Physical Function in Middle-aged and Older Adults With Type 1 Diabetes: Long-term Follow-up of the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study

Functional limitations and disabilities associated with aging are more prevalent, occur earlier, and progress more rapidly in individuals with diabetes compared with those without diabetes (1–5). However, little is known about the burden of functional limitations in older adults with type 1 diabetes. While this knowledge gap is due in part to the shortened life span historically associated with type 1 diabetes (6), life expectancy for individuals with type 1 diabetes has increased and was recently shown to be only slightly shorter than that of the U.S. general population (7–9). As more people with type 1 diabetes survive to advanced ages, it is important to understand how type 1 diabetes may impact physical function.

The Diabetes Control and Complications Trial (DCCT) established the benefit of intensive therapy to delay the onset and slow the progression of microvascular and neuropathic complications in type 1 diabetes (10). In the long-term observational follow-up of the DCCT cohort, the Epidemiology of Diabetes Interventions and Complications (EDIC) study, rates of diabetic retinopathy, kidney disease, neuropathy, cardiovascular disease (CVD), and mortality were lower in EDIC participants assigned to intensive therapy compared with those assigned to conventional therapy during DCCT, despite convergence of levels of glycemia during EDIC. This phenomenon is known as “metabolic memory” (11,12). The aims of this study were to 1) examine the prevalence of functional limitations in the EDIC cohort, 2) identify demographic and clinical factors associated with functional limitations, and 3) determine whether the benefits of prior DCCT intensive therapy extend to measures of functional limitations. We hypothesized that hyperglycemia and the presence of diabetes-related complications would be associated with a greater prevalence of functional limitations.

The DCCT and EDIC have previously been described (10,13–15). Briefly, between 1983 and 1987, the DCCT enrolled 1,441 individuals 13–39 years of age with type 1 diabetes and randomly assigned them to either intensive (N = 711) or conventional (N = 730) diabetes therapy. Intensive therapy was intended to achieve glucose levels as close to the nondiabetic range as safely possible using frequent self-monitoring of blood glucose to guide insulin doses administered via multiple daily injections or insulin infusion pumps. With conventional therapy the goal was to maintain clinical well-being using one to two daily injections of insulin without routine glucose monitoring. The primary prevention cohort (N = 726) had a diabetes duration of 1–5 years, no diabetic retinopathy, and urine albumin excretion rate (AER) <40 mg/24 h at baseline. The secondary intervention cohort (N = 715) had diabetes duration of 1–15 years, minimal to moderate nonproliferative diabetic retinopathy, and albumin excretion rate <200 mg/24 h at baseline. In 1993, the DCCT concluded, having demonstrated beneficial effects of intensive therapy on the development and progression of microvascular complications (10). Conventional therapy participants were taught intensive therapy, and diabetes clinical care was transitioned to community providers. In 1994, all surviving DCCT participants were invited to enroll in EDIC for assessment of the development and progression of and risk factors for more advanced diabetes complications including CVD, age-related morbidities, and survival (13).

The current study included 1,094 of 1,185 active EDIC participants (mean ± SD age 59.5 ± 6.8 years, diabetes duration 37.9 ± 4.9 years) and was conducted during the 25th through 26th years of EDIC follow-up (2018–2019) (14–16) (Supplementary Fig. 1). Institutional review boards at all 27 EDIC clinical centers approved the protocol. Written informed consent was obtained from all EDIC participants. EDIC staff were centrally trained and certified to conduct the assessments.

The primary outcome was a Short Physical Performance Battery (SPPB) score <10 (16). The SPPB is an objective measure of lower-extremity mobility, and the SPPB score, ranging from 0 (worst) to 12 (best score), represents the sum of scores for each SPPB component (balance, gait speed, and repeated chair stands) (16). Lower SPPB scores are associated with worse physical function in older adults and SPPB scores <10, or reductions in scores over time, have been shown to predict falls, disability, and mortality (16–19). SPPB scores were excluded from the analysis for 56 EDIC participants from one clinical center due to incorrect administration of the 4-meter gait speed test.

The SPPB assessment and scoring were developed in a community-based cohort of nondisabled adults over the age of 70 years (16,17). In anticipation of possible ceiling effects in EDIC (78.1% of EDIC participants were <65 years of age), participants who successfully completed the SPPB balance measures were asked to attempt a more challenging 30-s unipedal stand test of balance (20). Up to three attempts to stand unaided on one foot for 30 s were permitted, and the test was scored as successful (held stance for 30 s) or not (held stance for <30 s).

