Clinical Trials Targeting Aging and Age-Related Multimorbidity

Abstract

Background

There is growing interest in identifying interventions that may increase health span by targeting biological processes underlying aging. The design of efficient and rigorous clinical trials to assess these interventions requires careful consideration of eligibility criteria, outcomes, sample size, and monitoring plans.

Methods

Experienced geriatrics researchers and clinical trialists collaborated to provide advice on clinical trial design.

Results

Outcomes based on the accumulation and incidence of age-related chronic diseases are attractive for clinical trials targeting aging. Accumulation and incidence rates of multimorbidity outcomes were developed by selecting at-risk subsets of individuals from three large cohort studies of older individuals. These provide representative benchmark data for decisions on eligibility, duration, and assessment protocols. Monitoring rules should be sensitive to targeting aging-related, rather than disease-specific, outcomes.

Conclusions

Clinical trials targeting aging are feasible, but require careful design consideration and monitoring rules.

The incidence of age-related chronic diseases rises exponentially with age (1). This parallels the exponential increases with age in rates of major disease-specific deaths tracked by the U.S. National Center for Health Statistics, including those for heart disease, cancer, stroke, type 2 diabetes mellitus, and Alzheimer’s disease (2,3). It has repeatedly been shown that the major, and by far the most potent, risk factor cutting across age-related chronic diseases is age itself. The impact of interventions targeting individual chronic diseases on health span may thus be limited by the competing risks of other age-related chronic diseases (4).

There is growing evidence for a biologic construct underlying aging, leading to the potential that interventions may be developed to slow its progression (5). The primary goal is not to increase the number of years lived, but to increase the number of years lived with better health and function. This has great societal importance. Beyond the obvious impact of multiple diseases on individual well-being, care for elderly individuals accounts for 43% of total U.S. health care spending, approximately $1 trillion per year (4). This number is expected to rise to $6 trillion per year by 2050. Reducing these costs is critical for social and economic prosperity: Even a modest increase in health span (2.2 years) may reduce expenses by $7 trillion over the next 50 years (4).

Two thirds of elderly persons in the United States have two or more chronic diseases and that the 14% who have six or more diseases account for 47% of total Medicare spending (6). Care for these patients is clinically challenging because treatment guidelines for individual conditions can conflict, treatment priorities can be difficult to establish, and the effectiveness of individual drugs may be altered (7,8).

Advances in understanding the biological basis of aging are leading to the potential of identifying behavioral and pharmacologic intervention targets for increasing health span (9–12). The National Institute on Aging Interventions Testing Program has been established to organize research toward this goal across model organisms (13,14). As interventions emerge from this program as candidates for human intervention, clinical trials will be mounted to assess their efficacy. Members of the Multimorbidity Clinical Trials Consortium have collaborated on this manuscript to inform the development and conduct of these trials. We discuss design and analytical issues, including the choice of outcomes, eligibility criteria, monitoring rules, and analytical strategies. We present projections of rates at which outcomes occur, as benchmarks for estimating the statistical power for future trials.

Choice of Outcomes

Aging presents many potential targets for therapeutic interventions that can be evaluated in clinical trials. Mortality is one obvious outcome. Other targets include related outcomes such as active life expectancy or expected disability-free survival (15,16), and markers of geriatric syndromes (eg, falls, gait, incontinence, frailty indices, depression scales, assessments of physical and cognitive functions, and activities of daily living) (17,18).

Another hallmark of aging is the accumulation of age-related chronic diseases (5,19). These often co-occur with disabilities and geriatric syndromes, and their cumulative incidence is accelerated among individuals facing compressed health spans and poorer quality of late life (5,19). There are many age-related chronic conditions; the combinations in which they occur are diverse (20,21). To focus our discussion, we considered two composite outcomes drawn from the list of 20 chronic diseases developed by the U.S. Department of Health and Human Services (7). We use two of these diseases (hypertension and hyperlipidemia) as inclusion criteria to identify individuals at increased risk for accelerated aging. We selected 12 others that we considered particularly attractive targets for interventions to influence aging: arthritis, cancer, cardiac arrhythmias, chronic kidney disease, chronic obstructive pulmonary disease, congestive heart failure, coronary artery disease, dementia, depression, diabetes, osteoporosis, and stroke. The six we did not select (asthma, autism spectrum disorder, hepatitis, human immunodeficiency virus, schizophrenia, and substance abuse disorders) were judged to be less likely to be sensitive to approaches currently being researched by the National Institute on Aging Interventions Testing Program. To these 12 chronic conditions, we added all-cause death.

