Immunological recovery, failure and factors associated with CD-4 T-cells progression over time, among adolescents and adults living with HIV on Antiretroviral Therapy in Northern Ethiopia: A retrospective cross sectional study

Background This study was aimed to assess immunological recovery, failure, and factors associated with CD-4 T-cells progression over time, among adolescents and adults living with HIV on Antiretroviral Therapy in Northern Ethiopia. Methods A retrospective cross sectional study was done on 19,525 HIV patients on ART. Data were collected using a data retrieval checklist from a database. All eligible data in the database were exported to Microsoft excel 2010 and then data verification and filtration were done before exporting to STATA 14.0 for analysis. Factors associated with recent CD-4 count were modeled by using Generalized Linear Model poison family. Results Among the patients with advanced HIV infection (< 200 CD-4 T-cell/ mm3) at baseline, only 28.35%, 95% CI (27.45–29.26) of them had immunological recovery (≥ 500 T-cells/mm3). Only 2.14%, 95%CI (1.94%- 2.35%) of the patients had immunological failure. Baseline CD-4 count (Incidence Rate Ratio (IRR) = 1.0007, 95%CI = 1.00069–1.00078), patients from military health care facility (IRR = 1.11, 95%CI = 1.06–1.16), good adherence (IRR = 1.12, 95%CI = 1.04–1.21) and viral load suppression (IRR = 1.31, 95%CI = 1.28–1.33) were positively associated with recent CD-4 count in the full model. Whereas, being male (IRR = 0.85, 95%CI = 0.83–0.86), patients with on Anti-Retroviral Therapy (ART) regimen of 1e (TDF-3TC-EFV), 2f (AZT-3TC-ATV/r), and 2h (TDF-3TC-ATV/r) (IRR = 0.92, 95%CI = 0.91–0.94), (IRR = 0.65, 95%CI = 0.55–0.76) and (IRR = 0.71, 95%CI = 0.63–0.81) respectively were negatively associated with the recent CD-4 count in the full model. Conclusions Immunological recovery was achieved by 1/3 of the patients despite being on highly active ART (HAART). Therefore, intensive adherence counseling, follow-up and support should be focused on patients with viral non suppression to enhance immunological recovery.

Introduction specifically for Tigray region and some parts of Northern Ethiopia. Tigray region is the 6th largest by area and the 4th most populous of the 9 Regional States of Ethiopia [18]. The Region had an estimated population of 5,055,999 in 2016. Public health care services in Tigray are delivered through 2 specialized hospitals, 15 general hospitals, 22 primary hospitals, 202 health centers and 712 health posts. In addition, there are more than 500 private health care facilities including two general hospitals [19].
Blood samples are sent to the regional laboratory for VL determination from all the health care facilities offering HIV ART services. The sample referral form contains the following information: name of the patient, Medical Registration Number (MRN), Unique ART number, name of the health facility, some demographic data, clinical, treatment, baseline and recent CD-4 count and reason for determining viral load. The source of the data was from all the people living with HIV, enrolled in ART care for at least 6 months whose blood sample was sent for VL determination through standard sample transportation technique to the regional laboratory/THRI from April, 2015 to January, 2019. The study was done among 19525 patients which had complete data on demographic, clinical, immunological, and viral load in the database of THRI.

Sampling procedure
To come up with the final sample, all records of patients from the Tigray regional state available in the database of THRI were reviewed and then all the study participants which fulfilled the eligibility criteria were included in the study (Fig 1).

Data collection tools and procedure
The data were collected from the database of THRI using a data retrieval checklist. All the data in the database were exported to Microsoft excel 2010 and then data verification and filtration were done before exporting to STATA 14.0.
CD4+ T-cells count was measured at baseline and every six months during follow up at the respective health care facilities or at the nearby referral laboratory where blood samples were transported. Nevertheless, only the results of the baseline and recent CD-4 count of each client was sent via the request form for VL determination to THRI by the health care facilities regardless of the time of initiation of ART. Hence, baseline and recent CD-4 counts were taken for analysis in this study as an independent and dependent variables respectively.

Data quality assurance
Data completeness and consistency was checked using Microsoft excel. Data cleaning was done with box plot and running frequencies for each variable in STATA version 14.0 to check outliers and inconsistencies for accuracy purpose. The normality of the continuous variables was checked by scatter plot matrix and/or normal probability plots. Low and high positive controls were checked during VL determination. During the CD4+ T-cells count determination by each machine; low, medium and high quality controls were done to evaluate run validity in each laboratory, where the CD-4 count was done.