The secondary outcome, presence of self-reported functional limitations, was assessed by written questionnaire with use of items from the National Health and Nutrition Examination Survey (NHANES) (21). Self-reported functional limitations were classified across five functional domains: activities of daily living (ADLs), instrumental activities of daily living (IADLs), leisure and social activities, lower-extremity mobility, and general physical function (21–24). A domain-specific functional limitation was considered present if a question for at least one item within the domain was answered as “some difficulty,” “much difficulty,” or “unable to do” and absent if the questions for all items in the domain were answered as “no difficulty”. Responses of “do not do this activity,” “prefer not to answer,” and “don’t know” were treated as missing.

Demographic and clinical factors expected to influence functional limitations among EDIC participants were selected a priori based on a literature review. Demographic factors were current age, sex, and level of formal education. Clinical factors (glycemic characteristics, nonglycemic characteristics, micro- and macrovascular complications, and study design characteristics) were measured using methods that have been previously described (10,13–15). Glycemic characteristics included diabetes duration, history of severe hypoglycemia (coma or seizure), and three measures of glycemia: the time-weighted mean DCCT/EDIC HbA_1c (weighted mean HbA_1c from DCCT baseline to EDIC year 25/26), mean EDIC HbA_1c (mean annual HbA_1c from the start of EDIC to follow-up year 25/26), and current HbA_1c level (EDIC year 25/26). Nonglycemic characteristics were measured concurrently with functional limitation assessment and included heavy alcohol use (≥27 g/day), smoking status, BMI (kg/m², continuous and categorical), cognitive function, and psychological distress. Cognitive function was assessed with the Montreal Cognitive Assessment (MoCA) (25). Psychological distress (including depression) was defined according to Symptom Checklist-90-Revised Global Severity Index (GSI) score ≥63 (26). Micro- and macrovascular complications were assessed with standardized methods throughout DCCT and EDIC (10,13–15). Proliferative diabetic retinopathy was defined by neovascularization observed on centrally graded fundus photographs or self-reported and/or confirmed scatter photocoagulation at any time during DCCT/EDIC. Reduced estimated glomerular filtration rate (eGFR) was defined as a sustained eGFR <60 mL/min/1.73 m² (estimated with the Chronic Kidney Disease Epidemiology Collaboration equation) or chronic renal dialysis or transplant at any time during DCCT/EDIC. Diabetic peripheral neuropathy was defined as the presence of confirmed clinical neuropathy at the most recent full neurological assessment in EDIC year 13/14 (2006–2007). Cardiovascular autonomic neuropathy was defined according to R-R variation <15, R-R variation 15–19 plus Valsalva ratio ≤1.5, or a postural drop ≥10 mmHg in diastolic blood pressure at the most recent full autonomic neuropathy assessment in EDIC year 16/17 (2009–2010). Any nonfatal CVD included nonfatal myocardial infarction, stroke, angina, coronary revascularization, or congestive heart failure as reported by participants and supported by medical records and silent myocardial infarction identified on centrally evaluated electrocardiograms (obtained every other year during DCCT and annually during EDIC). All cardiovascular events were adjudicated and classified by the Mortality and Morbidity Review Committee (27). Study design characteristics, including DCCT treatment group (intensive vs. conventional) and DCCT study cohort (primary prevention vs. secondary intervention), were assessed to explore whether the effect of metabolic memory, observed for a number of outcomes during EDIC, might extend to measures of functional limitations (12).

Descriptive analyses were stratified by sex, as women typically bear a greater burden of functional limitations and disability and perform worse on objective measures of physical function than men (2,4,16,28,29). The Wilcoxon rank sum test (with continuity correction) was used to assess differences between women and men in quantitative measurements, and the χ² test (with Yates’ continuity correction) was used to assess sex differences in categorical measurements.

The associations between clinical factors (glycemic, nonglycemic, micro- and macrovascular complications, and study design) and functional limitation outcomes were assessed with logistic regression models minimally adjusted for age, sex, and education level (graduate school, college graduate, some college or trade school, or secondary school graduate or less). A separate multivariable logistic regression model was developed for SPPB <10 and for self-reported functional limitations with use of a backward elimination procedure. Each model was subsequently tested with use of forward selection (30). The final multivariable model retained the selected covariates significant at P < 0.05. The odds ratios (ORs) and 95% CIs are presented.