Rate of Accumulation of New Age-Related Chronic Diseases

One outcome of clear clinical importance is the rate at which the number of these chronic diseases increases over time. This can be captured by how raw counts of the 12 chronic diseases identified in Choice of Outcomes and death accumulate during follow-up. We provide benchmark data for this outcome. Alternatives are indices, such as Charlson Comorbidity Indices, in which chronic disease conditions are weighted and summed to provide scores (21), however no clear consensus for how weights should be defined exists (5).

Incidence of New Major Families of Chronic Diseases

The chronic diseases we consider share many risk factors and underlying mechanisms. To demonstrate that an intervention is effective in slowing aging, rather than having an effect focused on a specific mechanistic pathway, it may be important to group these into families according to major shared risk factor pathways and consider intervention effects across families. One approach is to identify five major families: diseases related to (i) atherosclerosis and cardiovascular disease (cardiac arrhythmias, congestive heart failure, coronary artery disease, and stroke), (ii) cancers (excluding non-melanoma skin cancers), (iii) dementia (including Alzheimer’s disease), and (iv) diabetes, and additionally (v) all-cause deaths. Although these five families exclude some of the 12 age-related conditions (eg, arthritis, chronic kidney disease, chronic obstructive pulmonary disease, depression, and osteoporosis) considered earlier, as a composite outcome they are likely to capture major components of aging.

Eligibility Criteria

It may be reasonable to expect that the effectiveness of interventions targeting aging depends on whether these are applied as primary prevention of multimorbidity, that is, before there is clinical evidence of accelerated aging such as the occurrence of at least one major aging-related clinical event, versus downstream therapy as secondary prevention, that is, after at least one age-related chronic disease has emerged. For trials to be feasible, it may be important to select participants expected to be at enhanced risk for accelerated aging, due to older age, increased risk factor burdens, or potentially genetic predisposition. Deficits in functional outcomes, such as slower gait speed or lower cognitive function, may be important (22). Intermediate metabolic risk factors, such as impaired glucose tolerance, markers of inflammation, or renal insufficiency (23,24) and evidence of unhealthy behaviors, such as low physical activity or obesity (25), may also be considered.

Estimation of Incidence and Progression Rates

To project rates that outcomes accrue, for determining power and sample sizes for clinical trials, investigators from three large longitudinal studies of older adult cohorts conducted the following exercise. They examined two outcomes: (i) the rate at which new chronic diseases and deaths accumulate over time and (ii) the incidence (ie, time until the first occurrence) of a new major family of chronic diseases (from the list of four given in Incidence of New Major Families of Chronic Diseases) or death. Individuals aged 65 to 79 years were grouped into two strata: whether or not they had at least one age-related chronic disease family at the beginning of follow-up. The primary prevention stratum was to have no history of the families of chronic diseases described earlier, but to be at increased risk for these (eg, due to hypertension, hyperlipidemia, and/or functional limitations). The secondary prevention stratum included individuals with a history of one or two major families of chronic conditions. Rates were generated separately by age and sex.

For the outcome of individual chronic disease accumulation, rates were based on 4- or 5-year spans of follow-up. For the outcome of incidence of a new family of major chronic diseases, rates at 2, 4, and 6 years were provided. We purposefully did not attempt rigorous standardization of methods. Instead, investigators from each study used the data resources at hand to identify cohorts and project rates. By using cohort studies, we examined trends occurring during the course of “usual care.” This follows from the expectation that trials will not provide medical care outside of that required for administering their interventions. Here is a brief description how this exercise was approached: how cohorts were selected, how chronic conditions were assessed, and how rates were projected. Supplementary Table 1 provides a summary.

