Skip to main content
NCBI home page
The American Journal of Clinical Nutrition logoLink to The American Journal of Clinical Nutrition
. 2021 Apr 7;114(1):16–28. doi: 10.1093/ajcn/nqab025

Sex-specific 25-hydroxyvitamin D threshold concentrations for functional outcomes in older adults: PRoject on Optimal VItamin D in Older adults (PROVIDO)

Michelle Shardell 1,2,✉, Anne R Cappola 3, Jack M Guralnik 4, Gregory E Hicks 5, Stephen B Kritchevsky 6, Eleanor M Simonsick 7, Luigi Ferrucci 8, Richard D Semba 9, Nancy Chiles Shaffer 10, Tamara Harris 11, Gudny Eiriksdottir 12, Vilmundur Gudnason 13, Mary Frances Cotch 14, Eric Orwoll 15, Kristine E Ensrud 16, Peggy M Cawthon 17
PMCID: PMC8246604  PMID: 33826696

ABSTRACT

Background

Threshold serum 25-hydroxyvitamin D [25(OH)D] concentrations for extraskeletal outcomes are uncertain and could differ from recommendations (20–30 ng/mL) for skeletal health.

Objectives

We aimed to identify and validate sex-specific threshold 25(OH)D concentrations for older adults’ physical function.

Methods

Using 5 large prospective, population-based studies—Age, Gene/Environment Susceptibility-Reykjavik (n = 4858, Iceland); Health, Aging, and Body Composition (n = 2494, United States); Invecchiare in Chianti (n = 873, Italy); Osteoporotic Fractures in Men (n = 2301, United States); and Study of Osteoporotic Fractures (n = 5862, United States)—we assessed 16,388 community-dwelling adults (10,376 women, 6012 men) aged ≥65 y. We analyzed 25(OH)D concentrations with the primary outcome (incident slow gait: women <0.8 m/s; men <0.825 m/s) and secondary outcomes (gait speed, incident self-reported mobility, and stair climb impairment) at median 3.0-y follow-up. We identified sex-specific 25(OH)D thresholds that best discriminated incident slow gait using machine learning in training data (2/3 cohort-stratified random sample) and validated using the remaining (validation) data and secondary outcomes.

Results

Mean age in the cohorts ranged from 74.4 to 76.5 y in women and from 73.3 to 76.6 y in men. Overall, 1112/6123 women (18.2%) and 494/3937 men (12.5%) experienced incident slow gait, 1098/7011 women (15.7%) and 474/3962 men (12.0%) experienced incident mobility impairment, and 1044/6941 women (15.0%) and 432/3993 men (10.8%) experienced incident stair climb impairment. Slow gait was best discriminated by 25(OH)D <24.0 ng/mL compared with 25(OH)D ≥24.0 ng/mL in women (RR: 1.29; 95% CI: 1.10, 1.50) and 25(OH)D <21.0 ng/mL compared with 25(OH)D ≥21.0 ng/mL in men (RR: 1.43; 95% CI: 1.01, 2.02). Most associations between 25(OH)D and secondary outcomes were modest; estimates were similar between validation and training datasets.

Conclusions

Empirically identified and validated sex-specific threshold 25(OH)D concentrations for physical function for older adults, 24.0 ng/mL for women and 21.0 ng/mL for men, may inform candidate reference concentrations or the design of vitamin D intervention trials.

Keywords: physical function, older adults, vitamin D, extraskeletal health, gait speed

Introduction

Observational studies have shown that higher concentrations of serum 25-hydroxyvitamin D [25(OH)D], a biomarker of vitamin D body stores, relate to better physical function in older adults (1–3). Of interest is the hypothesized role of vitamin D in gait speed, dubbed the “sixth vital sign,” given its strong association with disability and survival at older ages (4, 5). A mechanistic role for vitamin D to preserve gait speed is plausible because gait speed depends in part on muscle strength, and the active vitamin D metabolite 1,25-dihydroxyvitamin D [1,25(OH)2D] binds to vitamin D receptors, which are present on skeletal muscle cells (6), leading to protein synthesis and muscle cell growth (7, 8). Because 25(OH)D is the substrate of 1,25(OH)2D, higher 25(OH)D concentrations may contribute to higher 1,25(OH)2D activity in muscle cells.

Despite promising observational studies, randomized vitamin D trials of extraskeletal outcomes have been inconclusive or null (9–12). Explanations proposed for discrepancies between observational studies and randomized trials often focus on limitations of the former, such as residual confounding of lifestyle (e.g., sunlight exposure, diet) or biodemographic factors (e.g., age, sex, skin color). However, vitamin D trials can also have limitations such as enrolling participants with sufficient baseline 25(OH)D concentrations who might not benefit from vitamin D intervention, which could have contributed to null results. Vitamin trials in general that enrolled participants with sufficient micronutrient concentrations have been criticized (13). A design challenge for vitamin D trials is that definitions for sufficient 25(OH)D concentrations for physical function lack consensus despite having been studied (14–17).

The Institute of Medicine (IoM, now the National Academy of Medicine) and Endocrine Society each have recommended different serum 25(OH)D concentrations for skeletal health based, in part, on observational data. The IoM recommended population concentrations ≥20 ng/mL from integrated evidence for skeletal outcomes, and the Endocrine Society recommended patient concentrations ≥30 ng/mL based on parathyroid hormone (18, 19). However, both groups agreed that evidence was insufficient to endorse 25(OH)D concentrations for extraskeletal outcomes, including physical function (20). Assay manufacturers and large laboratories (e.g., DiaSorin, LabCorp, Quest Diagnostics) often cite reference ranges for clinical practice that guide treatment and have defined insufficiency as <20 ng/mL or <30 ng/mL, per the recommendations (21). Over 33% of older adults have 25(OH)D concentrations between 20 and 30 ng/mL, making the debate crucial to resolve (22). Despite the null vitamin D trials (9–11), lack of consensus for 25(OH)D has led to uncertainty as to whether the trials have included enough participants with 25(OH)D low enough to benefit from intervention. Thus, we aimed to empirically identify and validate threshold 25(OH)D concentrations predictive of physical function decline in older adults.

Methods

Participating cohorts

Five cohorts participated in the PRoject on Optimal VItamin D in Older adults (PROVIDO) Consortium that included community-dwelling adults with measured 25(OH)D and longitudinally measured gait speed. Participating cohorts were: Age, Gene/Environment Susceptibility-Reykjavik (AGES); Health, Aging, and Body Composition (Health ABC); Invecchiare in Chianti (InCHIANTI); Osteoporotic Fractures in Men (MrOS); and Study of Osteoporotic Fractures (SOF) (for cohort descriptions see Supplemental Method 1). Included participants were aged ≥65 y at baseline; black or white race; and had measured baseline 25(OH)D, gait speed, and complete covariates (Table 1). Data transfer and analysis occurred between July 2015 and October 2020. Physical function was assessed from September 1992 through August 2011 (Table 1). This study was approved for use of deidentified data by the University of Maryland, Baltimore Institutional Review Board, and need for informed consent was waived.

TABLE 1.

