcomes within each of the identified classes. The purpose of this study was to (i) identify subgroups of older adults best describing longitudinal change in performance on the Mini Mental State Examination (MMSE) and grip strength, (ii) examine the associations between MMSE and grip strength trajectories within each subgroup and (iii) evaluate the role of cognitive and physical activity and demographic variables on differences in class membership. Grip strength was chosen as a biomarker of physical functioning given its sensitivity to physiological change [13] and its use as a valid marker of frailty [14]. Its simplicity to administer, portability and affordability make it easy to apply in numerous settings [13]. The MMSE was selected as the cognitive measure given that a systematic review by Clouston and colleagues (12) found that the MMSE was more strongly related with grip strength than domain-specific measures of cognition in older adults. Method Data Origins of Variance in the Oldest-Old (OCTO-Twin) The OCTO-Twin study included dizygotic and monozygotic twin pairs aged 80 years of age and older [15, 16]. The sample was selected from older adults in the population-based Swedish Twin Registry [17]. Five cycles of longitudinal data were collected at 2-year intervals. The full sample consisted of 702 individuals (351 same-sex pairs). Individuals were excluded due to missing data on the covariates. The sample used in the analysis consisted of 369 individuals ranging from 79.4 to 97.9 years of age. The rate of attrition ranged from 10.5% to 28.8% every two years, primarily due to death (see Table 1). Ethical approval for OCTO-Twin was obtained from the Ethics Committee at the Karolinska Institute in Stockholm and the Swedish Data Inspection Board. Informed consent was obtained from all participants. Descriptive statistics at each occasion for all variables are provided in Table 1. Table 1. Descriptive statistics and attrition rate for MMSE and grip strength at each wave of data collection Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Note. N = 369. *Sample size at each data wave and attrition rate every 2 years. Edu = years of education. Height in cm. Table 1. Descriptive statistics and attrition rate for MMSE and grip strength at each wave of data collection Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Variables Baseline Time 2 Time 3 Time 4 Time 5 MMSE – mean (SD) 25.59 (4.8) 24.62 (6.7) 23.51 (7.9) 22.98 (8.4) 21.55 (8.7) MMSE – N (%)* 367 299 (18.5) 229 (23.4) 168 (26.6) 120 (28.6) Grip – mean (SD) 8.60 (3.0) 7.93 (2.9) 6.98 (3.1) 6.53 (3.0) 6.35 (2.9) Grip – N (%)* 314 281 (10.5) 224 (20.3) 160 (28.6) 114 (28.8) Males – N (%) 118 (32) Age – mean (SD) 82.9 (3.26) Edu – mean (SD) 7.28 (2.6) Height – mean (SD) 161.21(9.1) Note. N = 369. *Sample size at each data wave and attrition rate every 2 years. Edu = years of education. Height in cm. Measures MMSE The MMSE, administered in Swedish, was used to asses global cognitive functioning at each measurement occasion [18]. Grip strength A Martin vigorimeter (Elmed Inc., Addison, IL, USA; medium size bulb) to measure maximum force in pounds per square inch was used [19]. Participants performed the task three times per hand, with the final score being the maximum force exerted. Covariates Age, education and sex were included as covariates on the intercept and slope, and as markers of class membership. Height, level of physical activity, and engagement in cognitive activities were included as markers of class membership. Age, education and height were centred at their mean value. Sex was coded as 0 for males and 1 for females. Level of physical activity was measured by asking respondents: ‘Are you presently doing or have you previously done anything special to train your body or ‘keep your body fit’?’