Health and Retirement Study

The Health and Retirement Study (HRS) is a nationally representative longitudinal sample of the U.S. population 50 years and older that was interviewed approximately every 2 years since 1992 (26). It is refreshed with new members every 6 years to maintain age representativeness. We included members aged 65–79 years, beginning in either 2006 or 2008 depending on which wave they had biomarker measurements, so that initial eligibility criteria could be assessed. They were followed at 2-year intervals through 2012, with biomarker measurement 4 years after the first measurement, that is, some cohort members were only followed for 4 years and some for 6. Outcomes were assessed every 2 years.

Individuals were included if they had no chronic kidney disease (cystatin C > 1.55 mg/L), no current treatment for cancer (self-reports of current treatment), no dementia (testing or proxy report), and no difficulty in walking across a room. In the nationally representative sample of 6,128 persons 65–79 years old, 33.8% were eliminated because they did not meet eligibility criteria.

The remaining 4,058 were divided into primary and secondary prevention strata. Individuals in the primary prevention stratum had no history of cancer, cardiovascular disease, diabetes, or mild cognitive impairment, but were selected to be at enhanced risk for the development of age-related chronic diseases due to at least one hypertension or hyperlipidemia or difficulty in performing one out of five activities of daily living. The secondary prevention stratum included participants with one or two of the following: a history of cancer, cardiovascular disease, or diabetes.

The rate that new chronic conditions accumulated from baseline was based on 10 diseases or conditions: arthritis, cancer, chronic kidney disease (measured by cystatin C), cognitive impairment (including mild cognitive impairment and dementia), depression, diabetes heart disease, hypertension (based on direct measurement or use of prescription medications), lung disease, and stroke. It was determined by calculating the number of new conditions divided by the participant’s follow-up time in years. The composite outcome of the incidence of new families of chronic diseases was defined as the first on-study incidence of any of the following outcomes: cancer, cardiovascular disease (including coronary artery disease, congestive heart failure, and stroke), diabetes, or death and was determined at each of three waves 2 years apart.

Rochester Epidemiology Project

The Rochester Epidemiology Project (REP) is a population-based records-linkage system located in Rochester, Minnesota that includes data on medical visits and diagnoses for nearly all persons who have ever lived in Olmsted County, Minnesota since 1966 (27). It has been used extensively to study the prevalence and incidence of multimorbidity (1,27).

The REP was used to define a cohort aged 65 to 79 years on January 1, 2005. We included persons at baseline who were free of a recent diagnosis of cancer (within 2 years) using ICD-9 diagnosis codes in the 5-year window before January 1, 2005 (January 1, 2000 to December 31, 2004), chronic kidney disease, and dementia. The presence of conditions was based on the list of diagnosis codes identified by the Department of Health and Human Services (1,7). We included data from the first 6 years of follow-up after baseline.

For the primary prevention stratum, we included persons without a history of cardiovascular disease (coronary artery disease, congestive heart failure, or stroke), without diabetes, and without a history of cancer. To select individuals at increased risk for age-related chronic diseases, we only included individuals who had at least one or more of hypertension, hyperlipidemia, and at least one problem with activities of daily living (eating by oneself, bathing oneself, dressing oneself, walking, or using the toilet). For the secondary prevention stratum, we selected persons with one or two of a history of cancer, cardiovascular disease, or diabetes.

The rate of accumulation of new chronic diseases was based on 12 age-related chronic conditions (arthritis, cancer, cardiac arrhythmia, chronic kidney disease, chronic obstructive pulmonary disease, coronary artery disease, congestive heart failure, dementia, depression, diabetes, osteoporosis, and stroke) and death. Rates were calculated at 5 years of follow-up by dividing the number of new conditions acquired within each person by the span of follow-up. The composite outcome of the incidence of new families of chronic diseases was defined as the first on-study incidence of any of the following outcomes: cancer, cardiovascular disease (including coronary artery disease, congestive heart failure, and stroke), dementia, diabetes, or death and was determined passively when medical care was received or death reported.

Women’s Health Initiative Observational Study

The Women’s Health Initiative Observational Study (WHIOS) consisted of 93,676 women who were enrolled during 1995–1998 at 40 academic centers across the United States (28). We accessed the data from women’s first 6 years, limiting it to those aged 65–79 years at enrollment. At that time and annually thereafter, women self-reported histories of conditions and medical procedures. Alzheimer’s disease was based on self- or proxy-report. Hypertension and dyslipidemia were based on current treatment. Diabetes was primarily based on self-report. Physical function was based on self-report.