Cohort characteristics by sex1

AGES Health ABC InCHIANTI SOF (women)/MrOS (men)
Women
 Total women enrolled, n (1)2 3326 1584 811 9704
 Black or white women aged ≥65 y at baseline, n (2)2 3326 1551 655 8354
 Missing 25(OH)D measurement, n (3)2 165 122 94 2122
 Missing baseline covariates, n (4)2 384 174 79 370
 Black or white women aged ≥65 y in our study, n (5) = (2) − (3) − (4)2 2777 1255 482 5862
 Years of baseline visit 2002 to 2006 1998 to 1999 1998 to 2000 1992 to 1994
 Years of follow-up visit 2007 to 2011 2001 to 2002 2001 to 2003 1997 to 1998
Men
 Total men enrolled, n (6)2 2438 1491 642 5994
 Black or white men aged ≥65 y at baseline, n (7)2 2438 1447 500 5605
 Missing 25(OH)D measurement, n (8)2 80 83 56 3219
 Missing baseline covariates, n (9)2 277 125 53 85
 Black or white men aged ≥65 y in our study, n (10) = (7) − (8) − (9)2 2081 1239 391 2301
 Years of baseline visit 2002 to 2006 1998 to 1999 1998 to 2000 2000 to 2002
 Years of follow-up visit 2007 to 2011 2001 to 2002 2001 to 2003 2005 to 2006
Open in a new tab

1AGES, Age, Gene/Environment Susceptibility-Reykjavik Study; Health ABC, Health, Aging, and Body Composition Study; InCHIANTI, Invecchiare in Chianti Study; MrOS, Osteoporotic Fractures in Men Study; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

2

Numbers in parentheses indicate cohort subsets that impact the final number of participants who contributed data to the present project. Row (5) is the final number of women who contributed data and row (10) is the final number of men who contributed data.

Primary outcome

Usual gait speed (meters per second) (walking at “usual” pace) was measured on 4-m (InCHIANTI) or 6-m (all other cohorts) courses. We used a published calibration equation to convert measured 6-m gait speed to 4-m gait speed (4-m speed = −0.0341 + 0.9816 × 6-m speed) (5). Incident slow gait (women: <0.8 m/s; men: <0.825 m/s) assessed at a follow-up visit was the primary outcome. Gait speed ≥0.8 m/s is an operational definition of community ambulation (23), where the higher threshold for men was used to address sex differences in height based on data from NHANES III (1988–1994) (Supplemental Figure 1 and Supplemental Method 2 include detailed methods).

Secondary outcomes

Secondary outcomes were continuous usual gait speed calibrated to a 4-m course, incident mobility impairment [self-reported difficulty walking 2–3 blocks (MrOS and SOF), one-quarter mile (Health ABC), 400 m (InCHIANTI), or 500 m (AGES)] and stair climb impairment (self-reported difficulty climbing 10 steps). Self-reported measures of physical function been have validated in previous work (24, 25). Supplemental Method 1 describes assessments in further detail.

Measurement of 25(OH)D

Cohorts measured 25(OH)D using RIA (Health ABC, InCHIANTI—DiaSorin, Inc.), chemiluminescent immunoassay (AGES—DiaSorin, Inc.), or liquid chromatography-tandem mass spectrometry (LC-MS/MS) (MrOS, SOF). We calibrated measures to RIA using published equations for analysis to more easily compare thresholds with published recommended concentrations. Specifically, IoM and Endocrine Society recommendations were based on DiaSorin RIA-measured 25(OH)D data in NHANES III and NHANES 1999–2006 (18, 19). The groups purposefully did not normalize data to ensure that their respective recommendations incorporated all sources of uncertainty (18). Thus, calibrating assays to RIA allowed more direct comparison with IoM and Endocrine Society recommendations. After we identified 25(OH)D thresholds (described in Statistical analysis below), we calibrated these thresholds and those recommended by the IoM and Endocrine Society to the Vitamin D Standards Program (VDSP) using 3 published VDSP equations developed from NHANES. These equations converted from RIA to LC-MS/MS (26) (Supplemental Method 3 describes laboratory procedures, calibration, and conversion to VDSP).

Measurement of covariates

We harmonized variable definitions across cohorts. Medical conditions (myocardial infarction, falls within the previous year, stroke, congestive heart failure, angina, diabetes mellitus, hypertension, and osteoporosis) were assessed using self-report, medications, and clinical measurement (Supplemental Method 4 describes algorithms). Demographics included age (years), race (white or black; self-report), marital status, and education. Lifestyle factors included protein, dairy, and alcohol intake assessed by FFQ; smoking status; and physical activity assessed by self-report. Cohorts used validated instruments to measure global cognition and depressive symptoms. Cognition instruments were transformed to a 0-to-100 scale for harmonization. Instrument-specific thresholds for depressive symptoms defined depressed mood (Supplemental Method 5 describes harmonization). Self-rated health was on a Likert scale. Other covariates included systolic blood pressure (millimeters mercury) and BMI (kg/m2). Baseline gait speed was also a covariate.

Statistical analysis

Overview of statistical analysis

We stratified all analyses by sex. Descriptive statistics included cohort- and sex-specific means and SDs for continuous variables and number (percentage) for categorical variables. We randomly split each cohort into training and validation datasets in the recommended 2:1 ratio (27). We used weighting as a method of standardization to adjust for covariates, cohort, study site, and visit month (28–30). We additionally used inverse-probability weighting to address missing data (28, 29). We constructed weights for standardization and missing data using machine-learning methods via the SuperLearner package in R Statistical Software (31) version 3.6.1 (Supplemental Method 6 details methods). For incident slow gait, mobility impairment, and stair climb impairment, we restricted all analyses to participants who were at risk of the respective outcome (i.e., without the outcome at baseline). All tests were 2-sided, and we considered P values <0.05 or 95% CIs excluding the null to be statistically significant. The validation dataset was half the size of the training dataset, thus analyses of validation data emphasized replicability and qualitative agreement rather than statistical significance. We conducted all analyses using R Statistical Software version 3.6.1.

Statistical analysis to identify 25(OH)D thresholds

We used weighted classification and regression trees to identify sex-specific 25(OH)D thresholds that best discriminated (most accurately classified) incident slow gait using the weights constructed to adjust for covariates and address missing data (32). We estimated prediction SEs using 10-fold cross-validation and selected the most parsimonious tree with prediction error within 1 SE of the tree with smallest average prediction error (32).

Statistical analysis to validate 25(OH)D thresholds

We performed 2-stage restricted maximum-likelihood meta-analysis to evaluate threshold performance. We quantified between-cohort relative heterogeneity using I2, which we addressed using random effects, and performed Cochran Q test (33, 34). We computed cohort-specific estimates using linear or modified Poisson marginal structural models by regressing primary and secondary outcomes on 25(OH)D thresholds; we used the inverse-density and inverse-probability weights adjusted for covariates and selective attrition (28, 29). We quantified uncertainty of threshold performance using 95% CIs. We also used meta-analysis to quantify performance of thresholds recommended by the IoM (20 ng/mL) and Endocrine Society (30 ng/mL). The emphasis of secondary outcomes is to assess consistency in direction with the primary outcome and consistency in magnitude of association between training and validation data. We performed meta-analysis methods using the metafor package in R Statistical Software version 3.6.1 (35).

Sample size calculation for incident slow gait

We used training data meta-analysis results for incident slow gait as effect sizes to compute sex-specific sample sizes for an explanatory randomized vitamin D intervention trial to prevent slow gait based on a 2-sided test with 5% type I error and 90% power. We used random-effects meta-analyses to compute sex-specific incidence of slow gait among those with low 25(OH)D (the hypothetical “control” group”) and used these values along with sex-specific RRs of incident slow gait based on the identified thresholds to compute the number needed to enroll in a trial. After computing the number needed to enroll, we also computed the number needed to be screened based on sex-specific 25(OH)D distributions (Supplemental Method 7).

Sensitivity analysis

We performed a sensitivity analysis to assess the robustness of study findings. Specifically, we performed instrumental variable (IV) random-effects meta-analysis of slow gait to address potential unmeasured confounding. Season of 25(OH)D assessment was the IV owing to seasonality of 25(OH)D and plausibility that season of baseline 25(OH)D assessment and follow-up slow gait are not confounded. We used an extended IV method to address potential seasonality of gait speed (36) (Supplemental Method 8 includes details).

