Analysis of Viral load change in Case of HIV/AIDS Patients under ART Follow-up in Arba Minch General Hospital

Background : Monitoring human immunodeficiency virus plasma viral load over time is critical to identify virologic treatment failure in patients taking antiretroviral therapy. The aim of this study was to determine whether the overtime viral load changes depends on patient characteristic measured at baseline of human immunodeficiency virus patients at Arba Minch General Hospital. Methods : This prospective follow up study was conducted using data obtained from medical records, patient interviews, and laboratory workup for six months. The study was employed among 152 adult patients that were selected by systematic random sampling. Longitudinal data analysis that accounts for the correlated nature of the data handled through linear mixed effect models were used to fit the data set in this study. Result : The mean viral load declines over time for each of the adherence level groups. The estimates of linear (p = 0.0006) and quadratic visit time (p=0.0256) effects and the baseline characteristics sex, age, adherence level, and Isoniazid preventive therapy had significant effects on change of viral load of patients over time. Conclusion : In order to improve the status of the patient’s viral load over time, considering the patients’ differences in adherence to antiretroviral therapy, sex, age, and Isoniazid preventive therapy are important.


Introduction
Human immune deficiency virus treatment access is a key to the global effort to end AIDS as a public health threat. Currently, 12 million individuals are receiving antiretroviral therapy (ART), and a rapid scale up in the number of individuals receiving ART is in progress (8,9). In an effort to scale up and decentralize ART services in Ethiopia, there are increasing numbers of health facilities providing ART and similarly the number of patients getting ART services is increasing (10).
The primary goals of initiating ART among HIV patients are to suppress HIV viral replication and to restore immune function. The clinical decision to check whether such goals have been achieved is made through periodic viral load testing and/or CD4 cell counting (11). Failure to achieve and maintain suppression can result in the development of drug resistance and also increases the risk of both horizontal and vertical viral transmission (12 -14) In addition, a low viral load indicates that treatment is effective. A high viral load in a person on treatment indicates either that the medication is not being taken properly or that the virus is becoming resistant to the medication [3]. HIV patients with high viral loads are more infectious and lead to higher HIV transmission rates [4,5].
Furthermore, plasma viral load monitoring is an important as a marker of response to ART, because a decline in viral load suggests that the patient is adherent to the regimen, that the appropriate doses are being administered, and that the virus is susceptible to the drugs in the regimen [7]. Viral load testing could increase in importance as a guide for clinical decisions on when to switch to second-line treatment and on how to optimize the duration of the first-line treatment regimen [8]. Recently published guidelines regarding the use of antiretroviral therapy now include the recommendation that both viral load and CD4 lymphocyte counts should be monitored regularly, and that plasma viral load should be reduced by as much and for as long as possible [8].

Study area and design
The study was conducted in Arba Minch General Hospital, Arba Minch town, Gamo zone, In this study, systematic sampling technique with k th interval (i.e. k = N/n), k=263/152=1.73≈2 (i.e. every other patients were selected). The first study participant was selected randomly from the first patients who visited the clinic.

Study variables
The response variable considered for this study was the viral load.

Data collection procedures
After written consent obtained, using data collection tools, the data were collected on sociodemographic, baseline clinical and treatment related characteristics of the participants by their clinical attendant (trained data collector) starting from the commencement of the care.
The patient's information was obtained from the participants, their data base and medical charts. Additional information was collected during the 6 months follow up of the study participants.
Specimen for the laboratory tests of CD4 and viral load tests was collected by the trained laboratory staff at the facility. For both tests 4-5 ml of whole blood was drawn from each participant using vacutainer tube separately with anticoagulant following standard venipuncture protocols for viral load testing. Plasma sample was assayed for the presence of HIV RNA using Amplicor Monitor standard assay, version 1.5 (Roche Molecular Systems).