To increase the risk among women in the primary prevention stratum, we required history of treatment with lipid-lowering medications and history of hypertension treatment and/or measured blood pressure >140/90 mmHg, and also included women who reported that vigorous activities were limited “a lot” by their health. For the secondary prevention stratum, we selected women reporting one or two of a history of cancer, cardiovascular disease, or diabetes at enrollment. Women currently on dialysis were excluded from both strata.

The rate of chronic disease accumulation was based on annual reports of Alzheimer’s disease, arthritis (rheumatoid arthritis or osteoarthritis), atrial fibrillation, cancer (excluding non-melanoma skin cancer), cardiovascular disease, congestive heart failure, depression (treated) diabetes, emphysema/chronic bronchitis, kidney dialysis, osteoporosis, stroke, and death. Participant’s individual accumulation rates were the number of new conditions divided by observed follow-up time over 5 years.

The composite outcome of the incidence of new families of chronic diseases was defined as the first on-study incidence of any of the following outcomes: cancer (self-report of any cancer except non-melanoma skin), cardiovascular disease (clinical myocardial infarction, silent myocardial infarction, death due to coronary heart disease, coronary bypass grafting, percutaneous transluminal coronary angioplasty, hospitalization for angina, deep vein thrombosis, stroke, coronary revascularization, carotid artery disease, or cardiac catheterization), dementia (self-report of diagnosis), diabetes, or death. Incidence rates were based on Kaplan–Meier plots.

Projections of Rates

Incidence of a New Family of Chronic Diseases

Table 1 describes participants contributing to the projections from each cohort. The HRS cohort included 4,046 individuals (43.8% primary prevention). The REP provided 6,208 individuals (50.5% primary prevention). The WHIOS provided 24,798 women (37.7% primary prevention).

Table 1.

Analytical Databases for Three Cohorts

	HRS		REP		WHIOS
	Women	Men	Women	Men	Women	Men
Baseline Characteristics	N = 2,241	N = 1,805	N = 3,421	N = 2,787	N = 24,708
Age (y)
65–69	887 (39.6)	712 (39.5)	1,339 (39.0)	1,190 (42.7)	9,557 (38.7%)
70–74	792 (35.3)	623 (34.5)	1,075 (31.4)	905 (32.5)	10.131 (41.0%)	NA
75–80	562 (25.1)	470 (26.0)	1,013 (29.6)	692 (24.8)	5,018 (20.3%)
Strata
Primary prevention	1093 (48.8)	679 (37.6)	2,015 (58.9)	1,122 (40.3)	9,305 (37.7%)	NA
Secondary prevention	1148 (51.2)	1126 (62.4)	1,406 (41.1)	1,665 (59.7)	15,401 (62.3%)
Medical history
Cancer	344 (15.4)	287 (15.9)	357 (10.4)	316 (5.1)	5,901 (23.0%)
Cardiovascular disease	623 (27.8)	701 (38.8)	700 (20.5)	1,119 (40.2)	9,863 (39.8%)	NA
Diabetes	503 (22.5)	516 (28.6)	626 (18.3)	694 (24.9)	2,368 (9.6%)

Note: HRS = Health and Retirement Study; NA = not applicable; REP = Rochester Epidemiology Project; WHIOS = Women’s Health Initiative Observational Study.

Rates of Chronic Diseases Accumulation Over Time

Table 2 provides estimates for the rates that individual chronic diseases accumulate over 4–5 years, that is, mean slopes of cumulative tallies separately by gender and stratum. Rates of accumulation tended to be slightly higher among men than women and to increase with age. Accumulation rates in the REP cohort were uniformly greater than for the HRS and WHIOS cohorts, which were fairly similar. Rates in the secondary prevention strata were slightly larger than those for the primary prevention strata.

Table 2.

