Ankle-Brachial Index and Energy Production in People Without Peripheral Artery Disease: The BLSA.

Background Lower ankle-brachial index (ABI) values within the 0.90 to 1.40 range are associated with poorer mitochondrial oxidative capacity of thigh muscles in cross-sectional analyses. Whether ABI decline is associated with greater declines in thigh muscle oxidative capacity with aging is unknown. Method and Results We analyzed data from 228 participants (100 men) of the BLSA (Baltimore Longitudinal Study of Aging), aged 39 to 97 years, with an ABI between 0.9 and 1.40 at baseline and at follow-up (mean follow-up period of 2.8 years). We examined mitochondrial oxidative capacity of the left thigh muscle, by measuring the postexercise phosphocreatine recovery rate constant (kPCr) from phosphorus-31 magnetic resonance spectroscopy. Greater kPCr indicated higher mitochondrial oxidative capacity. Although kPCr was available on the left leg only, ABI was measured in both legs. Longitudinal rates of change (Change) of left and right ABI and kPCr of the left thigh muscle were estimated using linear mixed effects models, and their association was analyzed by standardized multiple linear regressions. In multivariate analysis including sex, age, baseline kPCr, both left and right baseline ABI, and ABI change in both legs, (kPCr)Change was directly associated with ipsilateral (left) (ABI)Change (standardized [STD]-β=0.14; P=0.0168) but not with contralateral (right) (ABI)Change (P=0.22). Adjusting for traditional cardiovascular risk factors, this association remained significant (STD-β=0.18; P=0.0051). (kPCr)Change was steeper in White race participants (STD-β=0.16; P=0.0122) and body mass index (STD-β=0.13; P=0.0479). There was no significant association with current smoking status (P=0.63), fasting glucose (P=0.28), heart rate (P=0.67), mean blood pressure (P=0.78), and low-density lipoprotein (P=0.75), high-density lipoprotein (P=0.82), or triglycerides (P=0.15). Conclusions In people without peripheral arterial disease, greater decline in ABI over time, but not baseline ABI, was associated with faster decline in thigh mitochondrial oxidative capacity in the ipsilateral leg. Further studies are needed to examine whether early interventions that improve lower extremity muscle perfusion can improve and prevent the decline of muscle energetics.

Baltimore Longitudinal Study of Aging

postexercise phosphocreatine recovery rate constant

What Is New?

Even among those without overt peripheral artery disease, people who experience greater decline in ankle‐brachial index with aging also experience greater decline in leg muscle energy production.

What Are the Clinical Implications?

Further studies are needed to examine whether early interventions that improve lower extremity muscle perfusion can improve and prevent the decline of muscle energetics.

Impaired walking endurance associated with peripheral artery disease (PAD), defined as an ankle‐brachial index (ABI) of <0.90, has been linked to lower mitochondrial oxidative capacity in the muscles of the lower extremities. This association may be attributable in part to the discrepancy between oxygen delivery and demand , as well as oxidative damage to myocytes from repeated episodes of ischemia‐reperfusion. , , Even a mildly lower ABI within the range of 0.9 to 1.4 is associated with lower performance in mobility tasks , , and with lower mitochondrial oxidative capacity, assessed by phosphorous‐31 magnetic resonance spectroscopy.

Longitudinal studies have found a decline in ABI with aging in the presence , , , and absence of PAD. Whether this age‐associated longitudinal decline in ABI is also associated with longitudinal reduction in muscle oxidative capacity is unknown. Hence, we hypothesized that, in participants who are free of PAD, lower ABI at baseline and greater decline in ABI with aging are associated with greater declines in mitochondrial oxidative capacity. If even subclinical decline of ABI is associated with reduced skeletal muscle oxidative capacity, interventions that ameliorate blood flow and muscle perfusion may prevent the decline of mitochondrial function with aging and its consequences.

Methods

Study Sample

Participants of the BLSA (Baltimore Longitudinal Study of Aging) are community‐dwelling volunteers who undergo 3 consecutive days of medical and physiological testing. Visits occur at intervals of 1 to 4 years and are progressively more frequent with older age.

A total of 572 repeated measures of postexercise phosphocreatine recovery rate constant (kPCr), ABI, and other clinical and laboratory measures were collected from 228 participants between 2013 and 2018. Six participants who reported a diagnosis of PAD and 6 who had ABI <0.90 of either leg were excluded from the analysis.