Subgroup analysis

We used subgroup random-effects meta-analysis to test interactions between covariates and 25(OH)D thresholds. Prespecified subgroups were covariates that might influence 25(OH)D and physical function (22, 37, 38): age [78+ compared with 65–77 y (lower 3 quartiles)], BMI (25+ compared with <25), and any comorbid conditions (angina, myocardial infarction, congestive heart failure, diabetes, osteoporosis, or stroke). We performed subgroup analysis for race within Health ABC.

Results

After excluding participants with missing 25(OH)D measurement (n = 5941) or baseline covariates (n = 1547), analysis included 16,388 black or white adults aged ≥65 y from 5 population-based prospective cohort studies (Table 1). In women, mean within-cohort age ranged from 74.4 to 76.5 y, and mean 25(OH)D ranged from 18.0 to 25.2 ng/mL (Figure 1, Table 2). In men, mean within-cohort age ranged from 73.3 to 76.6 y, and mean 25(OH)D ranged from 21.1 to 26.5 ng/mL (Figure 1, Table 3). InCHIANTI participants had a lower proportion of high-school graduation, higher proportion of depressed mood, and poorer self-rated health compared with participants in other cohorts (Tables 2 and 3). Baseline characteristics were similar between training and validation datasets (Supplemental Tables 1 and 2).

FIGURE 1.

FIGURE 1

Open in a new tab

Participant flow from baseline to the analytic sample and incident slow gait at follow-up. Women (top) and men (bottom). AGES, Age, Gene/Environment Susceptibility-Reykjavik; Health ABC, Health, Aging, and Body Composition; InCHIANTI, Invecchiare in Chianti; MrOS, Osteoporotic Fractures in Men; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

TABLE 2.

Participant characteristics by cohort in women aged ≥65 y1

AGES (n = 2777) Health ABC (n = 1255) InCHIANTI (n = 482) SOF (n = 5862)
Characteristic Mean ± SD or n (%) Mean ± SD or n (%) Mean ± SD or n (%) Mean ± SD or n (%)
Age, y 76.4 ± 5.6 74.5 ± 2.9 74.4 ± 6.8 76.5 ± 4.6
White race 2777 (100.0) 695 (55.4) 482 (100.0) 5862 (100.0)
High-school graduate 1976 (71.2) 990 (78.9) 33 (6.8) 4724 (80.6)
Marital status
 Never married 181 (6.5) 71 (5.7) 30 (6.2) 452 (7.7)
 Formerly married(widowed/divorced/separated) 1286 (46.3) 702 (55.9) 213 (44.2) 2979 (50.8)
 Currently married 1310 (47.2) 482 (38.4) 239 (49.6) 2431 (41.5)
Physical activity
 Sedentary 1562 (56.2) 662 (52.7) 99 (20.5) 3271 (55.8)
 Moderately active 872 (31.4) 529 (42.2) 369 (76.6) 1829 (31.2)
 Highly active 343 (12.4) 64 (5.1) 14 (2.9) 762 (13.0)
Smoking status
 Never smoker 1475 (53.1) 726 (57.8) 394 (81.7) 3628 (61.9)
 Former smoker 946 (34.1) 422 (33.6) 47 (9.8) 1921 (32.8)
 Current smoker 356 (12.8) 107 (8.5) 41 (8.5) 313 (5.3)
Protein intake, g/d 38.8 ± 10.1 61.3 ± 24.7 69.9 ± 19.0 40.5 ± 18.1
Dairy intake, servings/d 1.18 ± 0.58 1.41 ± 1.19 2.56 ± 1.26 1.91 ± 1.70
Alcohol intake, drinks/d 0.10 ± 0.22 0.18 ± 0.49 0.53 ± 0.78 0.19 ± 0.44
Cognition (proportion of maximum) 0.90 ± 0.09 0.91 ± 0.08 0.83 ± 0.11 0.94 ± 0.07
Depressed mood 354 (12.7) 49 (3.9) 182 (37.8) 568 (9.7)
Self-rated health
 Very poor/poor 174 (6.3) 22 (1.8) 39 (8.1) 13 (0.2)
 Fair 813 (29.3) 179 (14.3) 186 (38.6) 86 (1.5)
 Good 829 (29.9) 519 (41.4) 225 (46.7) 942 (16.1)
 Very good/excellent 961 (34.6) 535 (42.6) 32 (6.6) 4821 (82.2)
Systolic blood pressure, mmHg 141.8 ± 20.4 136.3 ± 21.0 151.5 ± 20.0 133.4 ± 18.9
BMI, kg/m2 27.2 ± 4.8 27.5 ± 5.5 27.9 ± 4.6 26.4 ± 4.5
Hypertension 1988 (71.6) 774 (61.7) 347 (72.0) 3023 (51.6)
Congestive heart failure 12 (0.4) 8 (0.6) 9 (1.9) 123 (2.1)
Angina 68 (2.4) 32 (2.5) 11 (2.3) 212 (3.6)
Myocardial infarction 324 (11.7) 90 (7.2) 15 (3.1) 401 (6.8)
Diabetes mellitus 282 (10.2) 152 (12.1) 48 (10.0) 381 (6.5)
Stroke 157 (5.7) 29 (2.3) 14 (2.9) 278 (4.7)
Osteoporosis 890 (32.0) 276 (22.0) 134 (27.8) 1558 (26.6)
Fallen in past 12 mo 562 (20.2) 327 (26.1) 130 (27.0) 1750 (29.9)
Hip fracture 49 (1.8) 14 (1.1) 14 (2.9) 191 (3.3)
Serum 25(OH)D,2 ng/mL 19.2 ± 7.7 25.2 ± 12.2 18.0 ± 14.2 20.5 ± 13.1
Baseline 4-m gait speed,3 m/s 0.84 ± 0.20 1.08 ± 0.21 0.93 ± 0.24 0.90 ± 0.22
Baseline mobility impairment 757 (27.3) 263 (21.2) 104 (21.6) 900 (15.4)
Baseline stair climb impairment 929 (33.5) 285 (23.0) 114 (23.7) 737 (12.6)
Open in a new tab
1

Characteristics based on available data; not all characteristics were available on all participants. Supplementary Method 1 details missingness. AGES, Age, Gene/Environment Susceptibility-Reykjavik Study; Health ABC, Health, Aging, and Body Composition Study; InCHIANTI, Invecchiare in Chianti Study; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

2

Multiply by 2.496 to convert to nanomoles per liter.

3

Usual gait speed (m/s) (walking at “usual” pace) was measured on 4-m (InCHIANTI) or 6-m (all other cohorts) courses. We used a published calibration equation to convert measured 6-m to 4-m gait speed (4-m speed  =  −0.0341 + 0.9816 × 6-m speed) (5).

TABLE 3.

