Weight status and associated factors among HIV-infected people on antiretroviral therapy in rural Dikgale, Limpopo, South Africa

Background Underweight in human immunodeficiency virus (HIV)-infected people on antiretroviral therapy (ART) complicates the management of HIV infection and contributes to mortality, whereas overweight increases the risk of cardiovascular disease (CVD). Aim The study determined weight status and associated factors in people with HIV infection receiving ART. Setting Rural primary health care clinics in Dikgale, Limpopo province, South Africa. Methods A cross-sectional study in which data were collected using the World Health Organization (WHO) stepwise approach to surveillance (STEPS) questionnaire and calculated using WHO analysis programmes guide. Weight and height were measured using standard WHO procedures, and body mass index was calculated as weight (kg)/height (m2). Data on ART duration were extracted from patients’ files. CD4 lymphocyte counts and viral load were determined using standard laboratory techniques. Results Of the 214 participants, 8.9%, 54.7% and 36.4% were underweight, normal weight and overweight, respectively. Physical activity (OR: 0.99, p = 0.001) and male gender (OR: 0.29, p = 0.04) were negatively associated with overweight. Men who used tobacco were more likely to be underweight than non-tobacco users (OR: 10.87, p = 0.02). Neither ART duration nor viral load or CD4 count was independently associated with underweight or overweight in multivariate analysis. Conclusion A high proportion of people on ART were overweight and a smaller proportion underweight. There is a need to simultaneously address the two extreme weight problems in this vulnerable population through educating them on benefits of avoiding tobacco, engaging in physical activity and raising awareness of CVD risk.


Study design
The study was a descriptive cross-sectional study in which data were collected at one point in time.The study used quantitative methods that generated numerical data.Description of participants' measurements and responses expressed as frequencies and categorised assisted in determining the factors associated with being underweight and overweight using regression models.

Setting
The study was conducted in the three primary health care clinics, Seobi-Dikgale, Sebayeng and Dikgale that are situated within the Dikgale Health and Demographic Surveillance System (HDSS) site.The Dikgale HDSS site consists of 15 villages and is situated about 70 kms to the northeast of Polokwane, the capital city of Limpopo province.Limpopo province is one of the poorest provinces and is situated in the northern part of South Africa.

Sample size calculation
The sample size was obtained using the formula for proportion in a cross-sectional study as where n = sample size, Z = a statistic for 95% confidence interval is 1.96, P = expected proportion, d = the maximum permissible difference between sample proportion and population proportion calculated as 20/100 x P. 11 According to literature, the proportion of overweight in ARVtreated Africans ranges from 27.9% to 46.0%. 9,10Considering an average proportion of 36%, and a corresponding maximum margin of error of 7%, sample size was calculated as follows: Sample size = (1.96)0.36(1 -0.36) (0.07)

Study population and sampling
The study population comprised of approximately 74 people with HIV infection treated at Seobi-Dikgale clinic, 373 people at Sebayeng and 377 people at Dikgale clinic.Based on these databases and using convenient sampling method, 20 people were recruited from Seobi-Dikgale, 96 people from Sebayeng and 98 from Dikgale clinics.A total of 214 participants were included in the study, and one was excluded due to pregnancy.
Participants aged 15 years and above who came to collect their ART were asked to participate in the study and were only included after written informed consent was obtained through the completion of a consent form approved by the MEDUNSA Ethics committee (-MREC/HS/137/2011:PG).In the case of minors, written informed consent was obtained from their legal guardians.Pregnant women were excluded.

Data collection
Recruitment and clinical assessment were conducted from September 2013 to March 2014.The WHO stepwise approach to surveillance (STEPS) questionnaire 12 was used to obtain information on socio-demographic characteristics, fruit and vegetable intake, physical activity, tobacco use and alcohol consumption.The questionnaire was first used on few people who did not form part of the study.Physical activity and fruit and vegetable intake levels were calculated using WHO analysis programmes guide. 13Physical inactivity was defined as < 600 metabolic equivalent task-minutes/week (METminutes/week), and low fruit and vegetable intake was defined as < 5 servings/day. 13Data on ART duration were extracted from patients' files.