Data analysis
Analysis was done using STATA-14.0 to estimate the proportion of patients with immunological failure or recovery and to identify factors associated with CD-4 T-cells progression over time. Univariate analysis was used to determine the socio-demographic and clinical characteristics of the study population.
Number of recent CD-4 T-cell counts after initiating ART was modeled by using Generalized Linear Model (GLM) with a family of Poisson with a log link function and a robust standard error. Generalized Linear Model (GLM) is an extension of the linear modeling process that allows the model to be fit to data that follow probability distributions other than the normal distribution. Poisson regression model is a special case of a generalized linear model (GLM) with a log link, the reason why the Poisson regression also called Log-Linear Model. [20]. GLM Poisson has three components; namely random, systematic and link. Poisson regression is a form of a GLM where the response variable is modeled as having a Poisson distribution. All the critical assumptions that underlie the GLM, many of which apply to any regression model: such as statistical independence of the "n" observations, the correct specification of the variance function v, the correct specification of ф (1 for Poisson and binomial data), the correct specification of the link function g, correct form of the explanatory variables x, and lack of undue influence of individual observations on the fit were fulfilled to model using Poisson GLM. Appropriate model was selected by the lowest AIC. The strength of association among the independent and dependent variables was determined by the Incidence Rate Ratio (IRR) at 95% Confidence Interval (CI). Statistical significance was considered at pvalue <0.05 (two-sided) in all tests.
All significant variables in the bivariable analysis were entered into the multivariable GLM Poisson regression models based on biological plausibility, previous literature, and statistical significance in bivariable analyses. As pregnancy and lactating status applies to females, both variables were excluded from statistical model building of the multivariable analysis. Multivariable analysis was also done by stratifying the WHO stages to check whether the coefficients of variation remain constant or differ compared to the full model. Collinearity was omitted in the statistical modeling.

Ethics statement
Ethical clearance and approval was obtained from Tigray Health Research Institute (THRI) Institutional Review Board (IRB) / Ethical Review Committee with the reference no of THRI/ 00132/19. Permission to use the data was obtained from the Tigray Regional Health Bureau and THRI. The data were from a secondary database on HIV infected patients for viral load monitoring while on combination of antiretroviral treatment. All the baseline and recent CD4 counts were extracted from the database retrospectively, which were entered from the sample referral form. The data were not accessible by any other third party other than the study team. All the data did not carry personal identifiers. Informed consent was waived from the ethics committee.
Factors associated with CD-4 T-cells progression over time. Bivariable analysis. The bivariable analysis showed that the Incidence Rate Ratios (IRRs) of the independent variables were associated with CD-4 T-cells progression over time. At bivariable analysis level, CD-4 Tcells count at baseline (IRR = 1. 000784, 95%CI = 1.0007-1.0008), Males (IRR = 0.79, 95% CI = 0.78-0.81), and lactating mothers (IRR = 1.08. 95%CI = 1.01, 1.16) were significantly associated with the recent CD-4 T-cells count. Similarly, patients' age categories were significantly associated with the recent CD-4 T-cells count. However being pregnant was not statistically associated with the recent CD-4 T-cells count over time (Table 3) were significantly associated with the recent CD-4 T-cells count. However, patients with a regimen of 1f were not significantly associated with the recent CD-4 T-cells count (Table 3).

Multivariable analysis
A multivariable analysis using the GLM showed that the initial model without any predictor variables have an AIC of 159.3391. The constant was significant with p-value< 0.01 (Table 4).      (Table 5). On the other hand, patients whose viral load was determined for suspected ART failure initial VL compared to routine first VL is expected to have a rate ( (Table 5).
We made a stratified analysis by WHO stage to find out whether the strength of association of independent variables with the recent CD-4 T-cells count remain same or differ. Based on the analysis, baseline CD-4 T-cells count and VL suppression were positively associated and remain the same with the recent CD-4 T-cells count in all the WHO stages compared to the full model (without stratification by the WHO stage). However, being male in gender, age category in 40-44 and 45-49, and suspected ART failure initial VL for VL determination reason were negatively associated and remain almost the same with the recent CD-4 T-cells count in all the WHO stages. Despite these variables, there were differences after the stratified analysis in variables such as: age categories, service provided in military health care facility, facility type, adherence, in some of the virological test reasons and in some of the regimen types. Service provide in military health care facility was only significant in the WHO stage I sub cluster analysis (Table 6).