Participant characteristics by cohort among men aged ≥65 y1

AGES (n = 2081) Health ABC (n = 1239) InCHIANTI (n = 391) MrOS (n = 2301)
Characteristic Mean ± SD or n (%) Mean ± SD or n (%) Mean ± SD or n (%) Mean ± SD or n (%)
Age, y 76.6 ± 5.3 74.8 ± 2.8 73.3 ± 6.4 73.7 ± 5.8
White race 2081 (100.0) 804 (64.9) 391 (100.0) 2228 (96.8)
High-school graduate 1751 (84.1) 942 (76.0) 56 (14.3) 2153 (93.6)
Marital status
 Never married 87 (4.2) 63 (5.1) 27 (6.9) 66 (2.9)
 Formerly married(widowed/divorced/separated) 398 (19.1) 258 (20.8) 47 (12.0) 334 (14.5)
 Currently married 1596 (76.7) 918 (74.1) 317 (81.1) 1901 (82.6)
Physical activity
 Sedentary 1104 (53.1) 393 (31.7) 35 (9.0) 772 (33.6)
 Moderately active 590 (28.4) 730 (58.9) 320 (81.8) 1209 (52.5)
 Highly active 387 (18.6) 116 (9.4) 36 (9.2) 320 (13.9)
Smoking status
 Never smoker 588 (28.3) 371 (29.9) 120 (30.7) 829 (36.0)
 Former smoker 1257 (60.4) 741 (59.8) 189 (48.3) 1389 (60.4)
 Current smoker 236 (11.3) 127 (10.3) 82 (21.0) 83 (3.6)
Protein intake, g/d 41.5 ± 10.3 73.5 ± 32.5 82.8 ± 19.4 65.1 ± 27.2
Dairy intake, servings/d 1.16 ± 0.56 1.39 ± 1.27 2.41 ± 1.03 1.60 ± 1.07
Alcohol intake, drinks/d 0.22 ± 0.42 0.47 ± 0.93 1.93 ± 1.97 0.24 ± 0.21
Cognition (proportion of maximum) 0.88 ± 0.09 0.90 ± 0.08 0.87 ± 0.09 0.93 ± 0.06
Depressed mood 229 (11.0) 40 (3.2) 56 (14.3) 195 (8.5)
Self-rated health
 Very poor/poor 79 (3.8) 13 (1.0) 21 (5.4) 3 (0.1)
 Fair 498 (23.9) 178 (14.4) 79 (20.2) 25 (1.1)
 Good 673 (32.3) 429 (34.6) 227 (58.1) 273 (11.9)
 Very good/excellent 831 (39.9) 619 (50.0) 64 (16.4) 2000 (86.9)
Systolic blood pressure, mmHg 143.1 ± 20.0 134.5 ± 20.4 148.5 ± 18.4 138.8 ± 19.0
BMI, kg/m2 26. ± 3.8 27.0 ± 3.9 27.1 ± 3.3 27.5 ± 3.7
Hypertension 1403 (67.4) 688 (55.5) 266 (68.0) 1367 (59.4)
Congestive heart failure 29 (1.4) 18 (1.5) 10 (2.6) 46 (2.0)
Angina 71 (3.4) 41 (3.3) 16 (4.1) 60 (2.6)
Myocardial infarction 594 (28.5) 193 (15.6) 23 (5.9) 332 (14.4)
Diabetes mellitus 297 (14.3) 201 (16.2) 49 (12.5) 231 (10.0)
Stroke 130 (6.2) 24 (1.9) 21 (5.4) 128 (5.6)
Osteoporosis 222 (10.7) 87 (7.0) 51 (13.0) 155 (6.7)
Fallen in past 12 mo 302 (14.5) 252 (20.3) 49 (12.5) 461 (20.0)
Hip fracture 20 (1.0) 10 (0.8) 8 (2.0) 32 (1.4)
Serum 25(OH)D,2 ng/mL 21.1 ± 7.8 26.5 ± 9.8 23.9 ± 14.3 23.1 ± 9.0
Baseline 4-m gait speed,3 m/s 0.92 ± 0.20 1.19 ± 0.23 1.09 ± 0.26 1.15 ± 0.22
Baseline mobility impairment 399 (19.2) 189 (15.3) 49 (12.5) 251 (10.9)
Baseline stair climb impairment 438 (21.0) 147 (11.9) 56 (14.3) 151 (6.6)
Open in a new tab
1

Characteristics based on available data; not all characteristics were available on all participants. Supplementary Method 1 details missingness. AGES, Age, Gene/Environment Susceptibility-Reykjavik Study; Health ABC, Health, Aging, and Body Composition Study; InCHIANTI, Invecchiare in Chianti Study; MrOS, Osteoporotic Fractures in Men; 25(OH)D, 25-hydroxyvitamin D.

2

Multiply by 2.496 to convert to nanomoles per liter.

3

Usual gait speed (m/s) (walking at “usual” pace) was measured on 4-m (InCHIANTI) or 6-m (all other cohorts) courses. We used a published calibration equation to convert measured 6-m to 4-m gait speed (4-m speed  =  −0.0341 + 0.9816 × 6-m speed) (5).

A follow-up visit occurred a median of 3.0 y (IQR: 2.4–4.9 y) after baseline. Overall, 1112/6123 (18.2%) of women and 494/3937 (12.5%) of men experienced incident slow gait, 1098/7011 (15.7%) of women and 474/3962 (12.0%) of men experienced incident mobility impairment, and 1044/6941 (15.0%) of women and 432/3993 (10.8%) of men experienced incident stair climb impairment.

Identifying sex-specific 25(OH)D thresholds

Among women in the training dataset to assess incident slow gait (n = 4076), 25(OH)D <24.0 ng/mL [below-threshold 25(OH)D: 2701 (66.3%)] compared with 25(OH)D ≥24.0 ng/mL [above-threshold 25(OH)D: 1375 (37.3%)] best predicted incident slow gait (RR: 1.29; 95% CI: 1.10, 1.50) (Figure 2). Similarly, among men in the training dataset to assess incident slow gait (n = 2638), 25(OH)D <21.0 ng/mL [below-threshold 25(OH)D: 977 (37.0%)] compared with 25(OH)D ≥21.0 ng/mL (above-threshold 25(OH)D: 1661 (63.0%)] best predicted incident slow gait (RR: 1.43; 95% CI: 1.01, 2.02) (Figure 2). Scatterplots with locally estimated scatterplot smoothing (LOESS) fits demonstrate that these thresholds are located on cohorts’ slopes of sigmoid-shaped relations of 25(OH)D with incident slow gait and continuous gait, which all were monotonically decreasing (incident slow gait) or increasing (continuous gait) except for men in AGES (Supplemental Figures 2 and 3).

FIGURE 2.

FIGURE 2

Open in a new tab

Sex-specific 25(OH)D thresholds and incident slow gait. Random-effects meta-analysis was performed among training and validation datasets (left and right columns, respectively) and among women and men (top and bottom rows, respectively). RR refers to high 25(OH)D as the reference (≥24.0 ng/mL for women and ≥21.0 ng/mL for men); Q refers to Cochran Q test of between-cohort heterogeneity; I2 refers to Higgins I2 for proportion of variability explained by between-study heterogeneity. Cohort-specific adjusted RRs were estimated from marginal structural models; RR from all studies was estimated using 2-stage random-effects. The size of the black squares is proportionate to the weight of the cohort's contribution to the estimate from all studies. AGES, Age, Gene/Environment Susceptibility-Reykjavik; Health ABC, Health, Aging, and Body Composition; InCHIANTI, Invecchiare in Chianti; MrOS, Osteoporotic Fractures in Men; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

Sample size requirements for a vitamin D intervention trial

Estimates were used to calculate required sample size to achieve 90% power for explanatory randomized vitamin D intervention trials to prevent slow gait. Estimates suggested that screening 3912 women is needed to enroll 2610 participants with 25(OH)D <24.0 ng/mL in a 2-group trial. Similar calculations suggest that screening 5376 men is needed to enroll 2052 participants with 25(OH)D <21.0 ng/mL (Supplemental Method 7 includes details).