Mean (SD) Rates that New Individual Chronic Conditions Accumulate

Age (y)	New Chronic Diseases Per Year
	HRS		REP		WHIOS
	Primary	Secondary	Primary	Secondary	Primary	Secondary
65–69
Women	0.14 (0.17)	0.17 (0.21)	0.26 (0.30)	0.31 (0.39)	0.16 (0.17)	0.19 (0.21)
Men	0.15 (0.18)	0.19 (0.21)	0.29 (0.30)	0.33 (0.43)	NA	NA
70–74
Women	0.13 (0.16)	0.18 (0.19)	0.33 (0.34)	0.39 (0.45)	0.18 (0.18)	0.22 (0.24)
Men	0.19 (0.22)	0.19 (0.22)	0.36 (0.37)	0.40 (0.47)	NA	NA
75–79
Women	0.17 (0.21)	0.25 (0.25)	0.42 (0.43)	0.45 (0.45)	0.20 (0.20)	0.24 (0.25)
Men	0.24 (0.22)	0.24 (0.24)	0.47 (0.46)	0.48 (0.50)	NA	NA

Note: HRS = Health and Retirement Study; NA = not applicable; REP = Rochester Epidemiology Project; WHIOS = Women’s Health Initiative Observational Study.

Table 3 summarizes the projected incidence rates for a new family of chronic diseases for the three cohorts, grouped by age, sex, and primary and secondary prevention stratum. Rates are provided across 2, 4, and 6 years. These tend to increase with age and to be higher among men than women, but there is little difference between the primary and secondary cohorts. Rates from the WHIOS are markedly lower than the REP, with the HRS rates intermediate. Overall, the differences between the rates between the primary and secondary prevention cohorts are not large and vary in sign among age groups and genders.

Table 3.

Incidence Rates for a New Major Chronic Disease or Death at Three Time Points

Age (y)	2 Years		4 Years		6 Years
	Primary	Secondary	Primary	Secondary	Primary	Secondary
Health and Retirement Study
65–69
Women	0.133	0.146	0.236	0.247	0.299	0.309
Men	0.204	0.157	0.355	0.313	0.486	0.414
70–74
Women	0.112	0.174	0.258	0.255	0.343	0.376
Men	0.198	0.147	0.397	0.324	0.546	0.451
75–79
Women	0.160	0.209	0.297	0.348	0.371	0.481
Men	0.256	0.156	0.454	0.328	0.578	0.502
Rochester Epidemiology Project
65–69
Women	0.174	0.182	0.392	0.348	0.529	0.472
Men	0.266	0.227	0.471	0.411	0.641	0.557
70–74
Women	0.205	0.248	0.438	0.409	0.619	0.574
Men	0.324	0.261	0.579	0.474	0.728	0.607
75–79
Women	0.276	0.240	0.492	0.478	0.653	0.623
Men	0.354	0.321	0.632	0.562	0.736	0.709
Women’s Health Initiative Observational Study
65–69
Women	0.048	0.118	0.114	0.193	0.179	0.259
Men	NA	NA	NA	NA	NA	NA
70–74
Women	0.059	0.136	0.136	0.178	0.214	0.216
Men	NA	NA	NA	NA	NA	NA
75–79
Women	0.075	0.148	0.162	0.256	0.246	0.350
Men	NA	NA	NA	NA	NA	NA

Note: NA = not applicable.

Supplementary Table 2 lists the death rates in terms of person years of follow-up for the three cohorts. These are much lower than the rates at which new families of chronic diseases occur.

Summary of Rate Projections

The exercises from the three cohorts in projecting rates for potential outcomes can be used to develop sample size projections for clinical trials of interventions to reduce multimorbidity. As would be expected, rates increased over the age range we examined (65 to 79 years) and were slightly lower among women than men. We are encouraged that the projections of accumulation and incidence rates that we provide, depending on trial duration, targeted intervention effect, number of arms, and retention rates, are sufficiently large to lead to trials that may be feasible, that is, involve a few thousand participants for 4–6 years of planned follow-up. As examples, the median accumulation rates and standard deviations across the cells in Table 3 are both 0.24 and the median 4-year incidence rate is 0.411. To achieve 90% power to detect a 20% intervention effect on mean rates (two-tailed type 1 error of 0.05; two equal sized arms), approximately 1,050 total participants are required. Similarly, to achieve a 20% reduction in the 4-year incidence cumulative hazard, approximately 2,052 total participants are required. These power estimates require adjustments for lost follow-up, assessment schedules, interim monitoring, and other design features, but signal that such trials are feasible. Clearly, they appear to be much more tractable than trials with mortality as the primary outcome in cohorts similar to those we present: mortality rates are much lower.