Blood collection
Fasting venous blood samples were drawn by registered nurses.Whole blood was used to measure CD4 counts on the day of collection.Plasma from whole blood was separated through centrifugation at 2000 rpm for 15 minutes and stored at -80 °C until viral load determination.

Laboratory analysis
CD4 lymphocyte count was measured using the Pima Analyser (Inverness Medical, Tokyo, Japan).Viral load determination using the branched deoxyribonucleic acid (DNA) technique (Siemens, South Africa), was performed by Toga Molecular Biology and Pathology Medical Laboratory in South Africa.

Statistical analysis
Statistical analysis was performed using Statistical Package for Social Sciences version 23. 16 Variables were tested for normality using frequency histograms and line graphs.Data not normally distributed were logarithmically transformed for analysis.ANOVA was used to compare continuous variables and the chi-square test to compare categorical variables among underweight, normal weight and overweight groups.Bivariate logistic regression was used to determine individual influence of socio-demographic factors, HIV and ART variables on underweight and overweight status.Multivariate logistic regression modelling using a forward and backward stepwise approach was used to fit best predictive model and determine independent predictors.Factors significant at p ≤ 0.25 in univariate regression analysis and significant interaction terms at p < 0.05 were included in multivariate modelling.

Ethical considerations
Ethical approval was obtained from the Ethics Committee of University of Limpopo, Medunsa campus (MREC/ HS/137/2011: PG).Permission to conduct the study in the Dikgale HDSS clinics was obtained from the Department of Health-Provincial office and Primary Health Care Capricorn District office.

Socio-demographic and lifestyle characteristics of participants
Of the 214 participants on ART, the majority were women (171 Underweight was three times more prevalent in men (18.6%) than in women (6.4%), whereas overweight was twice as prevalent in women (40.9%) than in men (18.6%).The prevalence of underweight among tobacco users was double the prevalence among non-tobacco users (15.6% vs. 7.1%), whereas the prevalence of overweight was higher among nontobacco users compared with tobacco users (39.6% vs. 24.4%).However, the differences were not significant.A significantly higher proportion of physically inactive (< 600MET-minute/ week) than physically active (< 600MET-minute/week) people were overweight (100% vs. 16%, p < 0.05).None of the people whose daily intake was more than 5 servings/day of fruits and vegetables were underweight (Table 1).

Clinical characteristics of participants
The majority of participants were on efavirenz-based ART (86%) and nevirapine-based ART (12.5%).Only three of the participants (1.5%) were on a lopinavir or ritonavir-based ART regimen.The mean duration of ART in this study was 36 months (range: 1-121 months).Majority of the participants (85%) had an undetectable viral load, and the mean CD4 count was 462 ± 235 cells/mm 3 .
The prevalence of underweight was lower in people with suppressed viral load (< 50 copies/ml) compared with people with a viral load of greater than 5000 copies/ml (7.3% vs. 20.0%,p > 0.05), whereas the prevalence of overweight was significantly higher in people with suppressed viral load (< 50 copies/ml) compared with people with a viral load of greater than 5000 copies/ml (39.9% vs. 0%, p < 0.05).
Overweight people tended to have a higher mean CD4 count than the normal weight or underweight people, but levels of significance were not achieved.The levels of high sensitivity C-reactive protein were similarly high in underweight and in overweight people (7.01 mg/l vs. 5.75 mg/l, p > 0.05) but were significantly higher when compared with the levels in people with normal weight (Table 2).

Predictors of underweight in people on ART
In univariate analysis, gender (OR: 3.33, p = 0.02) and the interaction of gender and tobacco use (OR: 7.72, p = 0.001) were significantly associated with underweight.After controlling for employment, tobacco, fruit and vegetable intake, and viral load in multivariate model, the interaction of gender and tobacco use (OR: 10.87, p = 0.003) was an independent predictor of underweight in the final model (Table 3).

Predictors of overweight in people on ART
In univariate analysis, male gender (OR: 0.33, p = 0.009) and physical activity (OR: 0.99, p = 0.001) significantly associated negatively with overweight.After controlling for tobacco use, CD4 count, viral load and ART duration in multivariate model, male gender (OR: 0.29, p = 0.04) and physical activity (OR: 0.99, p = 0.001) were independent negative predictors of overweight.Neither ART duration nor viral load or CD4 count was independently associated with overweight in multivariate analysis (Table 3).