Validating sex-specific 25(OH)D thresholds: primary outcome

In women, in the validation dataset, the RR of slow gait comparing below-threshold compared with above-threshold 25(OH)D was 1.35 (95% CI: 0.94, 1.93) (Figure 2), and the IV RR was 1.75 (95% CI: 1.06, 2.87) (Supplemental Table 3, Supplemental Figure 4). The RR using the 20-ng/mL threshold was of similar magnitude (RR: 1.29; 95% CI: 0.79, 2.10) to that using the identified 24.0-ng/mL threshold, but the RR using the 30-ng/mL threshold was smaller (RR: 1.09; 95% CI: 0.74, 1.59) (Supplemental Figure 5). Published equations estimated that VDSP-calibrated thresholds are 0–2 ng/mL higher than RIA-calibrated and previously published thresholds (Supplemental Method 3).

In men, the RR of slow gait comparing below-threshold with above-threshold 25(OH)D was 1.58 (95% CI: 0.87, 2.86) in the validation dataset (Figure 2); the IV RR was 1.43 (95% CI: 1.06, 1.92) (Supplemental Table 4, Supplemental Figure 6). The 20-ng/mL threshold performed similarly (RR: 1.66; 95% CI: 0.93, 2.96) to the identified 21.0-ng/mL threshold, but the RR using the 30-ng/mL threshold was smaller (RR: 1.27; 95% CI: 0.84, 1.93) (Supplemental Figure 7).

Performance of 25(OH)D thresholds did not meaningfully differ between training and validation datasets (Supplemental Figures 8 and 9); therefore, datasets were combined for subgroup analysis to overcome sparsity. Subgroup analysis found no statistically significant subgroup interactions among either sex (Supplemental Figures 10 and 11). Within Health ABC race subgroups, RRs for below-threshold compared with above-threshold 25(OH)D were 1.35 (95% CI: 0.81, 2.25) in white women, 1.33 (95% CI: 0.87, 2.05) in black women (P-interaction = 0.97), 2.71 (95% CI: 1.32, 5.58) in white men, and 1.43 (95% CI: 0.76, 2.69) in black men (P-interaction = 0.19).

Validating sex-specific 25(OH)D thresholds: secondary outcomes

Women with below-threshold 25(OH)D had slightly slower prospective gait speed (0.93 m/s; 95% CI: 0.89, 0.98 m/s) than women with above-threshold 25(OH)D (0.96 m/s; 95% CI: 0.90, 1.02 m/s) in the training dataset (mean difference = −0.024 m/s; 95% CI: −0.047, −0.001 m/s) (Figure 3). Estimates in the validation dataset were similar (mean difference = −0.033 m/s; 95% CI: −0.071, 0.005 m/s) (Figure 3). Estimated risks of incident mobility impairment and stair climb impairment were slightly higher in women with below-threshold compared with above-threshold 25(OH)D and were of similar magnitude between training and validation datasets, but not statistically significant (Figure 4, Supplemental Figure 12). LOESS fits demonstrated that the threshold was generally on cohort-specific negative slopes; however, not all relations of 25(OH)D with secondary outcomes were monotonically decreasing, such as those among InCHIANTI participants (Supplemental Figure 13).

FIGURE 3.

FIGURE 3

Open in a new tab

Sex-specific 25(OH)D thresholds and gait speed at follow-up. Random-effects meta-analysis was performed in training and validation datasets (left and right columns, respectively) and among women and men (top and bottom rows, respectively). MD refers to mean difference with high 25(OH)D as the reference (≥24.0 ng/mL for women and ≥21.0 ng/mL for men); Q refers to Cochran Q test of between-cohort heterogeneity; I2 refers to Higgins I2 for proportion of variability explained by between-study heterogeneity. Cohort-specific adjusted mean differences were estimated from marginal structural models; MD from all studies was estimated using 2-stage random-effects. The size of the black squares is proportionate to the weight of the cohort's contribution to the estimate from all studies. AGES, Age, Gene/Environment Susceptibility-Reykjavik; Health ABC, Health, Aging, and Body Composition; InCHIANTI, Invecchiare in Chianti; MD, mean difference; MrOS, Osteoporotic Fractures in Men; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

FIGURE 4.

FIGURE 4

Open in a new tab

Sex-specific 25(OH)D thresholds and incident mobility and stair climb impairment in the training dataset. Random-effects meta-analysis for mobility impairment and stair climb impairment (left and right columns, respectively) among women and men (top and bottom rows, respectively) in the training dataset (Supplemental Figure 12 displays the validation dataset). RR refers to high 25(OH)D as the reference (≥24.0 ng/mL for women and ≥21.0 ng/mL for men); Q refers to Cochran Q test of between-cohort heterogeneity; I2 refers to Higgins I2 for proportion of variability explained by between-study heterogeneity. Cohort-specific adjusted RRs were estimated from marginal structural models; RR from all studies was estimated using 2-stage random-effects. The size of the black squares is proportionate to the weight of the cohort's contribution to the estimate from all studies. AGES, Age, Gene/Environment Susceptibility-Reykjavik; Health ABC, Health, Aging, and Body Composition; InCHIANTI, Invecchiare in Chianti; MrOS, Osteoporotic Fractures in Men; SOF, Study of Osteoporotic Fractures; 25(OH)D, 25-hydroxyvitamin D.

Men with below-threshold 25(OH)D had slightly slower prospective gait speed (1.03 m/s; 95% CI: 0.96, 1.10 m/s) than men with above-threshold 25(OH)D (1.06 m/s; 95% CI: 0.97, 1.14 m/s) in the training dataset (mean difference = −0.022 m/s; 95% CI: −0.042, −0.002 m/s) (Figure 3). The validation dataset estimate was of larger magnitude (mean difference = −0.056 m/s; 95% CI: −0.121, 0.010 m/s) (Figure 3). The 25(OH)D threshold was not statistically significantly associated with incident mobility impairment; however, it was significantly associated with incident stair climb impairment in the training (RR: 1.38; 95% CI: 1.09, 1.74) and validation (RR: 1.53; 95% CI: 1.07, 2.18) datasets (Figure 4, Supplemental Figure 12). LOESS fits demonstrated that the threshold was generally on cohort-specific negative slopes; however, not all relations of 25(OH)D with incident mobility impairment were monotonically decreasing, such as some among AGES and Health ABC participants (Supplemental Figure 14).

Discussion

In this pooled analysis of 5 population-based cohorts, serum 25(OH)D thresholds of 24.0 ng/mL and 21.0 ng/mL best predicted incident slow gait speed in community-dwelling older women and men, respectively. Replicated directions and magnitudes of association with secondary outcomes, validation data, and IV analysis generally supported the thresholds, which performed similarly across prespecified subgroups. In contrast to many observational studies of 25(OH)D and physical function (2), estimated associations here were of modest magnitude.

The 24.0-ng/mL threshold for women is closer to the IoM recommendation (20 ng/mL) than the Endocrine Society recommendation (30 ng/mL) for skeletal health. Comparing the three, 24.0 ng/mL most strongly predicted incident slow gait and the IoM recommendation performed nearly as well. The 21.0-ng/mL threshold in men is so close to the IoM recommendation that their performances were nearly identical. For both sexes, 20 ng/mL outperformed 30 ng/mL. Per VDSP (Supplemental Method 3), adding an estimated 0 to 2 ng/mL to the thresholds can convert from RIA to LC-MS/MS. Previous efforts to identify 25(OH)D thresholds for physical function in older adults aged ≥65 y were most often cross-sectional, based on a single cohort, or modeled splines using analyst-specified knots or local regression with analyst visualization. One study estimated optimal 25(OH)D thresholds for cross-sectional physical performance of 20 ng/mL in men and 23 ng/mL in women, similar to thresholds here (14); other studies estimated optimal cross-sectional thresholds from 36 to 40 ng/mL (15–17) and longitudinal improvement in a 10–20-ng/mL category but not in a 20–30-ng/mL category (17).