The Institutional Review Board of the Intramural Research Program of the National Institutes of Health granted ongoing Institutional Review Board approval to the BLSA, and all participants provided written, informed consent. Data and program code associated with this study are available on written request through the corresponding author.

Phosphorus‐31 Magnetic Resonance Spectroscopy

Thigh muscle mitochondrial oxidative capacity was assessed in the left leg only by magnetic resonance spectroscopy, as previously described. Briefly, a 3‐T magnetic resonance imaging scanner (Achieva; Philips Healthcare, Andover, MA) and a 10‐cm phosphorus‐31–tuned surface coil (PulseTeq; Surrey, UK) fastened over the left vastus lateralis muscle were used to acquire in vivo spectra of phosphorous‐containing metabolites. From a supine position in the bore of the magnet, participants performed a rapid ballistic knee extension exercise, which was first practiced before entering the magnet. , Before, during, and after the exercise, pulse‐acquired phosphorus‐31 spectra were obtained every 6 seconds, yielding a total of 75 spectra acquired over 7.5 minutes, of which the first 5 spectra were recorded at baseline. The pulse sequence incorporated an adiabatic radiofrequency excitation pulse with a 90‐degree flip angle, a repetition time of 1.5 seconds, and signal averaging over 4 successive acquisitions for signal/noise ratio enhancement. The length of exercise did not exceed 60 seconds, and the postrecovery period was at least 6 minutes. Exercise was concluded when a reduction in phosphocreatine (PCr) peak height of at least 25% from baseline was observed, with an optimal goal of 66%. If <25% PCr depletion was achieved, then the entire exercise study was repeated, if feasible, after a resting period of ≈10 minutes. Cases with <25% depletion were excluded from the analyses, and all models were statistically adjusted for PCr depletion during exercise to account for its effects on observed associations. Spectra were processed and metabolites were quantified using a publicly available software package (jMRUI) using a nonlinear least‐squares algorithm (AMARES). ,

Phosphocreatine recovery rate was determined by a fit of the postexercise time‐dependent PCr peak area to a monoexponential function of the form: where PCr0 is the end‐of‐exercise PCr peak area (ie, the PCr peak area at t=0, the beginning of the recovery period), ΔPCr is the decrease in peak area from its preexercise baseline value to PCr0 attributable to in‐magnet exercise, and τPCr is the PCr exponential recovery time constant.

The recovery rate constant, kPCr, is the inverse of τPCr (ie, kPCr=1/τPCr). There are few other energy demands during the resting period; therefore, kPCr is taken as in vivo muscle oxidative phosphorylation capacity, because with no or minimal contribution of anaerobic metabolism, kPCr resynthesis is primarily a function of maximum mitochondrial ATP production. , , , , Therefore, greater kPCr reflects shorter recovery time and thus higher mitochondrial oxidative capacity.

PCr depletion percentage was calculated as the decrease in PCr peak area from preexercise PCr baseline to PCr0. , To ensure intracellular pH did not drop below 6.8, intracellular pH was monitored and determined according to the chemical shift of inorganic phosphate relative to PCr whenever a rapid and deep PCr drop was observed and acidosis was suspected.

Several studies have validated the reproducibility of phosphorous‐31 magnetic resonance spectroscopy measurement of mitochondrial function in skeletal muscle, , , and the technique is particularly amenable to longitudinal studies. As suggested by the literature, to maximize reproducibility, we used a standardized data collection protocol that included continuity in equipment and technicians.

Blood Pressures and ABI

Bilateral ankle and brachial blood pressures were simultaneously measured 3 times using an oscillometric device (Colin VP‐2000; Omron Healthcare Inc, Kyoto, Japan). After averaging the triplicates, left and right ABIs were calculated as the ratio of the ankle to the brachial systolic pressures. Subjects who at any time during follow‐up had a left ABI <0.9 were excluded from this study.

Clinical Variables

Participants were classified as White or non‐White race based on self‐reported race. Height, weight, and heart rate were objectively measured according to standardized protocols. Body mass index was calculated as the ratio of weight to height squared. Left leg cross‐sectional images at midthigh were obtained using computed tomography (Somatom Sensation 10; Siemens, Malvern, PA). These images were processed to estimate muscle, subcutaneous fat, and intermuscular fat areas using commercial software (Genie 2.1; BonAlyse Oy, Jyvaskyla, Finland). Subjects were classified as being current smokers or not based on self‐report. Medications were grouped on the basis of World Health Organization Anatomical Therapeutic Chemical classification categories.