In a sensitivity analysis we investigated the robustness of the results with respect to the 91 (7.7%) otherwise active EDIC participants with incomplete or absent functional limitation outcome measures (Supplementary Fig. 1). Individuals missing both the SPPB and questionnaire data for reasons given as “unwillingness to participate,” “completion of tests unsafe,” or “physical/clinical condition makes completion of test not possible” were assigned SPPB score <10 and self-reported functional limitations as “present.” Those missing both SPPB and questionnaire data for any other reason were assigned SPPB score ≥10 and self-reported functional limitations as “absent.” Individuals with partial data (either the SPPB or questionnaire was completed, but not both) were categorized according to their available (nonmissing) data. Specifically, if the SPPB was completed but not the questionnaire, those with SPPB score <10 were assigned “present” self-reported functional limitations, while those with SPPB score ≥10 were assigned “absent” self-reported functional limitations. Conversely, if the questionnaire but not the SPPB was completed, those with “present” self-reported functional limitation were assigned SPPB <10 and those with “absent” self-reported functional limitations were assigned SPPB ≥10.

Both z and P values are reported for logistic regression models to further clarify the relative strengths of the associations. The absolute value of the z test, which is the regression coefficient divided by its SE, indicates the strength of the association of each exposure with the outcome. Larger absolute z values indicate stronger associations resulting in smaller P values. Results nominally significant at P < 0.05 (two sided) are reported. All analyses were performed with SAS 9.4 and R statistical software.

Data collected for the DCCT/EDIC study through 30 June 2017 are available to the public through the NIDDK Central Repository (https://repository.niddk.nih.gov/studies/edic/). Data collected in the current cycle (July 2017–June 2022) will be available within 2 years after the end of the funding cycle.

Characteristics of the 1,094 EDIC participants with stratification by sex are shown in Table 1. Men were slightly older than women (mean ± SD age 60.0 ± 6.4 vs. 59.0 ± 7.1 years, P = 0.018), but the proportions of men and women aged ≥65 years were similar. Sex differences were observed for heavy alcohol use, BMI categories, MoCA scores, current HbA_1c, and prevalence of peripheral neuropathy.

Functional limitations as assessed based on SPPB score <10 were present in 217 (21.2%) of EDIC participants (Table 2). Unipedal balance was assessed in the 867 participants (84.6%) who successfully completed the SPPB balance measures, and of these, ∼30% were unable to successfully hold the unipedal stand for 30 s. No sex differences were observed for the prevalence of SPPB score <10, for any component of the SPPB (balance, gait speed, repeated chair stands), or for the unipedal stand test (Table 2).

Self-reported functional limitations were reported by nearly half of participants (N = 522 [48.0%]). While the prevalence was higher among women than men, the difference was not significant (Table 2). General physical function domain limitations were reported by nearly all (N = 506 [96.9%]) participants with any self-reported functional limitations. Lower-extremity mobility limitations were reported by 218 participants (41.8%), IADL limitations by 190 (36.4%), ADL limitations by 176 (33.7%), and leisure and social limitations by 149 (28.5%). No sex differences were observed for the prevalence of self-reported functional limitations overall or for any domain.

Table 3 presents two separate logistic regression models, minimally adjusted for age, sex, and education, evaluating the associations of clinical factors (study design, glycemic, and nonglycemic and micro- and macrovascular complications) and SPPB score <10 (yes vs. no) and self-reported functional limitations (present vs. absent). Higher time-weighted mean DCCT/EDIC HbA_1c was most strongly associated with SPPB score <10 (z value = 5.73, OR 1.69 [95% CI 1.41, 2.02]; P value <0.001). All historical measures of severe hypoglycemia (≥1 vs. 0 cumulative events, 1–5 events vs. none, and >5 vs. none) were associated only with SPPB score <10. Cardiac autonomic neuropathy had the strongest association with self-reported functional limitations, (z value = 6.06, OR 2.43 [95% CI 1.82, 3.24]; P value <0.001). There was no significant association between DCCT treatment group and either outcome. DCCT cohort (secondary intervention vs. primary prevention) was associated with SPPB <10 (OR 1.83 [95% CI 1.34, 2.51]; P value <0.001) but not with self-reported functional limitations (OR 1.28 [95% CI 0.98, 1.66]; P value = 0.067).