Lack of consensus for recommended 25(OH)D concentrations could have contributed to heterogeneity in randomized vitamin D trial design. Consequently, vitamin D trials could have enrolled participants who lacked potential to benefit from intervention (9–12), a common criticism of vitamin trials (13). This issue is likely in recent trials due to temporal population increases in 25(OH)D concentrations (22). The United States Preventive Services Task Force (USPSTF) gave vitamin D supplements a B grade recommendation in 2012 to prevent falls in older adults and a D grade recommendation in 2018 (39–41). Unlike in 2012, the USPSTF in 2018 excluded trials solely comprising participants with deficient or insufficient 25(OH)D in primary analysis [and included trials with mean baseline 25(OH)D concentrations ≤31.8 ng/mL] to assess benefits and harms of population-wide vitamin D supplementation policies. Also, the USPSTF explicitly excluded trials of persons with vitamin D deficiency in its 2018 I (insufficient evidence) grade for vitamin D supplements >400 IU/d to prevent fractures in community-dwelling older adults (42). A null meta-analysis of vitamin D supplement trials for fractures included trials with mean baseline 25(OH)D concentrations ≤33.0 ng/mL (43), 2 recent prominent null trials of vitamin D supplements for preventing chronic conditions had mean baseline 25(OH)D concentrations of 28.0 ng/mL (44) and 30.8 ng/mL (45), and a meta-analysis on physical function included trials with participants whose mean (or upper range) baseline 25(OH)D was above thresholds identified herein or recommended by the IoM or Endocrine Society (11). Most recently, a null vitamin D trial recruited older adults with baseline 25(OH)D concentrations between 10 and 29 ng/mL and compared 200 IU/d (control) with higher doses (46). Notably, the control group had a mean 25(OH)D that increased from 22.1 ng/mL to 27.0 ng/mL in 3 mo, values that are close to thresholds here; increases were even greater at higher doses.

Different vitamin D intervention trial designs answer different questions, complicating their comparison with observational studies. Vitamin D trials of participants with sufficient 25(OH)D concentrations are best understood as pragmatic trials to test benefits and harms of vitamin D intervention policies as opposed to explanatory trials (47) to test biological efficacy of vitamin D. Explanatory trials are more comparable to observational studies than are pragmatic trials; however, an important barrier of comparison for functional outcomes is selecting participants who might benefit from treatment based on prerandomization 25(OH)D, which is hampered by the lack of consensus 25(OH)D thresholds for physical function. Thus, the biological efficacy of vitamin D interventions—whether supplements, sunlight exposure, or diet—to prevent declining physical function is uncertain. Another barrier to performing explanatory trials of vitamin D interventions for physical function is that such trials are resource intensive and participants in the control group often use vitamin D supplements. Although thresholds identified here can serve as eligibility criteria for explanatory trials, modest associations with slow gait and 25(OH)D distributions here suggest that such trials would be large and expensive, requiring screening ∼4000–5000 older adults to detect modest differences. Therefore, in the absence of such resource-intensive explanatory vitamin D trials for physical function or published VDSP-calibrated recommendations, 25(OH)D concentrations of 24.0 ng/mL for women and 21.0 ng/mL for men address the evidence gap for recommended 25(OH)D for physical function and perform similar to or better than the IoM recommendation of 20 ng/mL.

This study's strengths include multiple large, well-characterized cohorts of diverse geography and baseline health. Rigorous individual-participant analysis overcame publication bias (48), included adjustment for multiple covariates including baseline gait speed, and addressed unmeasured confounding, selective attrition, and reverse causality. These features might have led to the more modest findings than in previous observational studies. Also, randomly allocating participants into training and validation datasets demonstrated replicable effect sizes. Although this study's strengths address previously noted gaps (9, 10), it has limitations such as low racial diversity, cohort differences in assessment, single 25(OH)D measurement that differed between cohorts, and relative heterogeneity from geographic diversity that reduced power. Although we calibrated 25(OH)D using published equations, measurement error and heterogeneity likely contributed to conservative estimates. As in any observational study, residual unmeasured confounding is likely; thus, causality cannot be assumed. We addressed these limitations to some extent by using 2-stage modeling with random effects to first derive within-cohort estimates, performing subgroup analyses of race in Health ABC, using consistent variable definitions and calibration equations, performing IV analysis, and analyzing individual participant data. We a priori stratified by sex to address the importance of sex as a biological variable; external validation is needed to determine whether the sex-specific thresholds are clinically meaningfully different.

In conclusion, empirically identified and validated sex-specific threshold serum 25(OH)D concentrations for physical function in community-dwelling older adults were 24.0 ng/mL for women and 21.0 ng/mL for men. These thresholds could inform about candidate reference 25(OH)D concentrations for physical function or trials of vitamin D interventions (which can take the form of supplements, sunlight, and/or dietary intake) for modest potential impact on physical function, an important extraskeletal outcome. Additional external validation of these thresholds in diverse populations, including VDSP conversion, is warranted.

Supplementary Material

nqab025_Supplemental_File
Click here for additional data file. (2.5MB, docx)

Acknowledgements

The authors’ responsibilities were as follows—MS, ARC, JMG, GEH, SBK, LF, RDS, NCS, PMC: designed the research; JMG, SBK, EMS, LF, RDS, TH, GE, VG, EO, KEE, PMC, MFC: conducted the research; JMG, SBK, EMS, LF, TH, GE, VG, EO, KEE, PMC, MFC: provided data access; MS: performed statistical analysis; MS, ARC, JMG, PMC, MFC: wrote the paper; MS: had primary responsibility for final content; and all authors: read and approved the final manuscript.

The authors report no conflicts of interest.

Notes

The Age, Gene/Environment Susceptibility-Reykjavik (AGES) Study was supported by National Institute on Aging (NIA) contract N01-AG-12100 from the NIA Intramural Research Program, the Icelandic Parliament, and Hjartavernd (Icelandic Heart Association) with funding for vitamin D data collection provided by the National Eye Institute Intramural Research Program (ZIAEY000401). The Health, Aging, and Body Composition (Health ABC) Study is supported by NIA contracts N01-AG-6-2101, N01-AG-6-2103, and N01-AG-6-2106 from the NIA Intramural Research Program. Assessment of 25-hydroxyvitamin D concentrations was funded by NIA grant R01-AG029364. The Invecchiare in Chianti (InCHIANTI) Study is supported by the Italian Ministry of Health and in part by NIH contracts 263 MD 9164 13, 263 MD 821336, N01-AG-1-1, N01-AG-1-2111, and N01-AG-5-000 from the NIA Intramural Research Program. The Osteoporotic Fractures in Men (MrOS) Study is supported by grants from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) (U01 AR45580, U01 AR45614, U01 AR45632, U01 AR45647, U01 AR45654, U01 AR45583), the NIA (U01 AG18197, U01-AG027810), the National Center for Research Resources (NCRR) (UL1 RR024140), and the NIH Roadmap for Medical Research. The Study of Osteoporotic Fractures (SOF) Study is supported by grants from NIAMS (R01 AR35582, R01 AR35583, R01 AR35584) and the NIA (R01 AG005407, R01 AG005394, R01 AG027574, and R01 AG027576). Funding for MS was provided in part by NIA grants R01 AG048069, R56 AG068673, and P30 AG028747-15S1.

The funding organizations were independent of the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation of the manuscript. Prior to submission for publication, the manuscript was reviewed and approved by the National Institute on Aging.

