Predictors of adverse drug reaction among adult HIV-infected patients on antiretroviral therapy in government hospitals of Kaffa Zone, Ethiopia; November 2018: a retrospective cohort

Introduction incidence of adverse drug reactions (ADR) associated with antiretroviral therapy (ART) was higher in developing countries. In two teaching hospital in Ethiopia: Debremarkose 23% and Yirgalem 73.2% of study participants reported at least one ADR. Since there was limited information about ADR in the study area; we aimed to determine its incidence-rate and predictors. Methods we conducted retrospective cohort study using medical records of HIV-infected patients enrolled on ART between 2006 and 2017 in government hospitals of Ethiopia. ADR was defined as report of at least one unwanted response to ART. We run descriptive and cox regression analysis (CRA). Results incidence-rate of ADR was 4.1 per 100 person-years (py). Hazards of ADR among patients living at rural was almost two times than at urban; [Adjusted hazard ratio (AHR): 1.94(95% (CI): 1.18, 3.20)]. Stavudine (D4T)-Lamivudine (3TC)-Nevirapine (NVP) had about two times [AHR: 1.78(95%CI: 1.03, 3.08)], Zidovudine(AZT)-3TC-NVP had about two times [AHR: 2.34 (95%CI: 1.20, 4.57)], D4T-3TC-Efaviranze(EFV) had about three times [AHR: 2.86(95%CI: 1.38, 5.95)] and AZT-3TC-EFV had about two times [AHR: 2.16(95%CI: 1.21,3.90)] hazards of ADR than Tenofovir(TDF) based regimens. Being WHO clinical stage III had about two times hazard of ADR [AHR: 2.46 (95%CI: 1.22, 4.95)] and IV had about four times hazard of ADR [AHR: 4.32 (95%CI: 1.88, 9.93)] than stage I. Conclusion risk of ADR was higher among adult HIV-infected patients on ART living in rural, WHO clinical stage III and IV, and patients on AZT and D4T based regimen. AZT should not be given as an alternative treatment, increase access of TDF regimens.


Introduction
Since the start of acquired immune deficiency syndrome epidemic, more than 70 million people infected and 35 million died worldwide [1]. Sub-Saharan Africa remains the most severely affected region, with nearly 1 in every 25 adult (4.2%) living with HIV [2]. Treatment for HIV involves multiple drugs, each with own side effects [3]. ADR is defined as; a noxious, unwanted, unpredictable and unintended response to drug taken correctly [4].The spectrum of ADRs associated with ART varies between developed and developing countries, and typically ranges from 11-35.9% reported in developing countries [5]. In two studies in Ethiopia, 23% and 73.2% the study participants in two teaching hospitals (Debremarkose and Yirgalem respectively) reported at least one ADR [6,7]. Currently, TDF and AZT are given with other regimens for initiation of care for HIVinfected patients in Ethiopia; AZT alone is considered as an alternative. Teaching hospitals (tertiary hospitals) provide high-level specialized care, and might have different population compared with primary public governmental hospitals. To determine incidence and predictors of ADRs in Ethiopia, we evaluated data from government hospitals in Kaffa Zone, Ethiopia during July 2006 and August 2017.

Methods
Study setting: this study was conducted in Gebretsadik Shawo Memorial General Hospital (GSMGH) and Wacha Melese Zenawi Memorial Primary Hospital (WMZMPH) government hospitals of Kaffa Zone between June and August 2018. GSMGH and WMZMPH are the only general and primary hospitals in the zone serves as a referral site. ART service was started in both hospitals in 2006.

Study populations and design:
we conducted a retrospective cohort study. We reviewed all baseline and follow-up clinical records of HIVinfected adult patients aged > 15 years who started ART between July 2006 and August 2017. All patients on ART with complete baseline and at least single follow-up after enrolment were included; duplicated and illegible data were excluded.
Sample size determination and sampling technique: we calculated sample size for incidence and predictors of ADR using EPI info V.7. We used a single population formula to calculate the incidence of ADR, using a confidence level of 95%, a margin of error 5% and incidence of ADR 23% based on the study in Debremarkose hospital. Using this estimations, the sample size was 203 [6]. We compared this with a sample size for power with 80% power, 95% confidence, a 5% margin of error, and a relative risk of ADR based on educational level of 1.5, for a sample 530 [6]. During our study period, 797 patients registered on ART. Of those, 592 clinical records had complete baseline data with at least single follow-up, after enrollment for ART had been included. Dependent variables included incidence of ADR and ADR-free survival time. Independent variables included, variables measured during initiation of HIV-infected adult patients for ART; socio-demography, baseline CD4 count, body mass index, clinical stages, functionality status of the patient, comorbidity, ART treatment type and eligibility criteria for ART.
Data collection method and tool: data were collected using standard data abstraction sheet; informed by reviewing previous literature as well as available ART intake and follow-up forms. Data were collected by ART trained clinical nurses and public health officers. A three-day orientation was given for supervisor and data collector.
Data processing and analysis:data were checked, coded, cleaned and entered in to Epi-info version 7.2.2, then exported and analyzed by SPSS version 20. Variables with P-value (P < 0.25) on the bivariable CRA were selected as candidates for multivariable CRA. On multivariable CRA, variable with P<0.05 declared as statistically significant predictor of incidence of ADR. We used a Kaplan-Meier curve to assess the cumulative survival. We summarized the data using mean, median and AHR, and presented with graph,  Figure 1). In total, 112 ADRs were reported. Among those, 23 (20%) were anemia, and 16 (14%) were peripheral neuropathy ( Figure 2).  (Table 3, Table 3 (suite)).