Discussion
Antiretroviral treatment began in 2003 and free ART was launched in Ethiopia in 2005. An estimated 738,976 Ethiopians are currently living with HIV and all of them require antiretroviral treatment (ART). However, only 426,000 are currently taking ART [21]. Monitoring individuals receiving ART is important to ensure successful treatment, identify adherence problems and determine whether and which ART regimens should be switched in case of treatment failure [21]. This retrospective cross-sectional study was conducted to estimate the proportion of patients with immunological recovery, failure and to identify factors associated with recent CD-4 count progression over time after ART commencement. Immunological failures, recovery, and factors associated with CD-4 T-cells progression over time Table 5  The median baseline CD4+ count of this study was 201cells/μL. This finding was similar to a study conducted in Jimma, Ethiopia, which was 191 cells/μL [22]. However, the current finding was higher compared with studies conducted in the capital city of Addis Ababa, which were 115 cells/μL [23] and 177 cells/μL [24]. Similarly, lower median CD4+ counts were reported from the Tigray region (162 cells/μL) [17], Felege-hiwot referral hospital, Bahir Dar, Ethiopia (147 cells/μL) [25], southern Ethiopia (156 cells/μL) [26] and Kenya (152 cells/μL) [27]. Nonetheless, a study conducted in Liberia revealed a higher median CD4+ count (238 cells/μL) [28]. This variation may be explained due to differences in the minimum standards of CD-4 count to initiate ART treatments and delayed initiation of ART. This could also be attributed due to the poor awareness of the public.

Variables Category Adjusted IRR (95% CI) P-value
The proportion of patients with a baseline CD-4 count of <200 cells/ mm 3 , 200-499 cells/ mm 3 and � 500 cells/ mm 3 was 49.61%, 38.95% and 11.44% respectively in this study. A study conducted in Somali region, Ethiopia showed that 57.6% of the patients had a baseline CD4 + count � 200 cells/ mm 3 and 42.4% had >200 cells/ mm 3 [29]. Similarly, about 15.59%, 45.63% and 38.79% of the participants in this study had a recent CD-4 count of < 200 cells/ mm 3 , 200-499 cells/ mm 3 and �500 cells/ mm 3 respectively. A study conducted at the Felege Hiwot Referral Hospital reported that 30.8% of the patients had recent CD-4 count of � 500 cells/ mm 3 [30]. With regard to achieving a normal CD-4 Count (� 500 cells/ mm 3 ), the current study was similar to the study conducted in the Felege Hiwot referral Hospital where   37.6% of them reached a CD4 count of 500 or more cells/mm 3 [31], while other studies reported a proportion of 45.2% [32] and 59% [33] of the patients achieved normal range of CD-4 count. Potential explanations for the differences could be that Ethiopians might have a relatively lower normal CD4 count compared to other study populations, though it requires verification, differences in the duration of ART, epidemiology of opportunistic infections, age of study participants, and methods used to determine the CD4 count (5,10,12,34). The median (±IQR) recent CD-4 count was 423 (±347) cells/ mm 3 . This was lower compared to the mean CD4 count of healthy HIV negative Ethiopians which was 820 (±270) [34]. This indicates that the difference in the duration of ART might not be enough to achieve CD-4 count plateau. Immunologic recovery was attained by only 28.35% of the patients, which was lower when compared with another study 45.2% [32]. This study has revealed immunological failure in 2.14% and 19.42% of the patients based on WHO 2016 and 2013 definitions respectively [11,12]. Considering the recent WHO definition, this finding was lower compared with the study conducted in Felege-hiwot hospital Bahir Dar, Ethiopia (15.9%) [25], 6.5% in Tigray region [17], 15.7% in Federal Police Referral Hospital (FPRH) Ethiopia [23], 13.2% in Gondar referral Hospital [16], 6.1% in the Somali region of Ethiopia [29]. However, based on the findings of the WHO definition 2013 immunological failure were higher compared to different studies [16,17,23,25,29]. These variations might be due to the application of different WHO definitions, which was changed over time or it could be a real difference.
Good adherence compared to poor adherence was positively associated with CD-4 count in this study. Similar studies from Ethiopia, Kenya, Colombia and France have reported that patients with non-adherence to treatment were associated with immunological failures when compared with good adherence [17,27,29,35,36]. This can be justified, in the case of nonadherence, the level of antiretroviral drug concentration in the blood might not suffice to suppress the viral RNA replication and this in turn deplete CD-4 T-cells in the blood.
Higher baseline CD4 count was positively associated with the recent CD-4 count during treatment. Different studies have also supported this finding that higher baseline CD4 count was positively associated with the time to immunologic recovery [30,31]. Another study has also reported a baseline CD4 count was positively associated with the recent CD-4 count [37]. However, there were contradicting studies which reported a negative association between baseline CD4 count and CD4 count during the treatment period [38,39]. In fact, the nature of the parameters estimated in these contradicting studies is different. The current study measured the gain in CD4 count from each baseline CD4 count without considering the time variation from initiating the ART; but these studies which reported negative associations estimated the average rate of increment or slope per unit time (considering time variable) which could diminish over time until it reaches a plateau.
Age 15-19 years was positively associated with an increase of the recent CD4 count. There was some evidence which showed that commencing HAART at a younger age may be associated with an improved immunologic response [40,41]. On the contrary, other studies reported a null association between age and CD4 count increment [32,38]. These studies have a small sample size [32,38] which may not ensure homogeneity of different age groups.