. Response options were ‘no’, ‘yes, to some extent’ or ‘yes, to a great extent’. These were used to derive an indicator as sedentary (0) and active (1). The cognitive activity measure was created by summing six items (e.g. playing games) dichotomously as ‘no’ (0) or ‘yes’ (1) and one item about training the memory and keeping the mind active as ‘no’ (0), ‘yes, to a certain degree’ (1), or ‘yes, definitely’ (2). Total scores ranged from 0 to 8 with higher scores signifying higher engagement in cognitive activity. Analysis Multivariate GMM [20] were employed to identify unobserved groups of individuals with similar linear developmental bivariate trajectories of grip strength and MMSE scores. Class membership was estimated using a multinomial logistic regression adjusted for the covariates. GMM provide estimates of the average MMSE and grip strength trajectories (intercept and slope) and variation for older adults in each latent class [8, 9, 21]. The number of classes is unknown a priori, therefore, we followed recommended procedures to select the best fitting model. First, models were fitted with increasing number of classes. To determine the optimal number of classes, Bayesian Information Criterion (BIC) values were compared from each model. Models with lower BIC values were considered better fitting [22]. Classification quality using entropy values (range from 0 to 1 with higher values meaning that individuals are better discriminated between classes) [23] and interpretability of classes was further considered. Mplus version 7.2 was used to run the GMMs [24]. The model was estimated by maximum likelihood with robust standard errors. Missing data were assumed to be missing at random. Results Number of classes A three-class model of cognitive and physical functioning was selected, given its lower BIC values (2 class = 18,411.00; 3 class = 18,254.08) and good classification of individuals in each class (entropy 2 class = 0.85; entropy 3 class = 0.81). Parameter estimates and standard errors for the three-class model are presented in Table 2. Table 2. Estimates and standard errors for the three-class growth mixture model of cognitive and physical functioning Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Note. N = 369. SE = standard errors. Class 1 = high functioning group. Class 2 = moderate functioning group. Class 3 = low functioning group. Table 2. Estimates and standard errors for the three-class growth mixture model of cognitive and physical functioning Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Class 1 (N = 147) Class 2 (N = 114) Class 3 (N = 108) MMSE Grip Strength MMSE Grip Strength MMSE Grip strength Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Estimates (SE) Fixed effects  Intercept 28.62 (0.17)*** 11.49 (0.48)*** 26.84 (1.08)*** 10.54 (0.49)*** 23.71 (1.24)*** 8.46 (0.82)***   Age −0.09 (0.06) −0.12 (0.08) −0.34 (0.13)* −0.30 (0.12)* −0.18 (0.22) −0.18 (0.13)   Education 0.04 (0.05) 0.01 (0.09) 0.15 (0.25) −0.008 (0.14) 0.96 (0.41)* 0.10 (0.22)   Sex 0.08 (0.94) −3.02 (0.59)*** 0.39 (1.37) −2.93 (0.70)*** −2.90 (1.98) −1.48 (0.96)  Slope −0.12 (0.05)* −0.37 (0.08)*** −0.66 (0.19)** −0.57 (0.06)*** −2.10 (0.18)*** −0.84 (0.12)***   Age −0.003 (0.02) −0.007 (0.03) −0.03 (0.06) 0.003 (0.02) 0.09 (0.03) 0.04 (0.02)   Education 0.01 (0.04) −0.003 (0.03) 0.06 (0.10) −0.02 (0.04) −0.04 (0.24) −0.05 (0.03)   Sex 0.07 (0.09) 0.10 (0.09) −0.23 (0.55) 0.16 (0.11) −0.29 (0.37) 0.005 (0.36) Random effects  Intercept 0.41 (1.67) 3.90 (0.59)*** 0.41 (1.67) 3.90 (0.59) 0.41 (1.67) 3.90 (0.59)***  Slope 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) 0.04 (0.02) 0.009 (0.009) Covariance  Intercept–Intercept 0.04 (0.38) −0.30 (0.80) 4.89 (0.81)***  Slope–Slope 0.003 (0.02) 0.002 (0.03) 0.01 (0.02) Note. N = 369. SE = standard errors. Class 1 = high functioning group. Class 2 = moderate functioning group. Class 3 = low functioning group. Grip strength and MMSE trajectories Class 1: The largest class was composed of 147 individuals (40% of sample) with good physical and cognitive functioning and slow decline. In this high functioning group, individuals had an average baseline MMSE score of 28.62 (SE = 0.17) and declined at a rate of 0.12 (SE = 0.048). See upper solid line in Figure 1. The average baseline grip strength score was 11.49 kPa (SE = 0.48) and rate of decline was 0.37 kPa (SE = 0.077) per year. See lower solid line in Figure 1. No associations were found between grip strength and MMSE. See intercept–intercept and slope–slope covariances in Table 2. Figure 1. View largeDownload slide Average trajectories of physical and cognitive functioning for the three classes. The three upper lines represent cognitive functioning and the three lower lines physical functioning. The solid lines represent high functioning individuals, the dotted lines represent moderate functioning individuals, and the dotdash lines represent low functioning individuals. Figure 1. View largeDownload slide Average trajectories