Table 4 presents multivariable models predicting SPPB score <10 and self-reported functional limitations. Clinical factors associated with both outcomes were higher time-weighted mean DCCT/EDIC HbA_1c, higher BMI, psychological distress (GSI ≥63), and presence of cardiac autonomic neuropathy. Higher BMI was most strongly associated with SPPB score <10 (z value = 4.34), with each additional kg/m² unit of BMI associated with a 7% increased odds of SPPB score <10 (OR 1.07 [95% CI 1.04, 1.10]; P value <0.001). Psychological distress (GSI ≥63) had the strongest association with any self-reported functional limitations, with more than a fourfold increased odds of self-reported functional limitations (z value = 5.44, OR 5.57 [95% CI 3.00, 10.34]; P value <0.001). The odds of SPPB score <10 remained greater for the DCCT secondary intervention cohort than the primary prevention cohort in the multivariable model (OR 1.75 [95% CI 1.21, 2.52]; P value = 0.003).

Supplementary Table 1 compares the characteristics of the 1,094 EDIC participants who completed the outcome assessments with those of the 91 EDIC participants who did not. Those who did not complete the assessments were more likely to have been assigned to DCCT conventional therapy and have higher mean HbA_1c levels during DCCT and EDIC, lower BMI, a greater prevalence of microvascular and cardiovascular complications and were more likely to report being unemployed due to disability. The prevalence of SPPB score <10 as imputed for those with missing or incomplete outcome data was proportionately similar to the prevalence among those who completed both measures, and the prevalence of self-reported functional limitations as imputed for those with missing or incomplete outcome data was higher among those who completed both measures. Incorporating the imputed values for the participants with incomplete or missing outcome data into the multivariable model did not affect the results of the multivariable analysis (data not shown).

In this cross-sectional study we examined the prevalence of and identified demographic and clinical factors associated with functional limitations in >1,000 middle-aged and older adults with long-standing type 1 diabetes. The primary functional limitation outcome was SPPB score <10, while any self-reported functional limitation was the secondary outcome. With logistic regression analyses we evaluated the association of clinical factors, assessed across 30 years of DCCT and EDIC evaluations, with SPPB score <10 and self-reported functional limitations.

SPPB score <10, indicating lower-extremity functional limitations, was present in approximately one-fifth of EDIC participants, with no difference by sex. A considerably higher prevalence of SPPB score <10 (43% of nondisabled 70-year-old men and nearly 60% of nondisabled 70 year old women) was reported in the Established Populations for Epidemiologic Studies of the Elderly (EPESE) study (16). More age-appropriate comparison groups were reported from a population-based study in Norway, where investigators found SPPB score <10 in 1% of men and 2% of women aged 40–59 years, 4% of men and 5% of women aged 60–69 years, and 7% of men and 15% of women aged 70–74 years. These rates are considerably lower than those observed in EDIC. In that study, it was further noted that while SPPB scores were generally lower in women than in men across all age groups, SPPB score <12 was uncommon in women younger than 65 years of age and in men younger than 70 years of age (28). Only 22% of EDIC participants were >65 years of age at the time of the study, suggesting the EDIC cohort is still too young for observation of a sex disparity for SPPB score <10.

Participants who successfully completed the SPPB balance tests attempted a more challenging 30-s unipedal stand test. Nearly 30% of participants attempting the unipedal stand test were unsuccessful. An inability to balance, unaided, on one leg for 30 s has been described as abnormal and shown to be a powerful predictor of falls in adults in their mid-60s (20).

The prevalence of self-reported functional limitations in EDIC was 48%. As with the SPPB, no difference was observed by sex. While this prevalence is lower than the 50%–77% prevalence reported by others using NHANES physical function data, those populations were nearly a decade older than the EDIC cohort (22–24). The prevalence of self-reported functional limitations among women (51.0%) and men (45.3%) in EDIC is similar to those reported in slightly older populations of non-Hispanic White women (54.2%, mean ± SD age 64.6 ± 0.1 years) and men (42.1%, age 63.0 ± 0.1 years) without diabetes and substantially lower than those reported for non-Hispanic White women (79.3%, age 67.0 ± 0.2 years) and non-Hispanic White men (66.9%, age 66.0 ± 0.1 years) with diabetes (31). The prevalence of domain-specific functional limitations appears higher in EDIC as compared with older adults without diabetes but lower when compared with older adults in the general population with diabetes (23). The hierarchy of self-reported functional limitations in EDIC (highest prevalence observed in the domain of general physical function, followed by lower-extremity mobility, IADL, and ADL, and lowest in the leisure and social activity domain) mirrors what has been reported by others (22,23).