Females compared to males have positive associations with the recent CD-4 cell count. This finding did not agree with the finding [31], however, this finding agrees with different studies [38,42,43]. A similar study has also shown that females had better response to ART as compared to males [43]. This can be justified as females could attend voluntary counseling as part of their routine health care services during pregnancy, while male patients are poor in their health seeking behavior as they experience lower rates of HIV testing, and acceptance of linkage to HIV-care after a positive result [44]. As a result, females are more likely to be diagnosed in HIV infection earlier than males [45,46], and this could make the response to ART treatment poor for lately diagnosed males as the immune system could be damaged irreversibly in advanced stages of the disease [47]. Therefore, health seeking behavior and early presentation might have been different in males and females.
This study showed that patients in WHO stage II and III were negatively associated with increment in the recent CD-4 count compared with the WHO stage I. Different other studies have also shown that WHO clinical stage and classification of HIV/AIDS correlates well with CD4+ T-lymphocyte counts [48,49]. Although WHO stage IV is the reason for poor immunological recovery from the science of knowledge, however, WHO stage IV was not statistically associated for recent CD-4 count in this study. The absence of statistical association might partly be attributed due to low sensitivity of WHO clinical stage in predicting CD-4 count [50] and/or it could be a real scenario. The other reason might be, about 44.81% and 1.91% of the patients with less than 200 CD-4 T-cells/ mm 3 in this study were classified in WHO stage I and II respectively. However, patients with <200 CD-4 T-cells count are more likely to have advance HIV infection that might be classified into WHO staging IV or III. This shows that there was wrong way of classifying patients into WHO stages at baseline and hence affects the association.
Patients with viral suppression were positively associated with the recent CD-4 T-cells count. A similar other studies from Ethiopia, Nepal, Thailand and India reported that patients with HIV RNA level �1000 copies/ml were more likely to experience immunological failure in ART follow up as compared with those who had an HIV RNA level less than 1000 copies/ml [17,51,52,53]. Another study has also indicated that maintaining virological suppression results in greater increases in CD4 cell counts in the long term [54]. This is because as the load of viremia decreases, the depletion of CD-4 cells might be decreased.
Patients from military health care facilities had a higher CD-4 cells count as compared to patients from non-military health care facilities. Even though there were no similar studies which support or contradict this finding, patients from military health care facilities were mostly military staff that might have relatively higher adherence to treatment and nutritional status as compared to civil patients from non-military facilities which then results in a boost of CD-4 cells count after Antiretroviral treatment.
Patients with a first line regimen 1e (TDF-3TC-EFV), and second treatment line regimens 2f (AZT-3TC-ATV/r) and 2h (TDF-3TC-ATV/r) were negatively associated with recent CD-4 count as compared to 1c (AZT-3TC-NVP). Though there were no similar studies in relation to recent CD-4 count and antiretroviral regimens, there was reported finding that initiating ART using any one of the following ART regimens: 1c (AZT-3TC-NVP), 1d (AZT + 3TC + EFV) and 1e (TDF + 3TC + EFV) prevented treatment failure [51]. Even though there is concordance with 1c, the variation with other regimes might be due to the small size of sample size in the later study, which might not insure the accuracy by avoiding sampling errors such as large variability, bias or under coverage.
Compared to routine first Viral Load (6 months or more on ART); suspected ART failure initial Viral Load, suspected ART failure immunological and suspected ART failure clinical were negatively associated with the recent CD-4 count. There was a similar finding from Uganda, where being a suspected treatment failure patient and repeat test after suspected failure were positively associated with viral non suppression [55]. This can be justified, as HIV viremia increases the probability of CD-4 T-cells depletion might increase.

Strength and limitations of the study
This study was not without limitation. Due to the nature of secondary data, there were no data when the baseline and recent CD-4 T-cells count was conducted and hence the time when the immunological failure/ recovery had occurred is not known. In addition to the time of ART exposure the analysis misses some important variables such as the existence of co-infection and grade of ART experience in the HIV infected patients. Despite these limitations, this study was done on a large sample size with appropriate statistical analysis techniques that provides important information regarding viability of ART treatment program in Tigray, Northern Ethiopia.

Conclusions
Immunological recovery was lower, only one third of the patients achieved immunological recovery despite being on highly active ART (HAART). Sex, CD4 baseline, age, ART clients served in defense facility, WHO Stages II and III, adherence, virological test reasons of suspected ART failure clinical, suspected ART failure immunological and suspected ART failure initial Viral Load, regimen type of 1e, 2f and 2h and viral load suppression were significantly associated with CD-4 T-cells progression over time, among HIV patients on ART in Tigray region. Therefore, intensive adherence counseling, follow-up and support should be provided focusing on patients with virally non suppressed individuals so as to enhance immunological recovery. In addition, determinants of immunological recovery need to be investigated in detail to design an appropriate intervention.