of physical and cognitive functioning for the three classes. The three upper lines represent cognitive functioning and the three lower lines physical functioning. The solid lines represent high functioning individuals, the dotted lines represent moderate functioning individuals, and the dotdash lines represent low functioning individuals. Class 2: A class of 114 individuals (31% of sample) with slightly lower average baseline scores and steeper declining MMSE scores and grip strength was identified. In this moderate functioning group, the average baseline MMSE score was 26.84 (SE = 1.08) and average rate of decline was 0.66 (SE = 0.19). See upper dotted line in Figure 1. Average baseline grip strength was 10.54 kPa (SE = 0.49) and decline rate was 0.57 kPa (SD = 0.064). See lower dotted line in Figure 1. No associations were found between grip strength and MMSE. See intercept–intercept and slope–slope covariances in Table 2. Class 3: A low functioning group, composed of 108 individuals (29% of sample), demonstrated lower MMSE scores and weaker grip strength at the start of the study and steeper rate of decline compared to those in the other two classes. These individuals had an average baseline MMSE score of 23.71 (SE = 1.24) and decline rate of 2.10 kPa (SE = 0.18) per year (see upper dotdash line in Figure 1); and a baseline grip strength score of 8.46 kPa (SE = 0.82) and decline rate of 0.84 kPa (SE = 0.12) per year (see lower dotdash line in Figure 1). Individuals with stronger grip strength at the start of the study also demonstrated higher MMSE scores. No relationship was found between the rate of change in grip strength and the rate of change in MMSE See intercept–intercept and slope–slope covariances in Table 2. Covariates In the high functioning group (class 1), none of the covariates were associated with MMSE at study entry. Sex was associated with baseline grip strength, with females having lower grip strength scores. In the moderate functioning group (class 2), individuals who were older in age at baseline had lower MMSE scores. Female participants and individuals who were older had lower grip strength. In the low functioning group (class 3), individuals with higher educational attainment had higher baseline MMSE scores. Markers of class membership Participation in physical and cognitive activities was markers of class membership. Membership in class 2 was associated with engagement in physical activities (Estimate = 1.31*; SE = 0.53). Individuals who were engaged in physical activities had higher probability of being in the moderate functioning group compared to the low functioning group (class 3) than individuals who did not participate in physical activity. Membership in class 1 was associated with engagement in cognitive activities (Estimate = 0.68***; SE = 0.18). Individuals who reported higher levels of engagement in cognitive activities had a higher probability of being in the high functioning group compared to the low functioning group than individuals who engaged in fewer cognitive activities. Class membership was not influenced by age, height, education or sex. Discussion Our findings demonstrate the use of GMM to examine clusters of trajectories based on MMSE and grip strength scores and for profiling the characteristics of individuals within clusters. This is very important as many research questions focus on the dynamic relationship between selected developmental processes and how they may change over time. Our results show that age-related decline in physical and cognitive function was heterogeneous and not universal. We identified three distinct classes of trajectories for MMSE and grip strength in the oldest-old segment of the population. The high functioning group with good physical and cognitive performance showed minor decline and represents successful agers. The moderate functioning group who had lower cognitive and physical performance and steeper rate of decline than those in the high functioning group might be at greater risk of more substantial decline. On average, they entered the study with an MMSE score of 26.8 which is close to the cut-off score of 26 used to identify individuals at risk of mild cognitive impairment (MCI) and dementia [25]. The low functioning group, with the lowest cognitive and physical performance scores and steepest rate of decline, is likely those who are at greatest risk of