Discussion
In this study, 8.9% of the participants were underweight and 36.4% were overweight.Similar to our finding, the prevalence of underweight in the general South African population was recently found to be 7.0%. 17However, the 50.8% prevalence of overweight in the general population 17 was higher than the prevalence of overweight observed in our study.The prevalence of underweight and overweight among HIVinfected people on ART reported from rural KwaZulu Natal in South Africa 10 are similar to the prevalence observed in our   study.However, some low-to middle-income countries are still faced with a higher burden of underweight than overweight (Table 4).Although there is a greater tendency to gain weight while taking ART, weight loss has been observed, 3 and therefore, health professionals should monitor patients' weight.An increase in weight among people on ART may be due to a return to good health in which the appetite and caloric intake increases accompanied with a lack of physical activity. 1On the contrary, literature suggests that weight loss during the first 3 months after initiation of ART was associated with symptoms such as difficulty in breathing, nausea, vomiting and oral opportunistic infections. 3Oral opportunistic infections can limit the ability to chew and swallow food leading to loss of appetite and reduced caloric intake.Additionally, poor adherence to ART or drug failure that may result in disease progression may explain a weight loss in people on ART. 3 However, our study did not evaluate these symptoms or drug adherence.In the United States 6 and the United Kingdom, 7 underweight was observed in ≤ 1% and overweight in ≥ 50% of people on ART (Table 4).Although an increasing number of people on ART in developed countries are overweight, in developing countries the co-existence of underweight and overweight among HIVinfected people on ART is a burden that is far from over, despite efforts to eradicate malnutrition.Fears are that the problem of malnutrition (underweight) may be compounded in the future by climate change impacting on agricultural yields and food security. 23Underweight among patients on ART in low-to middle-income countries was shown to contribute to excess early mortality, possibly due to poor immune reconstitution arising from deficient nutritional status. 24Similar to the general population, 4,5 overweight in people on ART may increase the risk of developing CVD.The levels of C-reactive protein an inflammatory marker were similar in the underweight group and in the overweight group, as these extreme weight statuses are characterized by inflammation. 25Adipose tissue in overweight individuals may produce proinflammatory cytokine such as interleukin-6, a significant stimulator of C-reactive protein, 26 whereas opportunistic infections that may be present in underweight individuals are associated with increased levels of C-reactive protein. 27However, individuals with very low CD4 counts less than 50 cells/µl have been observed to produce a small C-reactive response. 27 this study, overweight was disproportionately more frequent in women than in men, whereas underweight was more frequent in men than in women, as was reported among HIV-infected people in KwaZulu Natal, South Africa 10 and in the general population of South Africa. 17In contrast, studies from Brazil 18 and Rwanda 19 reported a higher prevalence of underweight and overweight in women than men.In Ethiopia, the prevalence of underweight was similar among women and men. 20ysical inactivity 28 and low tobacco use 29,30 may contribute to the higher prevalence of overweight in women on ART.In addition, studies among black African women from sub-Saharan Africa suggest high target 'ideal' weight, less dissatisfaction with larger body shape and poor awareness of CVD risk as main contributing factors to overweight among HIV-infected women on ART. 7,17,31Black African men on ART on the contrary are less keen to be obese. 7r study showed that men who used tobacco were more likely to be underweight than men who did not use tobacco.Further analysis to explain this increased risk of underweight among smokers showed that a higher proportion of smokers than non-smokers (11.8% vs. 7.7%) had a viral load greater than 5000 copies/ml and were physically active (100% vs. 95%).Previously, HIV infection was shown to increase the risk of underweight through increased resting energy expenditure, 3 whereas the use of tobacco was shown to reduce appetite. 32People on ART who were not physically active were more likely to be overweight than those physically active as was also observed in Rwanda. 19Physical activity may reduce the risk of being overweight 33 and has a potential anti-inflammatory effect in people on ART. 34In our study, neither ART duration nor viral load or CD4 count was independently associated with underweight or overweight.Similarly, a study from United States showed that demographics and ART duration could not independently predict the increase of weight in people on ART. 6 The relationship between ART duration and BMI is inconsistent. 6,35,36In a longitudinal study, weight gain was observed in a short ART duration (24 weeks) that was followed by weight loss in people treated with stavudinebased ART and maintained weight gain in people treated with tenofovir after 144 weeks. 37Therefore, according to Crum-Cianflone et al. ( 2008), 6 the differences in ART duration may account for variations in this relationship between ART duration and BMI.
According to the WHO strategy to tackle nutritional problems, stakeholders at global, regional and local levels have a mandate to improve diets and physical activity patterns of the population. 38In line with this strategy, South Africa introduced the 'Move for your Health' programme aimed at promoting physical activity.An initiative of major private health insurers the 'Wellness programme' rewards and recognizes those maintaining a healthy weight. 39In addition, programmes such as social grants, the provision of nutritional supplements to children and those on ART are aimed at alleviating undernutrition.However, the impact and effectiveness of these initiatives remains unclear.
Our study is one of a few in South Africa to provide information on the problem of underweight among people on ART in a rural area.The limitations of this study include its cross-sectional design that limits inference.In addition, fruit and vegetable intake, physical activity, tobacco and alcohol use were self-reported, and recall bias may have influenced results.Although our study did not assess the full dietary intake, an assessment of fruit and vegetable intake showed that people in all weight status categories had a lower fruit and vegetable daily intake than recommended.In agreement to this finding, a study conducted in a South African rural and semi-rural area, similar to Dikgale HDSS, found that the intake of fruit and vegetables in that community was low. 40Overcoming this problem particularly in rural areas may therefore remain a challenge, as evidence suggests an association of perceived expense, and low income with low fruit and vegetable intake. 40,41A low level of fruit and vegetable intake may lead to deficiencies in vitamins, predisposing people to a wide range of conditions including CVD, cancers and cataracts 41,42 and may accelerate HIV disease progression. 43Most of our participants were older people as the majority of young adults live away from home as temporary migrants for employment reasons. 44onvenient sampling was used to recruit participants.However, recruitment was conducted for a whole month cycle, giving all patients collecting their medication equal opportunity to participate.We also acknowledge the small sample size of our study.Further longitudinal cohort intervention studies that would monitor weight changes in people on ART are needed.