Data processing and analysis
The statistical software used in this study was the Statistical Analysis Software (SAS)

Linear Mixed-Effects Model
The continuous outcome variable viral load count contains measured at months 0, 1, 2, 3, 4, 5, and 6. Since measurements were taken from the same subject over time, observations cannot be considered as independent. So, appropriate random effect models that account for the correlated nature of the data was used. The random-effects approach is extending the univariate linear regression model to longitudinal settings based on subject-specific regression model [11].
A longitudinal model is the estimation of changes in response over time and testing whether these changes are covariate dependent [12]. Special methods of statistical analysis are needed for longitudinal data because the set of measurements on one patient tend to be correlated, measurements on the same patient close in time tend to be more highly correlated than measurements far apart in time, and the variability of longitudinal data often changes with time. These potential patterns of correlation and variation may combine to produce a complicated covariance structure. This covariance structure must be taken into account to draw valid statistical inferences. To assess the need for serial correlation inclusion, the criterion of fitting linear mixed models with the same mean and random-effects structure was used as proposed by Verbeke and Molenberghs [11].

Model Selection Criteria
A key part of the analysis of data is model selection, which often aims to choose a parsimonious model. To have an appropriate model for the linear mixed model most commonly known model selection criterions; Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) [14] were considered for this study.

Ethical considerations
The letter of ethical clearance was obtained from the institutional review board (IRB) of College of Medicine and Health Sciences in Arba Minch University. Written consent was obtained from all study participants for blood draws and interviews. Participants with a high likelihood of virologic failure prioritized for tailored individualized treatment preparation and other interventions to improve treatment outcomes in the hospital.

Exploratory data analysis
The baseline socio-demographic, clinical and chemo prophylaxis characteristics of adult enrolled on first line ART were presented in Table 1   The summary statistics at baseline variables and viral load over time were presented in  The next step in longitudinal studies address the relationship of a response with explanatory variables, often including time. The individual profiles, the average evolution, the variance function, and the correlation structure are very helpful tool in the selection of appropriate models. In the following sections exploratory analysis for the data sets considered.

The Correlation Structure
The correlation structure describes how measurements within a subject correlate. The correlation matrix presented in Table 3 below shows the pairwise correlation between measurements at any pair of time points. It was observed that there were positive correlations between the pair of measurements. Also, scatter plot matrices are a way to roughly determine if we have a linear correlation between multiple variables. The scatter plot matrix displayed in Figure   4 below, shows that viral load count at months (0, 1, 2, 3, 4, 5, and 6) are strongly correlated. So, we could consider the correlation of data in the model.

Result of Linear Mixed Model
The outcome variable of interest is the viral load by which evolution over time may be assessed. Based on results of exploratory data analysis, the evolution of the viral load was assumed to have a linear and quadratic function over time effect as a preliminary mean structure and unstructured variance-covariance structures seemed a plausible starting for performing linear mixed model. The models were compared using the likelihood ratio test based on REML and results showed that there was no need of including the serial correlation.
The possibility of random-effects structure reduction was also assessed using the likelihood ratio test (mixture of chi-squares with equal weights 0.5) by deleting random effects in a hierarchical way starting from the highest order time effect comparing it with the previous model. After this, simultaneous contrast statements were used to explore the possibility of reducing the mean structure, starting from the unstructured mean effect which was rejected.
In all cases, the possibility of simplification of random effects from the models was the only model with random intercept.

Discussion
This study was designed to fit a model of how adherence level influence viral load over visit time and whether the evolution of viral load depends on patient characteristic measured at baseline of HIV/AIDS patients at Arba Minch General Hospital.
This study showed that variable sex had significant effect on viral load. The viral load of the female group was greater than male patients. A related study by Adal et al. [15] also showed an association between sex and HIV RNA load. In another study by Hewitt et al. [16] it was found that women appear to have an increased risk of progression to AIDS compared with men with the same viral load. However, study aimed to assesses the use of both viral load (HIV RNA) and CD4 cell count in the monitoring of HIV/AIDS progression, it was found that there was no sex effect on the progression of HIV based on viral load levels [6]. Moreover, the present study identified that the baseline age and INH therapy were significantly associated with the change in viral load over time.
In previous a study, it was investigated that the level of adherence to HAART is closely associated with suppression of the HIV viral load in plasma [17]. In the current study also found that adherence level had significant effect on viral load. Baseline the viral load of fair and poor adherence level group were greater viral load measurements in comparison with good adherence level used group.
In conclusion, the viral load of HIV/AIDS patients is significantly determined by the visit

Authors' contributions
KT, SH and MA conceived and designed the study, developed data collection instruments and supervised data collection. They participated in the testing and finalization of the data collection instruments and coordinated the study progress. KT, SH and MA performed the statistical analysis and wrote all versions of the manuscript. All authors read and approved the final manuscript.