Laboratory Studies

At each visit, morning blood samples were taken from the antecubital vein after an 8‐hour overnight fast. Plasma triglyceride and total cholesterol concentrations were determined enzymatically by a standard clinical machine (ABA‐200ATC Biochromatic Analyzer; Abbott Laboratories, Irving, TX). High‐density lipoprotein cholesterol concentrations were determined via an established precipitation procedure. Low‐density lipoprotein (LDL) cholesterol concentrations were estimated using the Friedewald formula. Glucose concentration was measured via the glucose oxidase method using a standard instrument (Beckman Instruments, Inc, Fullerton, CA). CRP (C‐reactive protein) was measured via ELISA (Immundiagnostik AG, Bensheim, Germany). Erythrocyte sedimentation rate (ESR) was measured in a Sediplast tube (LP Italiano, Venice, Italy).

Statistical Analysis

Given the observed differences in ABI by sex, descriptive characteristics of the sample at baseline were reported and compared between men and women. For categorical variables, proportions were compared using χ2 statistics, whereas continuous variables were reported as mean±SD, and comparisons were made via Student t test. Given the relatively small sample size and the absence of significant sex differences in kPCr, subsequent analyses were not stratified by sex but included sex as a covariate and sex interaction terms to test for sex differences in associations.

Rates of change in left and right ABI and left thigh muscle kPCr were estimated by linear mixed effects models, which allow for unbalanced, unequally spaced observations, such as those of the BLSA. The general approach for using this method was described previously in detail. In summary, individual longitudinal rates of change (Change) were calculated by fitting separate linear mixed effects models for each variable, regressing the indexed variable against follow‐up time in years as a random effect (ie, the estimated individual variation from the population average slope was used to calculate the individual rate of change [Change]).

Multiple linear regression models were then used to examine the independent association between left and right (ABI)Change and left (kPCr)Change. The initial model fitted for (kPCr)Change as a function of independent variables, including sex, baseline age, baseline kPCr, PCr depletion, and left ABI, in addition to rates of change of PCr depletion and left ABI. The second model included right ABI and its rate of change (but not left ABI and its rate of change), whereas the third model adjusted for covariates and traditional cardiovascular risk factors, including pulse wave velocity, race, current smoking, fasting glucose, body mass index, mean arterial pressure, and blood lipids. In additional models, left leg composition and inflammatory markers and their changes over time, baseline medications, and LDLChange were explored as potential correlates of (kPCr)Change. Sex‐(ABI)Change interaction terms were included to assess sex differences.

To illustrate the association between ABI and kPCr, baseline left ABI range was divided into 2 groups (namely, borderline/low‐normal [0.90–1.10] and high‐normal [1.11–1.4]). Similarly, left (ABI)Change was categorized into 3 groups based on tertiles. Least‐squares adjusted means of (kPCr)Change (including variables in Table 2, model 1) were calculated for left baseline ABI and (ABI)Change groups and compared using ANCOVA followed by a post hoc Tukey test. For all analyses, P<0.05 was the threshold for significance, and all analyses were performed via SAS for Windows (Version 9.4; Cary, NC).

Table 2

Multiple Linear Regression Models Predicting (kPCr)Change and Considering Left ABI and Left (ABI)Change (Model 1) and Covariates and Cardiovascular Risk Factors (Model 2)