Discussion
In our study of adverse drug reactions among HIVinfected persons, we found an overall incidence of 4.1 per 100 person-years of treatment, or approximately one in every twenty-five persons on treatment, per year. We found that rural persons were likelier to experience an ADR compared with urban persons, and that persons in later stages of AIDS were likelier to experience ADRs, compared with persons at earlier stages. In addition, hazards were lower for persons on TDF-based regimens than AZT and D4T based regimens. Risk was highest in the first five years on treatment, and gradually declined. Incidence of ADR in the current study was 4.1 per 100 py (95% CI: 3.4, 5.0), which was similar with findings from a study in Felege Hiwote referral hospital, Bahirdar, where the rate was 4.2 per 100 py (95% CI: 3.3, 5.4). However, this finding was inconsistent with findings from a study in the seven teaching hospital in Ethiopia; 9 per 100 py. The difference could be due to difference in study design, design of the previous study was a prospective cohort and data were collected objectively for research purpose.
Based on our finding, the risk of ADR among patients from rural areas were higher than patients from urban. Similarly, the proportion of ADR were higher at the rural on the studies conducted at Mali [8]. Even though studies were conducted in different setting in Ethiopia, residency was not significantly associated with ADR. The difference could be, patients in the rural may use traditional medicine and non-prescribed medicine.
In this study, the risk of ADR among D4T-3TC-NVP, AZT-3TC-NVP, D4T-3TC-EFV and AZT-3TC-EFV were higher than TDF base (TDF-3TC-NVP and TDF-3TC-EFV) combinations. This finding was similar with studies done in Duala Hospital; Cameroon, tertiary teaching hospital; Kenya and Debremarkose hospital; Ethiopia [6,9,10]. Additionally, similar result was observed on D4T based regimen in Hiwote Fana hospital Harer, Ethiopia [11]. In contrast incidence of ADR was higher in TDF/3TC/EFV in Jimma Hospital Ethiopia [12]. This could be due to small sample size in previous study, which was 233 and much smaller than our study. It is known that D4T base regimens were known to cause toxicity and removed from the treatment plan in different country including Ethiopia. Additionally, AZT based regimens were also known to cause ADR, but still it is given as an alternative in Ethiopia.
The risk of ADR among patients with WHO clinical stage III and IV were higher than WHO clinical stage I, this is similar with the studies conducted at Bahirdar, Ethiopia [13]. Similarly, in other study in Ethiopia; incidence of ADR was higher among patients on WHO clinical stage III. This could be due to drug interaction; patients on WHO clinical stage III and IV might be on other treatment with ART.

Conclusion
The risk of ADR was much higher among adult HIV positive rural patients on ART. Additionally, adult HIV-infected patients within WHO clinical stage III and IV on HAART had higher risk of developing ADR than stage I. The study also concluded that patients on AZT and D4T based regimen had considerably higher risk of developing ADR than TDF based regimen. To reduce ADR, AZT-based regimens should not be given as a TDF alternative and increase TDF based regimens. ART providers should be prepared to see higher ADRs in rural and WHO clinical stage III/IV patients starting on ART, and give more attention. Additionally, we recommend further study on why incidence of ADRs were higher among rural patients than urban.
What is known about this topic  D4T as a regimen was declined from the treatment plan because of its higher incidence of ADR.

What this study adds  This study showed that AZT as a treatment is soul cause of ADR;  The hazard of ADR among patients lived in the rural is much higher than patients lived in the urban;  The hazard of ADR was higher among patients on clinical stage-3 and 4 than patients on clinical stage-1.
AAM the principal investigator, contributed in designing the study, data collection, data analysis, interpreting the result, writing and revising. FBB and ATD contributed in supervising the whole research, assisted in design and revising. All authors have read, reviewed and approved the final version of the manuscript.

Acknowledgments
We acknowledged Hawassa University, Kaffa Zone Health department and the two hospitals for facilitation during data collection. We also extend our acknowledgment to data collectors and supervisors, without them, this project has not been real. Additionally, we extended our heartfelt thanks to Julie Harris (PhD) for her support during development of this manuscript.