Conclusion
In a rural community in the Limpopo province of South Africa, a high proportion of people on ART were overweight and smaller proportions were underweight.Men who used tobacco had a higher risk of being underweight than nontobacco users.Female gender and physical inactivity increased the risk of being overweight.There is need to simultaneously address the two extreme weight problems in people on ART through educating this vulnerable population on the benefits of staying away from tobacco use, engaging in physical activity and increasing awareness on CVD risk.
was considered and the expected sample size was 200 people.

TABLE 1 :
Socio-demographic and lifestyle variables among HIV-infected people on ART by weight status.
Data is presented as n (%) number (percentage) unless indicated.*, Analysis of variance test (ANOVA) p-value and chi-square test were used to obtain linear association p-value; s.d., standard deviation; ART, antiretroviral therapy; MET-minute/week, metabolic equivalent of task-minutes/week.BMI was considered as underweight (< 18.5 kg/m 2 ), normal (18.5-24.9kg/m 2 ) and overweight (≥ 25.0 kg/m 2 ).Data for Pima Analyser (PA) and fruit and vegetable are available for 157 participants.

TABLE 2 :
Clinical variables and conditions among HIV-infected people on ART by weight status.

TABLE 3 :
Multivariate logistic regression analysis: predictors of underweight and overweight in HIV-infected people on ART.Variables that were included in the multivariate forward and backward stepwise regression model.Variables in final multivariate model are presented with adjusted odds ratio.BMI was considered as underweight (< 18.5 kg/m 2 ), normal (18.5-24.9kg/m 2 ) and overweight (≥ 25.0 kg/m 2 ).

TABLE 4 :
Prevalence of underweight and overweight in patients on ART in previous studies.
Table 4 continues on the next page →

TABLE 4 (
Continues...): Prevalence of underweight and overweight in patients on ART in previous studies.