Variable	Model 1			Model 2
	β	STβ	P value	β	STβ	P value
Intercept	1.80E‐03	0.0000	0.0034	2.08E‐03	0.0000	0.0170
Initial age, y	−8.05E‐06	−0.1387	0.0136	−1.09E‐05	−0.1868	0.0036
Sex (men)	−5.44E‐05	−0.0486	0.3818	3.86E‐05	−0.0345	0.6130
kPCr, ms−1	−6.64E‐02	−0.5579	<0.0001	−6.94E‐02	−0.5897	<0.0001
PCr depletion, %	−1.65E‐06	−0.0274	0.7201	−4.19E‐06	−0.0716	0.3981
(PCr depletion)Change, %/y	−1.13E‐04	−0.2113	0.0064	−1.26E‐04	−0.2417	0.0041
Left ABI	2.77E‐04	0.0400	0.5229	2.26E‐04	0.0322	0.6201
Left (ABI)Change, y−1	9.81E‐02	0.1440	0.0168	1.19E‐02	0.1753	0.0051
(PWV)Change, m/s per y	…	…	…	2.04E‐04	0.1060	0.0943
Race (White)	…	…	…	1.93E‐04	0.1560	0.0122
Current smoker	…	…	…	−1.09E‐04	−0.0270	0.6346
Fasting glucose, mg/dL	…	…	…	−3.25E‐06	−0.0699	0.2820
BMI, kg/m2	…	…	…	1.76E‐05	0.1293	0.0479
HR, bpm	…	…	…	−1.25E‐06	−0.0255	0.6704
RB MAP, mm Hg	…	…	…	8.37E‐07	0.0166	0.7816
LDL, mg/dL	…	…	…	−3.49E‐07	−0.0191	0.7451
HDL, mg/dL	…	…	…	4.98E‐07	0.0167	0.8169
Triglycerides, mg/dL	…	…	…	−1.07E‐06	−0.0943	0.1526

ABI indicates ankle‐brachial index; BMI, body mass index; bpm, beats per minute; Change, longitudinal rates of change; HDL, high‐density lipoprotein; HR, heart rate; kPCr, postexercise phosphocreatine recovery rate constant; LDL, low‐density lipoprotein; PCr, phosphocreatine; PWV, pulse wave velocity; RB MAP, right brachial mean arterial pressure; and STβ, standardized β coefficient.

Results

The average baseline age of the study population was 75 years (range, 39–97 years). On average, participants had 3 visits and a mean follow‐up of 2.8 years (SD, 1.2 years) (Table 1.) Of participants, 70% were White race and 2.2% were current smokers. There were no differences between men and women in mean arterial pressure, triglyceride levels, left thigh muscle oxidative capacity (kPCr), phosphocreatine depletion, and CRP. However, compared with women, men had greater fasting glucose, body mass index, ABI, pulse wave velocity, and left leg lean/fat ratio but lower heart rate, LDL and high‐density lipoprotein levels, and ESR.

Table 1

Baseline Characteristics of the Study Population

Variable	Total (n=228)	Men (n=100)	Women (n=128)	P value
Initial age, y	75.0±9.5	75.8±8.9	74.3±10.0	0.2450
Follow‐up time, y	2.8±1.2	2.8±1.2	2.7±1.2	0.5642
No. of follow‐ups	2.5±0.8	2.6±0.9	2.5±0.8	0.5168
Race (White)	160 (70.2)	78 (78.0)	82 (64.1)	0.0284
Current smoker	5 (2.2)	4 (4.0)	1 (0.8)	0.1713
Fasting glucose, mg/dL	95.6±11.9	98.3±14.0	93.4±9.5	0.0046
BMI, kg/m2	26.6±4.2	27.6±3.9	25.9±4.3	0.0016
HR, bpm	70.7±11.5	68.4±11.2	72.4±11.4	0.0094
RB MAP, mm Hg	89.8±11.2	90.5±8.9	89.2±12.6	0.3680
LDL, mg/dL	95.9±30.7	88.5±27.7	101.7±31.8	0.0012
HDL, mg/dL	65.3±18.4	55.9±15.1	72.6±17.4	<0.0001
Triglycerides, mg/dL	92.2±49.9	100.0±62.3	86.2±36.8	0.0517
PWV, m/s	7.8±2.1	8.4±2.3	7.4±1.8	0.0005
kPCr, ms−1	0.0200±0.0047	0.0205±0.0048	0.0195±0.0045	0.1199
PCr depletion, %	39.7±9.2	39.8±9.4	39.7±9.0	0.9521
Left ABI	1.17±0.08	1.20±0.08	1.16±0.08	0.0003
Right ABI	1.17±0.08	1.19±0.08	1.16±0.08	0.0012
Left leg lean/fat ratio	1.89±0.95	2.68±0.87	1.27±0.40	<0.0001
CRP, mg/L	2.06±2.99	2.09±3.06	2.03±2.94	0.8869
ESR, mm/h	12.2±11.6	9.2±10.2	14.5±12.1	0.0005

Data are given as mean±SD or number (percentage). ABI indicates ankle‐brachial index; BMI, body mass index; bpm, beats per minute; CRP, C‐reactive protein; ESR, erythrocyte sedimentation rate; HDL, high‐density lipoprotein; HR, heart rate; kPCr, postexercise phosphocreatine recovery rate constant; LDL, low‐density lipoprotein; PCr, phosphocreatine; PWV, pulse wave velocity; and RB MAP, right brachial mean arterial pressure.

