Objectively measured physical activity among people with and without HIV in Uganda: associations with cardiovascular risk and coronary artery disease

Participant Selection

The Ugandan sTudy of HIV effects on the Myocardium and Atherosclerosis using Computed Tomography (mUTIMA-CT), conducted in Kampala, Uganda, is an ongoing prospective cohort investigation examining the impact of HIV on the myocardium and atherosclerosis. In this phase of the study, PLHIV were age-(within 3 years) and sex-matched in a 1:1 ratio with PWOH controls. PLHIV were recruited from the Joint Clinical Research Centre (JCRC) in Lubowa, Uganda, which is near Kampala. PWOH were recruited from community or hospital-based internal medicine clinics in Kampala. To be eligible for inclusion, PLHIV must have been on ART for more than six months without any changes in regimen within 12 weeks prior to enrollment. All participants, regardless of HIV status, were ≥ 45 years old and had at least one cardiovascular disease (CVD) risk factor, such as hypertension, low high-density lipoprotein cholesterol, diabetes mellitus, smoking, or a family history of early CAD. Exclusion criteria were a history of known CVD, peripheral artery disease, ischemic stroke, uncontrolled chronic inflammatory conditions, pregnancy, use of chemotherapy or immunomodulating agents, or an estimated glomerular filtration rate (eGFR) less than 30 ml/minute.

The original cohort comprised of 100 PLHIV and 100 PWOH enrolled from April 2015 to May 2017. Previous publications have detailed the findings and methods of the baseline examination of the original cohort^10–13. Follow-up examinations were conducted from 2017 to 2019, with total cohort size of 200 participants maintained by replacing those lost to follow-up with age- and sex-matched individuals. This follow-up consisted of 2 study visits separated by 7 days, during which physical activity metrics were collected via accelerometry. This current study is a cross-sectional analysis of those data among participants with accelerometry data duration of at least 10 hours per day on at least 4 days^14,15, which is described as valid wear time.

The study protocol was approved by the University Hospitals Cleveland Medical Center Institutional Review Board, the JCRC Research Ethics Committee, and the Uganda National Council for Science and Technology, with all participants providing written informed consent prior to any study procedures.

Physical Activity

Self-reported measures of physical activity were assessed using the International Physical Activity Questionnaire – Short Form (IPAQ-SF). This was developed as a tool for international monitoring of physical activity and sedentary time¹⁶. The IPAQ-SF contains a series of 7 questions, requiring participant recall of light, moderate, and vigorous physical activity time, as well as sedentary time over the preceding 7 days. The amount of light, moderate, and vigorous physical activity per day was multiplied by the number of days performing those activities to get the amount of activity per week. The amount of time spent sitting per day was multiplied by 7 to obtain the amount of sedentary time per week.

Objective measures of physical activity were assessed by accelerometry. Participants were instructed to wear an Actigraph®wGT3X+-BT monitor (Actigraph, LLC, Pensacola, FL) for 7 days on their non-dominant hip. Physical activity metrics of interest were steps per day, light physical activity time per week, moderate-to-vigorous physical activity time per week, and sedentary time per week. The number of steps per day was derived by dividing the number of total steps recorded by the number of valid wear days. Other physical activity metrics were derived by dividing the amount of activity time by the number of valid days and multiplying by 7. Data was sampled at 30 Hz, using 60-second epochs and the normal filter. Bouted physical activity was used to measure physical activity accumulation over 10 minutes or more. Different intensities of physical activity were defined by adult cut-points for tri-axial accelerometers¹⁷. In this validation, moderate-to-vigorous activity was defined as achieving at least 3 metabolic equivalents and was predicted by achieving at least 2,960 triaxial counts per minute. Valid wear time was determined by data recorded for at least 10 hours per day for at least 4 days. Accelerometry data was analyzed using Actilife software.

Atherosclerotic Cardiovascular Disease Risk

10-year ASCVD risk was quantified using the pooled-cohort equation¹⁸. The following variables were utilized: age, sex, and smoking status were obtained by self-report; diabetes was defined as a chart confirmed diagnosis or on medication for diabetes; hypertension treatment was defined as receiving medication to lower blood pressure. Systolic and diastolic blood pressure were obtained at the initial visit by trained study staff. Total cholesterol, high density lipoprotein, and calculated low-density lipoprotein concentrations were measured by the JCRC clinical lab from fasting blood drawn at the initial visit. Notably, we chose to use the “other race” term, since the African American term has not been validated in the East African population. Further, the nature of atherosclerotic cardiovascular disease risk may be different in the East African population compared to the population from which the pooled-cohort equation was derived¹⁰. Elevated ASCVD risk was defined as 10-year ASCVD risk ≥5%, which is the threshold to consider use of statin therapy to reduce ASCVD risk¹⁸.