also developing clinical dementia. Their average baseline score of 23.7 meets MCI criteria, [25] and almost meets the cut-off of 23 often used for clinical dementia [18]. Interestingly, we found no evidence for differences between the sexes in the low functioning group, suggesting that, for those individuals who are most frail, there are only minor or no differences between men and women. Our finding about the role of education suggests that less educated older adults in the low functioning group are at an increased risk of dementia and highlights the protective role of educational attainment in delaying the onset of the disease which is in line with the cognitive reserve hypothesis [26]. The importance of specific covariates in different classes would have been masked had we examined only one trajectory class, highlighting the importance of GMM. Participation in cognitive and physical activities was associated with an increased likelihood of being in the highly or moderately functioning group compared to the lower functioning group. This aligns with previous longitudinal research on the importance of physical and cognitive activity [27]. Unlike previous studies which focused on the importance of lifestyle activities on the intercept and rate of change (i.e. whether participation in activities can delay decline) of physical and cognitive functioning, the current study demonstrated that these activities increase the likelihood that an individual will follow the highly or moderately functioning trajectory over the course of their later life rather than following the poorly functioning trajectory. This finding suggests that it would be effective to implement interventions which entail increased participation in physical and cognitive activities. By identifying subgroups of individuals with similar changes in cognitive and physical functioning, we were able to better understand the relationship between these two dimensions of the aging process. Our finding that individuals with higher cognitive scores at baseline also had higher grip strength in only one class (low functioning group) suggests that the relationship between the two processes may be most relevant for those individuals who are especially frail and cognitively impaired. The fact that we only found a relationship in one class may help explain previous mixed results about the association between cognitive and physical functioning as other modelling approaches tend to assume that individual trajectories follow a single pattern of change. One limitation of this study is the use of the MMSE with its well-known ceiling effect and modest accuracy in identifying individuals with dementia [28]. Still, the MMSE is suitable for this study since it was not used as a diagnostic tool. Another limitation is that the sample size was too small to model the GMM separately for male and female participants knowing that older adults differ in grip strength norms [29]. Still, a recent bivariate study modelled both sexes together and found a relationship between grip strength and cognition [30]. To the best of our knowledge, this is the first study to examine the longitudinal relationship of subpopulations characterised by their physical and cognitive functioning trajectories. Our research demonstrates that cognitive and physical functioning is best described by more than one trajectory and that studies assuming a single population with an average developmental trajectory tend to miss important information about the class-specific relation between physical and cognitive functioning and class-specific markers. More research is needed to further identify risk factors to being in the low functioning group as these individuals might benefit the most from targeted interventions. Further research is also needed to examine whether similar subpopulations are found when including younger older adults and whether subpopulations of trajectories also exist for other biopsychosocial aging processes. Rather than assuming that aging individuals all follow a similar pattern of change, GMM allows for the modelling of subpopulations of trajectories which can have important implications for better tailored interventions. Key points Cognitive and physical functioning is best described by three distinct classes of trajectories. This study demonstrates the applicability of the multivariate growth mixture model (GMM), decline