The mean and SD of (kPCr)Change was 0.0000568±0.0000556 ms‐1/y, while that of left (ABI)Change was 0.00163±0.0082/y relative to a mean intravisit left ABI SD of 0.0395. Figure 1 illustrates the bivariate association between kPCr and left ABI rates of change. In multivariate linear regression analysis, left (ABI)Change, but not baseline left ABI, was a significant independent predictor of (kPCr)Change (standardized β=0.1440 [P=0.0168] and standardized β=0.0400 [P=0.52], respectively; Table 2, model 1). (kPCr)Change was not significantly associated with baseline right ABI or right (ABI)Change (P=0.81 and P=0.22, respectively; Table S1, model 1). The association of left (ABI)Change with (kPCr)Change was independent of baseline right ABI and right (ABI)Change before (Table S1, model 2) and after adjusting for traditional cardiovascular risk factors (Table S1, model 3). The association of left (ABI)Change with (kPCr)Change was also independent of initial age, baseline kPCr, and (PCr depletion)Change (Table 2, model 1), and remained significant after adjusting for additional covariates, including traditional cardiovascular risk factors (Table 2, model 2). White race and higher body mass index were associated with greater increase in (kPCr)Change (standardized β=0.1560 [P=0.0122] and standardized β=0.1293 [P=0.0479], respectively). Baseline pulse wave velocity, current smoking, fasting glucose, heart rate, mean arterial pressure, and cholesterol levels were not significantly associated with (kPCr)Change. Also, left thigh muscle lean/fat composition ratio and CRP and their changes, baseline medications, and LDLChange were not associated with (kPCr)Change (Tables S2 through S5). ESRChange was directly associated with (kPCr)Change, but only after adjusting for covariates and cardiovascular risk factors (P=0.0495; Table S4, model 4). Left (ABI)Change remained an independent and significant predictor of (kPCr)Change after adjustment for all these potential confounders. Via multiple regression analysis, the longitudinal left ABI‐kPCr relationship had a power of 0.62. To illustrate the association between left ABI and kPCr, we plotted model‐adjusted means of (kPCr)Change (based on Table 2, model 1) for baseline left ABI categories (0.90–1.10 and 1.11–1.4) and left (ABI)Change tertiles. As shown in Table 2, there was no significant difference between the 2 baseline ABI groups (Figure 2A). However, those in the highest tertile of left (ABI)Change had significantly higher mean (kPCr)Change than those in the lowest tertile (Figure 2B).

Figure 1

Bivariate association between postexercise phosphocreatine recovery rate constant (kPCr) and left ankle‐brachial index (ABI) rates of change.

 

Figure 2

Postexercise phosphocreatine recovery rate constant (kPCr) mean rates of change for those with baseline left ankle‐brachial index (ABI) in the borderline/low‐normal (left) and normal (right) ranges (A) and kPCr mean rates of change for lower, middle, and upper tertile groups of left ABI rate of change (B). Least‐squares(LS).

*P<0.05 vs lower.

Discussion

Principal Findings

In this study, we found that among people without PAD, ABI longitudinal decline, but not baseline value, was associated with a decline in thigh muscle oxidative capacity of the same leg but not of the contralateral leg; this longitudinal association remained significant after adjusting for covariates and cardiovascular risk factors.

Linear Association Between ABI and Muscle Oxidative Capacity

This study is consistent with our previously published findings that, when compared with those with high‐normal ABI (1.10–1.40), individuals with modestly reduced ABI (0.90–1.09) are more likely to have lower thigh muscle mitochondrial oxidative capacity, as assessed by kPCr, in cross‐sectional analyses. These new results, however, add a longitudinal dimension that reinforces the idea that even subclinical reduction of ABI negatively impacts skeletal muscle oxidative capacity. The association appears linear, suggesting that declines in ABI affect oxidative capacity at any level within the normal ABI range studied herein, independent of initial ABI. However, further studies with a larger sample size may be needed to finally determine whether impairment of oxidative capacity only critically occurs below a certain ABI threshold.