Coronary Artery Disease

At visit 2, all participants underwent coronary computed tomography angiography (CCTA) on a 128-slice Siemens Somatom scanner at Nsambya St. Francis Hospital in Kampala. Participants were ineligible to have intravenous contrast for CCTA if their estimated GFR was not greater than 60 mL/min/1.73m². Data acquisition was additionally limited for some participants by other technical reasons (e.g., inability to adequately lower the heart rate with betablockers) (Figure 1). The acquisition and image analysis protocols were developed in alignment with the Society of Cardiovascular Computed Tomography Guidelines¹⁹. Participants were given 100 mg of oral metoprolol two hours prior to the scan. They were additionally given another 50 mg dose 30 minutes prior to the scan if the heart rate was elevated above 60 beats per minute.

• Download figure
• Open in new tab

Figure 1.

Derivation of the study population from the mUTIMA-CT cohort. Definitions: PLHIV, people living with HIV; CTA, computed tomography angiography; eGFR, estimated glomerular filtration rate.

The CT scans were first read by a local radiologist for clinically significant findings and were subsequently read offline in batch by a single expert reader (MSB) for research. Scans that were of poor technical quality were excluded from analysis. Segment involvement score (SIS) was defined as the total number of disease segments.

Segment severity score (SSS) was calculated using a luminal obstruction weight for that segment (x1 if <25% obstruction, x2 if 25–50%, x3 if 50–70%, x4 if 70–99% and x5 if totally occluded), giving a maximum possible SSS of 90 for an 18-segment model. The presence of CAD was defined as SIS>0 and more severe CAD was defined as an SSS greater than the median among those with CAD.

Statistical Analysis

Demographic and clinical characteristics were described by frequency (percent) for categorical variables and median (interquartile range) for continuous variables, stratified by HIV status and sex. Physical activity measures were described in a similar manner. Spearman’s correlation coefficient was used to evaluate the relationship between self-reported and gold-standard objective measures of physical activity. Bland-Altman plots were used to assess the degree of agreement between the two measures of physical activity. The objective measures were subtracted from the self-reported measures to obtain measurement differences and the limits of agreement were defined as the mean difference ± 1.96 times the standard deviation of the differences. Spearman’s correlation was additionally used to assess the relationship between age and objective measures of physical activity. Additionally, the Wilcoxon rank-sum test was used to compare physical activity metrics in their continuous version across binary subgroups of HIV status and sex. Chi-square tests were used to compare the proportion of those achieving at least 150 minutes of moderate-to-vigorous exercise per week and those achieving at least ten thousand steps per day by HIV status and sex.

Unadjusted and adjusted Poisson regression models with robust standard errors were used to estimate prevalence ratios to analyze the relationship between binary measures of physical activity, by median cut points, with elevated ASCVD risk, presence of CAD among those with available CTA data, and presence of more severe CAD among those with CAD. Binary measures of physical activity were used to meet the linearity assumption of log transformation of the outcomes of interest with their physical activity predictors. Median cut points were used to allow for sufficient sample size on both sides of the selected cut points. Covariates – age and sex – were chosen a priori to build the adjustment models. HIV status and sex interaction terms, along with likelihood ratio tests, were used individually to assess for effect modification in the relationship between binary objective physical activity metrics and the presence of CAD. A schematic of the data analysis approach may be found in Supplementary Figure 1.

R version 4.3.1 was used for analysis: P<0.05 was considered statistically significant.

Study cohort

The cohort began with 200 participants (Figure 1). The overall cohort for this analysis included 168 participants with valid accelerometry wear time, 48% of whom were PLHIV. 139 participants among those with valid accelerometry wear time had coronary angiography available for analysis (46% PLHIV). The most common reasons for missing coronary angiography were tachycardia and low eGFR.

Baseline characteristics of study participants

The median (IQR) age of the overall cohort was 57 (53, 62) years old and 64% were female (Table 1). The median age was similar across HIV status and sex. 84% of PLHIV had viral load suppression, and males were more likely to have HIV viral load suppression than females (94% vs 77%). Overall, smoking rates were low (2.4%) and no females reported smoking. The median systolic blood pressure was higher in males than in females [155 mmHg (139-171) vs 146 (131-162)]. Males had higher ASCVD 10-year risk than females [13% (8-24) vs 6% (3-10)]. Males had lower body mass index than females [25 (23-28) vs 31 (27-34)].