There are many reasons that rates may vary among cohorts. These include differences among recruitment practices, assessment protocols, and risk factors. One key difference is how outcomes are ascertained. The REP provided consistently higher rates than the HRS and WHIOS cohort. Likely, this is at least partially due to its use of electronic medical record surveillance of a fairly circumscribed cohort. This approach may more fully capture events and is less dependent on participant’s memory and perhaps less sensitive to differential attrition. Another difference may be related to central adjudication, such as that used in the WHIOS study for some of its outcomes: Use of central adjudication may alter ascertainment rates (29).

Statistical Issues

The interpretation of findings from clinical trials with composite outcomes has been frequently discussed (30). Several issues require consideration. For example, it may be difficult to ascribe efficacy to individual components of the composite, even when overall findings are positive. For the trial to be interpreted as successfully targeting aging, its effect must be expressed broadly among families of chronic diseases. If its benefits are apparent, for example, solely within the family of cardiovascular disease, this is important, but it is difficult to see this as a demonstration of an effect on aging unless there is a concomitant decrease in other age-related conditions. One way to assess this is through the use of shared parameter models, in which a latent variable is included in the parameterization of event rates for individual conditions and the intervention effect on this variable is assessed. A related approach is using multivariate survival models to assess incidence times of multiple endpoints (31,32). Another alternative is to follow analyses of the composite with formal tests for heterogeneity in the intervention effects on the individual components (33). Although power may be limited, this approach is designed to detect situations when there are marked differences in how the intervention is related to components. Nonparametric approaches to analyzing composite outcomes have also been proposed (34).

Rules guiding interim testing may also not be straightforward. For example, in a trial targeting accelerated aging as expressed through multimorbidity, under what situations should the trial be stopped if there is only current evidence for benefit for a single family of chronic diseases and little current evidence for a more diffuse effect that may be interpreted as targeting accelerated aging? One possibility is to conduct interim testing only on mortality, while reserving final analyses for the composite (35). Other possibilities are to use strategies such as Bonferroni adjustments to ensure that boundaries guiding stopping rules for each component are appropriately conservative. Adjustment for competing causes of mortality influences trials with endpoints other than total mortality (36).

It is likely that lost follow-up will be related to accumulation of chronic conditions. For example, participants who enter nursing homes may not continue to be contacted, and it is likely that these individuals will have increased rates of chronic disease accumulation. Great progress has been made in identifying strategies to improve the validity of clinical trials despite lost follow-up and to assess the sensitivity of findings to differential follow-up (37).

Conclusions

Clinical trials of interventions to slow the rate that age-related chronic conditions accumulate are likely to become increasingly common. Exercises we have conducted to project rates from three large cohorts support the feasibility of their development: Cohorts at high risk for these outcomes can be identified. Clearly many assumptions and designs should be considered and optimal choices should depend on the specific intervention. Careful deliberation is necessary for developing analysis plans and monitoring rules for these trials.

Supplementary Material

Supplementary material can be found at: http://biomedgerontology.oxfordjournals.org/

Funding

The Health and Retirement Study was funded by the National Institute on Aging (NIA; U01 AG09740). The Rochester Epidemiology Project infrastructure is funded by the National Institutes of Health (R01AR030582 and R01AG034676; PI: W.A.R., MD, MPH). The Women’s Health Initiative was funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services. The American Federation for Aging Research has provided support for this work (J.P.C. and N.B.). Additional support has been provided by the Nathan Shock Center of Excellence for the Biology of Aging (P30AG038072, N.B.), the Glenn Center for the Biology of Human Aging (Paul Glenn Foundation for Medical Research) (N.B.), grant 1R24AG044396 from the NIA (PI: Kirkland; co-PI: N.B.), grant P30 AG021332 from the NIA to the Wake Forest Older Americans Independence Center (S.B.K.), and HHS contract HHSN26801100004C (PI: Shumaker; co-PI: M.A.E).