The underlying mechanism of this association remains undefined, but these findings might indicate that low‐normal ABI is indicative of limited ability to increase blood flow during exercise, leading to reduced postexercise oxidative capacity. The significant association of (kPCr)Change with changes in ipsilateral but not contralateral ABI suggests that local flow‐limiting factors, such as structural plaques or impaired endothelial reactivity, may contribute to this phenomenon by limiting oxygen delivery to mitochondria. Such theories are also consistent with our prior work indicating that differences in muscle perfusion partly explain the age‐associated decline in muscle oxidative capacity. Interestingly, we found that the association of hemodynamics and muscle energetic changes was not accounted for by typical age‐associated changes, including changes in muscle/fat ratio. Further studies examining systemic and local contributors to this process, including changes in type I versus type II muscle fibers, femoral artery blood flow, and endothelial function with aging, are needed.

Indications of an Early Relationship Between ABI and Muscle Energetics

One of the interesting findings of this analysis is that the heterogeneity in ABI changes was concordant with that in muscle oxidative capacity; specifically, our analysis of the consistency between change in ABI and change in kPCr showed that participants in the lowest tertile (negative changes) and those in the highest tertile (positive changes) of ABI rate of change were more likely to have concordant changes in kPCr (Figure 2B).

Although we cannot make any causal statements based on an observational study, the changes observed in this analysis (ie, increasing and decreasing ABI with corresponding changes in kPCr) suggest a possible dynamic relationship between these 2 parameters. Such a dynamic relationship, if it exists, would not be explained only by flow‐limiting lesions, which tend to be fixed or progressive. However, more labile hemodynamic alterations that ensue with arterial aging, such as changes in pressure wave reflection and peripheral pulse pressure amplification, can contribute to the changes in ABI observed in this healthy cohort.

It is possible that the reduction in oxidative capacity observed in this study might be reversible. Furthermore, even fluctuations in perfusion may cause a vicious cycle of ischemia‐reperfusion injury that causes irreversible damage and in some individuals may evolve into clinically overt PAD (ie, ABI <0.90). , , However, these hypotheses could not be tested because of limitations in the number of follow‐up visits with complete data in this study. Hence, our findings call for further investigations, including clinical trials testing the hypothesis that increasing lower extremity perfusion in individuals with borderline/low‐normal ABI values may help preserve mitochondrial function and possibly prevent subclinical decline with aging.

Limitations

First, the association between change in ABI and change in oxidative capacity remained significant after adjusting for multiple potential confounders but was modest in magnitude. Thus, these findings should be interpreted with caution and should be replicated in an independent population and with longer follow‐up. Second, the BLSA is a study of healthy aging with well‐educated participants who have relatively high socioeconomic status. Thus, these findings may not extrapolate to the general population. Third, the BLSA is an observational study, and therefore, the associations observed herein are not sufficient to conclude causal relationships between hemodynamics and oxidative metabolism. Fourth, this study did not measure plaque burden, either proximally or distally, preventing the examination of its contribution to the associations observed. Fifth, a relationship between mitochondrial activity and functional performance is yet to be established. Sixth, the relatively small sample size and short longitudinal follow‐up time of our study might have produced unstable longitudinal estimates that prevented observations of nonlinear associations and sex differences that should be examined in future studies. Seventh, this study lacks advanced hemodynamic characterizations, including the evaluation of femoral endothelial function, and further studies are needed to examine the role of hemodynamic properties in the longitudinal changes observed in muscle energetics.

Conclusions

Modest longitudinal decline in ABI in people without PAD was associated with a simultaneous decline in thigh muscle oxidative capacity. A dynamic relationship between ABI and muscle energy is suggested by the predominantly concordant changes in ABI and muscle oxidative capacity and the lack of association between kPCr rate of change and baseline ABI. Future studies are necessary to determine whether interventions that increase lower extremity perfusion prevent the decline in lower extremity oxidative capacity and mobility impairment observed in many aging individuals.

Sources of Funding

This study was funded by the Intramural Research Program of the National Institute on Aging.

Disclosures

All authors have no disclosures to report.

Tables S1–S5

Click here for additional data file.