Relationship between self-reported and objective measures of physical activity

There was no statistically significant correlation between self-reported measures of physical activity and objective gold-standard measures of physical activity (ρ=0.06, P=0.529 for sedentary time; ρ=-0.04, P=0.703 for light activity time; ρ=-0.16, P=0.172 for moderate-to-vigorous activity time; Supplementary Figure 2). Additionally, participants overestimated their moderate-to-vigorous activity and underestimated their light activity; and these errors in estimation were enhanced as self-reported or objectively measured physical activity increased. Objective measures of physical activity were used for the remainder of the analysis.

Objective physical activity outcomes of study participants

The median (IQR) accelerometry valid wear days for the total cohort was 6 (6-7) days during the 7 days in which objective measures of physical activity were collected. The median step counts per day was higher for males than for females [6,534 steps (5,073-9,178) vs 5,382 steps (3,808-7,316), P= 0.005], and males were more likely to achieve at least 10,000 steps per day than females (16% vs 5%, P=0.023). Males had more moderate-to-vigorous physical activity per week [68 minutes (12-144) vs 15 minutes (0-50), P<0.001] than females and a greater proportion of men achieved at least 150 minutes of moderate-to-vigorous physical activity per week than females (21% vs 5%, P=0.002). Males had less light physical activity per week [788 minutes (497-1,202) vs [1,059 (730-1490), P=0.001] than females. There was no significant difference in objective physical activity outcomes when comparing PLHIV and POWH.

There was a statistically significant positive correlation between age and sedentary time (Spearman’s correlation coefficient, ρ=0.22, P=0.004) and a statistically significant negative correlation between age and moderate-to-vigorous physical activity (ρ=-0.25, P=0.001), and steps per day (ρ=-0.31, P<0.001) (Supplementary Figure 3). The negative correlation between age and light physical activity was not statistically significant (ρ=-0.12, P=0.113).

Association of objective physical activity outcomes and atherosclerotic cardiovascular disease risk and coronary artery disease

Prior to adjusting for age and sex, participants who achieved at least 15.7 hours of light physical activity per week (the median in the total cohort) had non-statistically significant 19% lower prevalence of elevated ASCVD risk compared to those who had less than 15.7 hours of light physical activity per week (unadjusted prevalence ratio (PR) 0.807, 95% CI 0.649-1.002, P=0.053). This association was weakened after adjustment for age and sex. Prior to adjustment, there was a general non-statistically significant association of decreased risk of elevated ASCVD risk with more physical activity and increased risk of elevated ASCVD risk with more sedentary time for the overall cohort.

For the total cohort, no statistically significant association was found between physical activity metrics and presence of CCTA-detected CAD prior to and after adjustment. Prior to adjustment, there was a general non-statistically significant trend of lower risk of CTA-detected CAD with more physical activity and greater risk of CAD with greater sedentary time for the overall cohort.

After stratifying by HIV status and prior to adjustment, there was a trend of lower risk of CT-detected CAD in PLHIV with more light physical activity compared to PWOH [unadjusted PR 0.259, 95% CI 0.057-1.188, P=0.082 vs unadjusted PR 1.461, 95% CI 0.675-3.160, P=0.336] with a statistically significant interaction P-value of 0.025. After adjustment, the interaction was no longer statistically significant.

Similarly, there was no significant interaction of sex on the relationship between physical activity metrics and the presence of CTA-detected CAD prior to and after adjustment.

Prior to and after adjustment, there was a general non-statistically significant trend of lower risk of more severe CTA-detected CAD with more physical activity and higher risk of more severe CTA-detected CAD with more sedentary time.

Strengths and Limitations

This work contains several important strengths including the use of gold-standard objective measures of physical activity and atherosclerotic CAD. We have shown that self-reported physical activity is not a good measure of objectively measured physical activity, arguing for the use of objective measures whenever feasible. Our prior research has also shown that ASCVD risk scores do not correlate well with objectively measured subclinical atherosclerosis in Uganda, and thus risk scores should not be used as a surrogate marker of subclinical ASCVD¹⁰.

However, our study also has limitations. First, it is possible that our sample size was not sufficient to detect significant associations between physical activity metrics and the outcomes described. Secondly, this was a cross-sectional analysis where measures of physical activity were assessed for a short period of 7 days. Although wearing the physical activity monitor during 7 days with a minimum of 4 days of valid data is recommended³⁰, this did not capture the majority of lifetime physical activity habits that would inform the relationship with ASCVD risk, CAD, and severity of CAD. Thirdly, it is possible some participants may have altered their physical activity habits with the awareness of monitoring. Additionally, as this was an observational study, the univariable trends and associations seen cannot be construed as causal. Lastly, this was a single-center study based near the commercial capital of Uganda and the results might not be generalizable to those living in more rural areas.