Supplemental Methods 1–8, Supplemental Figures 1–14, and Supplemental Tables 1–4 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

Abbreviations used: AGES, Age, Gene/Environment Susceptibility-Reykjavik; Health ABC, Health, Aging, and Body Composition; InCHIANTI, Invecchiare in Chianti; IoM, Institute of Medicine; IV, instrumental variable; LC-MS/MS, liquid chromatography-tandem mass spectrometry; LOESS, locally estimated scatterplot smoothing; MD, mean difference; MrOS, Osteoporotic Fractures in Men; SOF, Study of Osteoporotic Fractures; USPSTF, United States Preventive Services Task Force; VDSP, Vitamin D Standards Program; 1,25(OH)2D, 1,25-dihydroxyvitamin D; 25(OH)D, 25-hydroxyvitamin D.

Contributor Information

Michelle Shardell, Institute for Genome Sciences, University of Maryland School of Medicine, Baltimore, MD, USA; Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, MD, USA.

Anne R Cappola, University of Pennsylvania School of Medicine, Philadelphia, PA, USA.

Jack M Guralnik, Department of Epidemiology and Public Health, University of Maryland School of Medicine, Baltimore, MD, USA.

Gregory E Hicks, Department of Physical Therapy, University of Delaware, Newark, DE, USA.

Stephen B Kritchevsky, Sticht Center for Healthy Aging and Alzheimer's Prevention, Wake Forest School of Medicine, Winston-Salem, NC, USA.

Eleanor M Simonsick, National Institute on Aging, Bethesda, MD, USA.

Luigi Ferrucci, National Institute on Aging, Bethesda, MD, USA.

Richard D Semba, Wilmer Eye Institute, Johns Hopkins University School of Medicine, Baltimore, MD, USA.

Nancy Chiles Shaffer, National Institute on Aging, Bethesda, MD, USA.

Tamara Harris, National Institute on Aging, Bethesda, MD, USA.

Gudny Eiriksdottir, Icelandic Heart Association, Reykjavik, Iceland.

Vilmundur Gudnason, Icelandic Heart Association, Reykjavik, Iceland.

Mary Frances Cotch, National Eye Institute, Intramural Research Program, Division of Epidemiology and Clinical Applications, Bethesda, MD, USA.

Eric Orwoll, Oregon Health & Science University, Portland, OR, USA.

Kristine E Ensrud, University of Minnesota Department of Medicine and Division of Epidemiology, Minneapolis, MN, USA.

Peggy M Cawthon, California Pacific Medical Center Research Institute, San Francisco, CA, USA.

Data Availability

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval from the publications committees of the participating cohorts.

