Nutritional status and its effect on treatment outcome among HIV infected clients receiving HAART in Ethiopia: a cohort study

We conducted retrospective cohort study at Jimma University Specialized Hospital, the only teaching and referral hospital with bed capacity of 450 in the South Western part of the country providing specialized health service for approximately 9000 inpatients and 80,000 outpatients each. The ART clinic of the hospital started providing service to people living with HIV/AIDS (PLWA) in 2002. Since establishment the clinic had 3700 patients following care and treatment [11]. The primary data was collected from September 11, 2006 to September 10, 2011. Data was extracted from the medical record from January 30 to February 28, 2014. The sample size was calculated by single proportion formula used for cohort studies, which assumes proportion of mortality in malnourished group to be 61.8 % and proportion of mortality in well-nourished group to be 46.8 with 95 % confidence interval, 80 % power and 1:1 ratio of unexposed versus exposed. The sample size calculated was 340 patients; one hundred seventy (170) patients in both malnourished and well-nourished groups. The medical records of adult patients who started HAART between September 2006 and September 2011 were isolated. The isolated medical charts were categorized into malnourished and well-nourished groups based on their BMI at the start of HAART. Malnutrition was defined as a BMI <18.5, while BMI ≥18.5 was defined as a well-nourished as per WHO criteria. All patients whose age was greater than 14 were included in the study. Pregnant women’s (BMI and nutrient metabolism vary during pregnancy), patients with incomplete data on weight, height and outcome variables, transferred-out during follow-up were excluded from the study. The data were collected by clinically trained data collectors using data collection tool adapted from national ART clinic intake form, ART follow up form and anti-retroviral drugs and patient information sheet. The endpoint of this study was death. The patients were followed until the occurrence of event (death) or until 2 years (end of study). Patients who were alive at the end of study were censored. The survival time was calculated in days using date of starting treatment and date of an event or date censored.

The data was cleaned, edited and completed using Epi data version 3.1 and exported to SPSS for window version 21 for analysis. The main analysis in this study was linear regression for CD4 recovery and survival analysis for death. Variables with p value of <0.25 in bivariate analysis and BMI were fitted to multivariable cox-proportional analysis model to identify independent predictors of CD4 change at all time points. Kaplan–Meier survival analysis was done to estimate the survival time. Log-rank test was used to compare the KM curves for two or more categories of patients on HAART. Predictors of event (death) over a period of time t, was analyzed by Cox proportional hazard model. The level of significance was set at p value less than 0.05 and 95 % confidence intervals (CI) were used throughout. Multi-collinerity was checked for independent variables before fitting into multivariable analysis.

There was no significant difference between malnourished and well-nourished patients in terms of CD4 recovery after HAART initiation. This finding is similar with reports by Elizabeth et al. who reported nutritional status at the start of HAART as measured by BMI, FFMI, FMI and skin folds did not predict good post HAART CD4 change at 6, 12 and 24 months in a cohort of HIV infected Rwanda women [12]. There was no significant difference in CD4 recovery between malnourished patients and well-nourished patients after initiation of HAART [12]. Likewise, Brandon et al. showed that baseline BMI does not predict CD4 change at 48 weeks (11 months) but that this baseline BMI did predict CD4 change at 96 and 144 weeks [13]. Lack of longterm outcomes related to baseline BMI in our study may be related to different ethnic background, geographical region, diet, smaller sample size and shorter follow-up period.

CD4 count at the initiation of treatment was significantly associated with a change in CD4 at 6, and 12 months when adjusted for other variables. A CD4 count greater than 350 cells/mm³ at baseline was associated with smaller increase in CD4 at 6 and 12 months in multivariable model compared to patients with lower CD4 counts. This finding was also documented in London, UK where lower CD4 count was associated with greater increase in CD4 after 3 month of HAART initiation both in multivariable model [14]. In both our study and Bennett and his colleagues, females were found to have better CD4 recovery compared to males after at least a year of treatment [15]. Finally, we found that younger patients had greater increase in CD4 at 6, 12, and 24 months similar to study in Europe which shows younger age favors CD4 cell restoration [16] and increasing age was risk factors for not achieving CD4 count >200 [17]. Similarly, this finding is in line with the finding from study done in seven African countries which shows older age at HAART initiation was associated with suboptimal CD4 recovery CD4 [18]. Older age was independent predictor of mortality, which was comparable with study done in Johannesburg, South Africa which shows older patients were at greater risk of early death compared to younger patients [19]. Moreover, this finding was in line with study from South Eastern Nigeria and with what Adena et al. reported which shows patients whose age greater than 45 were at greater risk of early death [20].

Malnourished patients were almost two times more likely to die early compared to well-nourished group though it is not significant in Multivariable Cox proportional model. This finding was in line with study done in Zewditu Memorial hospital, Ethiopia which shows patients with BMI of <18.5 kg/m² at the start of treatment were 1.13 [95 % CI 0.23, 5.43] times at greater risk of death compared to those with BMI >18.5 kg/m² [21]. This finding was also consistent with the finding from study done in Singapore where patients with BMI less than 18.5 were 1.4 times more likely to die early compared to patients with BMI greater than or equal to 18.5 kg/m² [12]. Studies in Tanzania and Malawi where malnourished patients were associated with early death compared to well-nourished group [22, 23]. Likewise, Raoul and his colleagues reported low BMI was an independent predictor of death in West Africa [18]. Other studies [13‐15] also reported BMI as an important predictor of death.

Tenofovir based regimen was associated with highest risk of death compared to Stavudine and Zidovudine based regimen. This is similar with what Kelechi and his colleagues reported [24] but in contrast to the finding South Africa which shows there was no difference in mortality between Tenofovir and Stavudine [19]. Again different ethnic background, geographical region, diet, smaller sample size and shorter follow-up period in our study. Furthermore, single antiretroviral drugs were compared in our study while Kavindhran et al. [19] compared regimens of three or more antiretroviral drugs.

Independent of specific antiretrovirals, Fluconazole prophylaxis was found to be strong predictor of death in our study and indifference to a Uganda study suggesting that this antifungal did not impact mortality under similar conditions [25]. Most patients in our study took fluconazole when CD4 cell counts were <200/ml.