in physical and cognitive function is heterogeneous and not normative. Modelling subpopulations of trajectories is needed for better tailored interventions. Conflict of interest None. Funding Research reported in this publication was supported by the National Institute on Aging of the National Institutes of Health under Award Number P01AG043362, Integrative Analysis of Longitudinal Studies of Aging and Dementia. The OCTO-Twin study was funded by the National Institute on Aging of the National Institutes of Health (Grant number AG08861) and the Swedish Research Council for Health, Working Life and Welfare (Forte), and the Swedish Brain Power Consortium. The funding source played no role in the design, execution, analysis and interpretation of data, or writing of the study. References 1 Cooper R , Hardy R , Aihie Sayer A et al. . Age and gender differences in physical capability levels from mid-life onwards: the harmonisation and meta-analysis of data from eight UK cohort studies . PLoS One 2011 ; 6 : e27899 . doi:10.1371/journal.pone.0027899. Google Scholar CrossRef Search ADS PubMed 2 Brown JC , Harhay MO , Harhay MN . Walking cadence and mortality among community-dwelling older adults . J Gen Intern Med 2014 ; 29 : 1263 – 9 . Google Scholar CrossRef Search ADS PubMed 3 Terrera GM , Brayne C , Matthews F . One size fits all? Why we need more sophisticated analytical methods in the explanation of trajectories of cognition in older age and their potential risk factors . Int Psychogeriatr 2010 ; 22 : 291 – 9 . doi: http://dx.doi.org/10.1017/S1041610209990937; PubMed PMID: 223742514; 19906326. Google Scholar CrossRef Search ADS PubMed 4 Hayden KM , Reed BR , Manly JJ et al. . Cognitive decline in the elderly: an analysis of population heterogeneity . Age Ageing 2011 ; 40 : 684 – 9 . Google Scholar CrossRef Search ADS PubMed 5 Xie H , Mayo N , Koski L . Identifying and characterizing trajectories of cognitive change in older persons with mild cognitive impairment . Dement Geriatr Cogn Disord 2011 ; 31 : 165 – 72 . doi: http://dx.doi.org/10.1159/000323568; PubMed PMID: 859334228; 21346357. Google Scholar CrossRef Search ADS PubMed 6 Jerome GJ , Ko S-U , Kauffman D , Studenski SA , Ferrucci L , Simonsick EM . Gait characteristics associated with walking speed decline in older adults: results from the Baltimore longitudinal study of aging. Arch Gerontol Geriatr 2015 ; 60 : 239 – 43 . doi:10.1016/j.archger.2015.01.007. Google Scholar CrossRef Search ADS PubMed 7 Liu M , Hancock G . Unrestricted mixture models for class identification in growth mixture modeling. Educ Psychol Meas 2014 ; 74 : 557 – 84 . doi:10.1177/0013164413519798. Google Scholar CrossRef Search ADS 8 Muthén B . Latent Variable Analysis: Growth Mixture Modeling and Related Techniques for Longitudinal Data. The SAGE Handbook of Quantitative Methodology for the Social Sciences . Thousand Oaks, CA : SAGE Publications, Inc , 2004 ; pp. 346 – 70 . 9 Muthén B , Shedden K . Finite mixture modeling with mixture outcomes using the EM algorithm . Biometrics 1999 ; 55 : 463 – 9 . doi:10.1111/j.0006-341×.1999.00463.x. Google Scholar CrossRef Search ADS PubMed 10 Proust C , Jacqmin-Gadda H . Estimation of linear mixed models with a mixture of distribution for the random effects . Comput Methods Programs Biomed 2005 ; 78 : 165 – 73 . doi:http://dx.doi.org/10.1016/j.cmpb.2004.12.004. Google Scholar CrossRef Search ADS PubMed 11 Small BJ , Backman L . Longitudinal trajectories of cognitive change in preclinical Alzheimer’s disease: a growth mixture modeling analysis. Cortex 2007 ; 43 : 826 – 34 . doi:10.1016/S0010-9452(08)70682-8. Google Scholar CrossRef Search ADS PubMed 12 Clouston SA , Brewster P , Kuh D et al. . The dynamic relationship between physical function and cognition in longitudinal aging cohorts . Epidemiol Rev 2013 ; 35 : 33 – 50 . Epub 2013/01/26. doi: 10.1093/epirev/mxs004; PubMed PMID: 23349427; PubMed Central PMCID: PMCPMC3578448. Google Scholar CrossRef Search ADS PubMed 13 Bohannon RW . Hand-grip dynamometry predicts future outcomes in aging adults . J Geriatr Phys Ther 2008 ; 31 : 3 – 10 . Epub 2008/05/21. PubMed PMID: 18489802. Google Scholar CrossRef Search ADS PubMed 14 Syddall H , Cooper C , Martin F , Briggs R , Aihie Sayer A . Is grip strength a useful single marker of frailty? Age Ageing 2003 ; 32 : 650 – 6 . Epub 2003/11/06. PubMed PMID: 14600007. Google Scholar CrossRef