References

  • 1. Houston DK, Neiberg RH, Tooze JA, Hausman DB, Johnson MA, Cauley JA, Bauer DC, Shea MK, Schwartz GG, Williamson JDet al. Low 25-hydroxyvitamin D predicts the onset of mobility limitation and disability in community-dwelling older adults: the Health ABC Study. J Gerontol A Biol Sci Med Sci. 2013;68(2):181–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Annweiler C, Henni S, Walrand S, Montero-Odasso M, Duque G, Duval GT. Vitamin D and walking speed in older adults: systematic review and meta-analysis. Maturitas. 2017;106:8–25. [DOI] [PubMed] [Google Scholar]
  • 3. Shardell M, Semba RD, Kalyani RR, Hicks GE, Bandinelli S, Ferrucci L. Serum 25-hydroxyvitamin D, plasma klotho, and lower-extremity physical performance among older adults: findings from the InCHIANTI study. J Gerontol A Biol Sci Med Sci. 2015;70(9):1156–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Fritz S, Lusardi M. White paper: “walking speed: the sixth vital sign.” J Geriatr Phys Ther. 2009;32(2):46–9. [PubMed] [Google Scholar]
  • 5. Studenski S, Perera S, Patel K, Rosano C, Faulkner K, Inzitari M, Brach J, Chandler J, Cawthon P, Connor EBet al. Gait speed and survival in older adults. JAMA. 2011;305(1):50–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Simpson RU, Thomas GA, Arnold AJ. Identification of 1,25-dihydroxyvitamin D3 receptors and activities in muscle. J Biol Chem. 1985;260(15):8882–91. [PubMed] [Google Scholar]
  • 7. Wassner SJ, Li JB, Sperduto A, Norman ME. Vitamin D deficiency, hypocalcemia, and increased skeletal muscle degradation in rats. J Clin Invest. 1983;72(1):102–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Mosekilde L. Vitamin D and the elderly. Clin Endocrinol. 2005;62(3):265–81. [DOI] [PubMed] [Google Scholar]
  • 9. Autier P, Mullie P, Macacu A, Dragomir M, Boniol M, Coppens K, Pizot C, Boniol M. Effect of vitamin D supplementation on non-skeletal disorders: a systematic review of meta-analyses and randomised trials. Lancet Diabetes Endocrinol. 2017;5(12):986–1004. [DOI] [PubMed] [Google Scholar]
  • 10. Autier P, Boniol M, Pizot C, Mullie P. Vitamin D status and ill health: a systematic review. Lancet Diabetes Endocrinol. 2014;2(1):76–89. [DOI] [PubMed] [Google Scholar]
  • 11. Rosendahl-Riise H, Spielau U, Ranhoff AH, Gudbrandsen OA, Dierkes J. Vitamin D supplementation and its influence on muscle strength and mobility in community-dwelling older persons: a systematic review and meta-analysis. J Hum Nutr Diet. 2017;30(1):3–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Bischoff-Ferrari HA, Vellas B, Rizzoli R, Kressig RW, da Silva JAP, Blauth M, Felson DT, McCloskey EV, Watzl B, Hofbauer LCet al. Effect of vitamin D supplementation, omega-3 fatty acid supplementation, or a strength-training exercise program on clinical outcomes in older adults: the DO-HEALTH randomized clinical trial. JAMA. 2020;324(18):1855–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Morris MC, Tangney CC. A potential design flaw of randomized trials of vitamin supplements. JAMA. 2011;305(13):1348–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Sohl E, de Jongh RT, Heymans MW, van Schoor NM, Lips P. Thresholds for serum 25(OH)D concentrations with respect to different outcomes. J Clin Endocrinol Metab. 2015;100(6):2480–8. [DOI] [PubMed] [Google Scholar]
  • 15. Bischoff-Ferrari HA, Dietrich T, Orav EJ, Hu FB, Zhang Y, Karlson EW, Dawson-Hughes B. Higher 25-hydroxyvitamin D concentrations are associated with better lower-extremity function in both active and inactive persons aged > or = 60 y. Am J Clin Nutr. 2004;80(3):752–8. [DOI] [PubMed] [Google Scholar]
  • 16. Bischoff-Ferrari HA, Giovannucci E, Willett WC, Dietrich T, Dawson-Hughes B. Estimation of optimal serum concentrations of 25-hydroxyvitamin D for multiple health outcomes. Am J Clin Nutr. 2006;84(1):18–28. [DOI] [PubMed] [Google Scholar]
  • 17. Wicherts IS, van Schoor NM, Boeke AJ, Visser M, Deeg DJ, Smit J, Knol DL, Lips P. Vitamin D status predicts physical performance and its decline in older persons. J Clin Endocrinol Metab. 2007;92(6):2058–65. [DOI] [PubMed] [Google Scholar]
  • 18. IOM (Institute of Medicine) . Dietary reference intakes for calcium and vitamin D. Washington (DC): National Academies Press; 2011. [PubMed] [Google Scholar]
  • 19. Holick MF, Binkley NC, Bischoff-Ferrari HA, Gordon CM, Hanley DA, Heaney RP, Murad MH, Weaver CM; Endocrine Society . Evaluation, treatment, and prevention of vitamin D deficiency: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2011;96(7):1911–30. [DOI] [PubMed] [Google Scholar]
  • 20. Rosen CJ, Adams JS, Bikle DD, Black DM, Demay MB, Manson JE, Murad MH, Kovacs CS. The nonskeletal effects of vitamin D: an Endocrine Society scientific statement. Endocr Rev. 2012;33(3):456–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Krasowski MD. Pathology consultation on vitamin D testing. Am J Clin Pathol. 2011;136(4):507–14. [DOI] [PubMed] [Google Scholar]
  • 22. Herrick KA, Storandt RJ, Afful J, Pfeiffer CM, Schleicher RL, Gahche JJ, Potischman N. Vitamin D status in the United States, 2011–2014. Am J Clin Nutr. 2019;110:150–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Abellan van Kan G, Rolland Y, Andrieu S, Bauer J, Beauchet O, Bonnefoy M, Cesari M, Donini LM, Gillette Guyonnet S, Inzitari Met al. Gait speed at usual pace as a predictor of adverse outcomes in community-dwelling older people an International Academy on Nutrition and Aging (IANA) Task Force. J Nutr Health Aging. 2009;13(10):881–9. [DOI] [PubMed] [Google Scholar]
  • 24. Fried LP, Bandeen-Roche K, Williamson JD, Prasada-Rao P, Chee E, Tepper S, Rubin GS. Functional decline in older adults: expanding methods of ascertainment. J Gerontol A Biol Sci Med Sci. 1996;51(5):M206–14. [DOI] [PubMed] [Google Scholar]
  • 25. Fried LP, Ettinger WH, Lind B, Newman AB, Gardin J. Physical disability in older adults: a physiological approach. Cardiovascular Health Study Research Group. J Clin Epidemiol. 1994;47(7):747–60. [DOI] [PubMed] [Google Scholar]
  • 26. Schleicher RL, Sternberg MR, Lacher DA, Sempos CT, Looker AC, Durazo-Arvizu R, Yetley EA, Chaudhary-Webb M, Maw KL, Pfeiffer CMet al. A method-bridging study for serum 25-hydroxyvitamin D to standardize historical radioimmunoassay data to liquid chromatography-tandem mass spectrometry. Natl Stat Health Report2016;(93):1–16. [PubMed] [Google Scholar]
  • 27. Dobbin KK, Simon RM. Optimally splitting cases for training and testing high dimensional classifiers. BMC Med Genomics. 2011;4:31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11(5):550–60. [DOI] [PubMed] [Google Scholar]
  • 29. Shardell M, Hicks GE, Ferrucci L. Doubly robust estimation and causal inference in longitudinal studies with dropout and truncation by death. Biostatistics. 2015;16(1):155–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Sato T, Matsuyama Y. Marginal structural models as a tool for standardization. Epidemiology. 2003;14(6):680–6. [DOI] [PubMed] [Google Scholar]
  • 31. Polley EC, van der Laan MJ. SuperLearner: super learner prediction, package version 2.0-9. 2011. [Google Scholar]
  • 32. Breiman L, Friedman JH, Stone CJ, Olshen RA. Classification and regression trees. Boca Raton (FL): CRC Press; 1984. [Google Scholar]
  • 33. Borenstein M, Higgins JP, Hedges LV, Rothstein HR. Basics of meta-analysis: I(2) is not an absolute measure of heterogeneity. Res Syn Meth. 2017;8(1):5–18. [DOI] [PubMed] [Google Scholar]
  • 34. Higgins JP. Commentary: heterogeneity in meta-analysis should be expected and appropriately quantified. Int J Epidemiol. 2008;37(5):1158–60. [DOI] [PubMed] [Google Scholar]
  • 35. Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Soft. 2010;36(3):1–48. [Google Scholar]
  • 36. Shardell M, Ferrucci L. Instrumental variable analysis of multiplicative models with potentially invalid instruments. Statist Med. 2016;35(29):5430–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Shapses SA, Lee EJ, Sukumar D, Durazo-Arvizu R, Schneider SH. The effect of obesity on the relationship between serum parathyroid hormone and 25-hydroxyvitamin D in women. J Clin Endocrinol Metab. 2013;98(5):E886–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Shea MK, Houston DK, Tooze JA, Davis CC, Johnson MA, Hausman DB, Cauley JA, Bauer DC, Tylavsky F, Harris TBet al. Correlates and prevalence of insufficient 25-hydroxyvitamin D status in black and white older adults: the health, aging and body composition study. J Am Geriatr Soc. 2011;59(7):1165–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. U. S. Preventive Services Task Force, Grossman DC, Curry SJ, Owens DK, Barry MJ, Caughey AB, Doubeni CA, Epling JW Jr, Kemper AR, Krist AHet al. Interventions to prevent falls in community-dwelling older adults: US Preventive Services Task Force recommendation statement. JAMA. 2018;319(16):1696–704. [DOI] [PubMed] [Google Scholar]
  • 40. Guirguis-Blake JM, Michael YL, Perdue LA, Coppola EL, Beil TL. Interventions to prevent falls in older adults: updated evidence report and systematic review for the US Preventive Services Task Force. JAMA. 2018;319(16):1705–16. [DOI] [PubMed] [Google Scholar]
  • 41. Moyer VA, U.S. Preventive Services Task Force . Prevention of falls in community-dwelling older adults: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med. 2012;157(3):197–204. [DOI] [PubMed] [Google Scholar]
  • 42. U.S. Preventive Services Task Force, Grossman DC, Curry SJ, Owens DK, Barry MJ, Caughey AB, Doubeni CA, Epling JW Jr, Kemper AR, Krist AHet al. Vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults: US Preventive Services Task Force recommendation Statement. JAMA. 2018;319(15):1592–9. [DOI] [PubMed] [Google Scholar]
  • 43. Zhao JG, Zeng XT, Wang J, Liu L. Association between calcium or vitamin D supplementation and fracture incidence in community-dwelling older adults: a systematic review and meta-analysis. JAMA. 2017;318(24):2466–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44. Pittas AG, Dawson-Hughes B, Sheehan P, Ware JH, Knowler WC, Aroda VR, Brodsky I, Ceglia L, Chadha C, Chatterjee Ret al. Vitamin D supplementation and prevention of type 2 diabetes. N Engl J Med. 2019;381(6):520–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Manson JE, Cook NR, Lee IM, Christen W, Bassuk SS, Mora S, Gibson H, Gordon D, Copeland T, D'Agostino Det al. Vitamin D supplements and prevention of cancer and cardiovascular disease. N Engl J Med. 2019;380(1):33–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Appel LJ, Michos ED, Mitchell CM, Blackford AL, Sternberg AL, Miller 3rd ER, Juraschek SP, Schrack JA, Szanton SL, Charleston Jet al. The effects of four doses of vitamin D supplements on falls in older adults: a response-adaptive, randomized clinical trial. Ann Intern Med. [Internet]2020. doi:10.7326/M20-3812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Thorpe KE, Zwarenstein M, Oxman AD, Treweek S, Furberg CD, Altman DG, Tunis S, Bergel E, Harvey I, Magid DJet al. A pragmatic-explanatory continuum indicator summary (PRECIS): a tool to help trial designers. J Clin Epidemiol. 2009;62(5):464–75. [DOI] [PubMed] [Google Scholar]
  • 48. Ross JS. Confronting bias. JAMA Intern Med. 2015;175(2):307. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

nqab025_Supplemental_File
Click here for additional data file. (2.5MB, docx)

Data Availability Statement

Data described in the manuscript, code book, and analytic code will be made available upon request pending approval from the publications committees of the participating cohorts.

Articles from The American Journal of Clinical Nutrition are provided here courtesy of American Society for Nutrition

RESOURCES