Concurrent and longitudinal measures of HbA_1c and diabetes duration were associated with both functional limitation outcomes in separate logistic regression models. History of severe hypoglycemia was associated only with SPPB score <10. Higher time-weighted mean DCCT/EDIC HbA_1c was associated with SPPB score <10 and self-reported functional limitations, and longer diabetes duration was associated with self-reported functional limitations only. These findings support our hypothesis of an association between long-term poor glycemic control and functional limitations and are consistent with the results of others who have noted associations between higher HbA_1c levels and/or longer duration of diabetes with functional limitations (1,4,5,23,32). To our knowledge, this is the first report of an association of a history of severe hypoglycemia with an objective measure of physical function.

Higher BMI, psychological distress as assessed by GSI ≥63, and lower total MoCA score were associated with both outcomes in separate logistic models. In multivariable analyses, higher BMI and GSI ≥63 were significantly associated with both SPPB score <10 and self-reported functional limitations, while lower MoCA scores were significantly associated with SPPB score <10. Obesity (BMI >30 kg/m²), which was present in 38.7% of EDIC participants, is consistently recognized as a risk factor for functional limitations and has been shown to partially account for the excess risk of functional limitations associated with diabetes (4). Furthermore, obesity may promote or exacerbate other comorbidities, such as hypertension, heart disease, kidney disease, and arthritis, all of which may impact physical function (1,22,29,33,34). Our findings align with those of others who have reported associations of depression, anxiety, and cognitive impairment with functional outcomes (1,35–37).

The presence of micro- and macrovascular complications was associated with both outcomes in separate logistic regression models. In multivariable models, a history of nonfatal CVD, diabetic peripheral neuropathy, and cardiac autonomic neuropathy was associated with SPPB score <10. Cardiac autonomic neuropathy was also associated with self-reported functional limitations. Peripheral neuropathy has been identified by others as a risk factor for reduced physical performance, especially lower-extremity function, and for disability (1,38). This relationship is not surprising as diminished sensation, impaired proprioception, altered weight bearing, muscle wasting, and weakness may all adversely affect lower-extremity strength and balance. The association of cardiac autonomic neuropathy with the measured outcomes may reflect aberrant heart rate and blood pressure responses to postural changes and exercise. The association of severe hypoglycemia with SPPB score <10 may be further evidence of an association between autonomic dysfunction (hypoglycemia unawareness) and reduced physical function. The association of measures of autonomic dysfunction and functional limitations is a novel finding requiring confirmation.

DCCT cohort (secondary intervention vs. primary prevention) was associated with SPPB score <10 but not with self-reported functional limitations. The DCCT secondary intervention cohort was distinguished by their longer diabetes duration and observable (vs. absent) microvascular complications at DCCT randomization (10). DCCT cohort therefore may be a surrogate, albeit less robust, measure of diabetes duration. DCCT treatment group (intensive vs. conventional) was not associated with either outcome, suggesting that the effect of the 2% separation in HbA_1c between the intensive and conventional treatment groups during the DCCT has been overshadowed by essentially equivalent glucose control during 26 years of EDIC follow-up, consistent with the waning of metabolic memory over time (12). Indeed, within a few years of EDIC follow-up, the glycemic separation observed between the former intensive and conventional groups had vanished, and nearly equivalent levels of glycemia have been observed since (39).

The main strengths of this study are the well-characterized population of middle-aged and older adults with long-standing type 1 diabetes, the precise measures of diabetes duration and glucose control, excellent retention, and careful characterization of micro- and macrovascular complications over ∼30 years of DCCT and EDIC follow-up. The potential bias introduced by the 91 EDIC participants who did not complete the outcome assessments was addressed through sensitivity analysis. When we used imputed outcomes for these 91 participants, the results were unchanged (data not shown).