Search ADS PubMed 15 Johansson B , Hofer SM , Allaire JC et al. . Change in memory and cognitive functioning in the oldest-old: The effects of proximity to death in genetically related individuals over a six-year period . Psychol Aging 2004 ; 19 : 145 – 56 . Google Scholar CrossRef Search ADS PubMed 16 McClearn GE , Johansson B , Berg S et al. . Substantial genetic influence on cognitive abilities in twins 80 or more years old . Science 1997 ; 276 : 1560 – 3 . Google Scholar CrossRef Search ADS PubMed 17 Cederlöf R , Lorich U . The Swedish Twin Registry. In: Nance WE , Allen G , Parisi P , eds . Twin research: Biology and epidemiology . New York, NY : Alan R. Riss , 1978 ; pp. 189 – 95 . 18 Folstein MF , Folstein SE , McHugh PR . Mini-mental state’: a practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res 1975 ; 12 : 189 – 98 . Google Scholar CrossRef Search ADS PubMed 19 Sipers WM , Verdijk LB , Sipers SJ , Schols JM , van Loon LJ . The Martin vigorimeter represents a reliable and more practical tool than the Jamar dynamometer to assess handgrip strength in the geriatric patient . J Am Med Dir Assoc 2016 ; 17 : 466.e1 – 7 . Epub 2016/04/24. doi:10.1016/j.jamda.2016.02.026; PubMed PMID: 27107163. Google Scholar CrossRef Search ADS 20 Wu J , Witkiewitz K , McMahon RJ , Dodge KA . A parallel process growth mixture model of conduct problems and substance use with risky sexual behavior . Drug Alcohol Depend 2010 ; 111 : 207 – 14 . doi: http://dx.doi.org/10.1016/j.drugalcdep.2010.04.013; PubMed PMID: 2010-20916-004. Google Scholar CrossRef Search ADS PubMed 21 Berlin KS , Parra GR , Williams NA . An introduction to latent variable mixture modeling (part 2): longitudinal latent class growth analysis and growth mixture models. J Pediatr Psychol 2014 ; 39 : 188 – 203 . doi:10.1093/jpepsy/jst085. Google Scholar CrossRef Search ADS PubMed 22 Nylund KL , Asparouhov T , Muthén BO . Deciding on the number of classes in latent class analysis and growth mixture modeling: a Monte Carlo simulation study . Struct Equ Modeling 2007 ; 14 : 535 – 69 . doi:10.1080/10705510701575396. Google Scholar CrossRef Search ADS 23 Celeux G , Soromenho G . An entropy criterion for assessing the number of clusters in a mixture model . J Classif 1996 ; 13 : 195 – 212 . doi:10.1007/BF01246098. Google Scholar CrossRef Search ADS 24 Muthén LK , Muthén BO . Mplus User’s Guide , 7th ed . Los Angeles, CA : Muthén & Muthén , 1998 -2015. 25 Nasreddine ZS , Phillips NA , Bédirian V et al. . The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment . J Am Geriatr Soc 2005 ; 53 : 695 – 9 . doi:10.1111/j.1532-5415.2005.53221.x. Google Scholar CrossRef Search ADS PubMed 26 Terrera GM , Minett T , Brayne C , Matthews FE . Education associated with a delayed onset of terminal decline . Age Ageing 2014 ; 43 : 26 – 31 . doi: 10.1093/ageing/aft150; PubMed PMID: PMC3861340. Google Scholar CrossRef Search ADS PubMed 27 Lindwall M , Cimino CR , Gibbons LE et al. . Dynamic associations of change in physical activity and change in cognitive function: coordinated analyses of four longitudinal studies . J Aging Res 2012 ; 2012 : 493598 . doi:10.1155/2012/493598. Google Scholar CrossRef Search ADS PubMed 28 Mitchell AJ . A meta-analysis of the accuracy of the mini-mental state examination in the detection of dementia and mild cognitive impairment . J Psychiatr Res 2009 ; 43 : 411 – 31 . doi: http://dx.doi.org/10.1016/j.jpsychires.2008.04.014. Google Scholar CrossRef Search ADS PubMed 29 Proctor DN , Fauth EB , Hoffman L et al. . Longitudinal changes in physical functional performance among the oldest old: insight from a study of Swedish twins . Aging Clin Exp Res 2006 ; 18 : 517 – 30 . Epub 2007/01/27. PubMed PMID: 17255642. Google Scholar CrossRef Search ADS PubMed 30 Björk MP , Johansson B , Hassing LB . I forgot when I lost my grip—strong associations between cognition and grip strength in level of performance and change across time in relation to impending death . Neurobiol Aging 2016 ; 38 : 68 – 72 . doi:10.1016/j.neurobiolaging.2015.11.010. Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Published by Oxford University Press on behalf of the British Geriatrics Society. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Age and AgeingOxford University Press

Published: Sep 1, 2018

There are no references for this article.

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