There are several limitations to this study, including the lack of an appropriate comparator population without diabetes. The EDIC cohort is still young relative to individuals usually included in studies of functional limitations. Further, the EDIC cohort is composed entirely of people with type 1 diabetes, thus limiting comparisons of differences in age, diabetes type, and diabetes duration to published reports of functional limitations. The EDIC cohort is racially and ethnically homogeneous (∼95% non-Hispanic White) and is well educated (64% with college or advanced degrees) and thus not typical of the average population with type 1 diabetes, which limits the generalizability of our findings. As such, results from EDIC participants may reflect higher physical functioning than the average population with type 1 diabetes. Additional studies should be conducted in other populations with type 1 diabetes to validate these findings. The primary outcome measure, SPPB <10, may not be sufficiently sensitive given the relatively young age of the EDIC cohort. The questionnaire used includes questions about limitations due to “any long-term physical, mental or emotional problem or illness” and is not specific with regard to diabetes and its complications. We did not consider the effect of nondiabetes-related comorbidities or the number of comorbidities, as has been done by others (22,23). For these reasons, our findings must be interpreted cautiously. Still, this study provides an important foundation for longitudinal assessments addressing the interactions of aging and type 1 diabetes.

In this cross-sectional study we leveraged phenotypic data collected over >30 years of follow-up from the DCCT/EDIC study to identify factors associated with functional limitations in >1,000 middle-aged and older adults with long-standing type 1 diabetes. While many of the clinical factors associated with functional limitations are consistent with findings from other studies, including glycemia, BMI, psychological and cognitive function, peripheral neuropathy, and CVD, novel findings from this study include an association of severe hypoglycemia and cardiac autonomic neuropathy with measures of functional limitations. These findings suggest that glucose control, body weight, and mental health could all be targets for risk reduction strategies.

To our knowledge, this is the first study to examine functional limitations in an exclusively type 1 diabetes cohort, providing the necessary foundation for ongoing and planned longitudinal studies of risk factors and predictors of incident disability, frailty, falls, and fractures. This work should guide future studies in evaluating older subjects in different contexts and with varying life histories to fully understand how type 1 diabetes affects physical functioning. The findings from this and future studies may inform strategies aimed at reducing the burden and consequences of functional limitations in individuals aging with type 1 diabetes.

Funding. DCCT/EDIC (clinical trial reg. nos. NCT00360815 and NCT00360893, ClinicalTrials.gov) has been supported by cooperative agreement grants (1982–1993, 2012–2017, 2017–2022) and contracts (1982–2012) with the Division of Diabetes Endocrinology and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases (current grant nos. U01 DK094176 and U01 DK094157) and by the National Eye Institute, the National Institute of Neurologic Disorders and Stroke, the General Clinical Research Centers program (1993–2007), and Clinical and Translational Science Center program (2006–present), Bethesda, MD. Industry contributors have provided free or discounted supplies or equipment to support participants’ adherence to the study: Abbott Diabetes Care (Alameda, CA), Animas (Westchester, PA), Bayer Diabetes Care (North America Headquarters, Tarrytown, NY), Becton, Dickinson and Company (Franklin Lakes, NJ), Eli Lilly (Indianapolis, IN), Extend Nutrition (St. Louis, MO), Insulet Corporation (Bedford, MA), LifeScan (Milpitas, CA), Medtronic Diabetes (Minneapolis, MN), Nipro Home Diagnostics (Ft. Lauderdale, FL), Nova Diabetes Care (Billerica, MA), OMRON (Shelton, CT), Perrigo Diabetes Care (Allegan, MI), Roche Diabetes Care (Indianapolis, IN), and Sanofi (Bridgewater, NJ).

The sponsor of this study is represented by the NIDDK Project Scientist who serves as part of the DCCT/EDIC Research Group and plays a part in the study design and conduct as well as the review and approval of manuscripts. The NIDDK Project Scientist was not a member of the writing group of this paper. The opinions expressed are those of the investigators and do not necessarily reflect the views of the funding agencies.

Industry contributors have had no role in the DCCT/EDIC study.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.L.M., V.R.T., and J.-Y.C.B. wrote the manuscript. V.R.T. and J.-Y.C.B. conducted the statistical analyses. C.L.M., V.R.T., J.-Y.C.B., P.L., B.H.B., I.B., J.M.L., A.M.J., R.G.-K., and W.H.H. wrote sections of the manuscript and reviewed and edited the manuscript. B.H.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.