Cox regression of factors associated with the onset of chronic kidney disease in HIV/AIDS patients in Albert Luthuli Municipality of South Africa

Background: Current research indicates that chronic kidney disease is a global problem which poses a major health threat to people of poor countries with HIV/AIDS and on antiretroviral treatment. In this study, the prevalence of chronic kidney disease and the factors associated with it were investigated among the HIV/AIDS patients in a rural community of South Africa. Methods: A cohort of HIV/AIDS patients was retrospectively followed from 2010 to 2017 until chronic kidney disease was diagnosed or until the end of the observation period at two hospitals (Carolina and Embhuleni). Patient information was obtained from the routine hospitals’ records, and the data were analysed using Cox regression and survival analysis (Kaplan-Meier hazard functions and ratios, and log-rank tests) methods. Results: Out of a random sample of 320 HIV/AIDS patients, 51 patients (15.9%) had chronic kidney disease. The factors associated with chronic kidney disease were: gender (p-value=0.0356), age (p-value=0.00077), baseline creatinine (p-value=0.00253), follow-up alanine transaminase (p-value=0.0152), ART treatments (p-value<0.00193) and hospital (p-value=0.00258). Discussion: Whilst antiretroviral treatment is associated with some improvement in virology and immunology in HIV-infected patients, research is still needed for the assessment of the impact of ART and other risk factors on renal function in marginalised communities in Africa. Conclusion: The research findings on HIV/AIDS patients in Albert Luthuli Municipality concurred with several previous research findings on risk factors to CKD. The expected action to alleviate the health threat due to CKD in South Africa is to educate the nation on prevention, early detection and on the management of the disease.


Background
HIV/AIDS has been a major health problem worldwide for more than three decades now.
According to the WHO global statistics: since the beginning of the epidemic, 75  prevalence rate after KwaZulu-Natal province. Gert Sibande district which is in Mpumalanga province is leading all districts in the country with 46.1% HIV prevalence rate (Motsoaledi, 2013). Gert Sibande district has Albert Luthuli as one of its municipalities whose HIV prevalence stood at 43.2% (Nkosi, 2017).
Human Immunodeficiency virus (HIV) affects every organ system in the body by direct damage or by rendering the host susceptible to opportunistic infections (Mekuria et al., 2016).
The commonest sites of infection include kidneys. A kidney contains millions of nephrons which are made up of blood vessels called glomeruli. The function of kidneys is to filter blood of waste products such as creatinine and urea and extra water which is removed in form of urine. Chronic kidney disease (CKD) is a slow and prolonged failure of the kidneys to filter impurities from the blood. The progression of the kidney damage is marked by the rise in creatinine and urea whose evaluation in serum helps to assess Glomerular Filtration Rate (GFR). CKD poses a major health threat to people living in poor countries, especially when it is combined with HIV, antiretroviral treatment (ART) or communicable and noncommunicable diseases (Glaser et al., 2016). The objective of this study is to use Cox regression method to determine the prevalence of CKD and to identify the risk factors associated with CKD among HIV/AIDS patients on ART by using data from Albert Luthuli municipality hospitals in South Africa.

Clinical determination of CKD status
The patient's CKD status was determined by referring to the South African National Health where =( 1 , 2 , … , ) is a p-dimensional vector of regression coefficients to be estimated from the data, and ℎ 0 ( ) is the unspecified baseline hazard function that does not have to be estimated. The hazard model in model (1) makes no assumptions about the shape of the hazard function over time. A hazard function could be constant, increasing, decreasing, or it could be a combination of two or three of these graph trends.
Model (1) can be written in terms of the survivor function (Kleinbaum & Klein, 2012) The assumptions of model (1) which may be violated by the data are: (i) the covariates do not vary with time and hence the hazard rate ratios of pairs of patients do not vary with time; (ii) censoring and survival are independent; and (iii) the log hazard rate is indeed a linear function of the covariates. Model (1) was fitted to the data using coxph function in R Version 3.5.1 to determine the risk factors associated with CKD. Firstly, the variables in Table 1 were In this study, the groups were the levels of the significant categorical covariates or categorized for ≥ 0, where ℎ ( ) is the hazard function of the i th group patients. Data for the study were recorded on research tools and then captured on Microsoft excel database and checked against original records by two competent individuals. Data was analysed using R Version 3.5.1.

Cox model fit for risk factors associated with the onset of CKD
The function 'coxph()' in R fits a Cox proportional hazards regression model.

Kaplan-Meier survival functions and hazard ratios for baseline creatinine
Baseline creatinine was categorised as done in the standard Laboratory report on creatinine in South African hospitals. Figure 1 shows that survivor functions of the baseline creatinine strata are statistically different at 0.05 significance level (p-value= 0.004, from the log-rank test) and this confirms that baseline creatinine level is associated with CKD hazard. This is further confirmed by the pairwise comparisons of the strata using the hazard ratios in Table 2 which show that patients with baseline creatinine level which is above normal level are about 2.5 times likely to experience CKD relative to patients with baseline creatinine which is at normal level. This risk can be as low as about 1.3 times and as high as about 4.7 times ( Table 2). The difference in CKD experience between patients with baseline creatinine level below normal relative to patients with baseline creatinine at normal level is statistically not significant ( Table   2). Table 3 shows that female patients relative to male patients have a reduced risk to CKD of about 0.6 times. However, the effect of gender is statistically not significant in the experience of CKD hazard. This is confirmed by the survivor functions of gender strata which are not statistically different at 0.05 significance level (p-value =0.05, from the log-rank test)( Figure   2) and also by the confidence limits for gender strata in Table 3 which include '1' which is a value of no effect.  Kaplan-Meier survival functions and hazard ratios for age groups Figure 3 shows that survivor functions of the age strata are not statistically different at 0.05 significance level (p-value= 0.08, from the log-rank test) and this is confirmed by the confidence limits for strata in Table 4 which include '1' which is a value of no effect.

Kaplan-Meier survival functions and hazard ratios for gender
Patients who are above 50 years relative to patients aged 16-30 years are about 2.7 times likely to experience CKD (Table 4). Patients aged 31-50 years relative to patients aged 16-30 years are about 1.5 times likely to experience CKD. However, this difference is not significant since the 95% confidence limits include 1 (a result of no effect) and also the graphs of 31-50 and 16-30 age groups cross each other as shown in Figure 3.

Kaplan-Meier survival functions and hazard ratios for treatment (regimen one)
The five treatment groups are statistically different (log rank p-value=0.003) (Figure 4). Treatment EFV+D4T+3TC relative to treatment NVP+D4T+3TC has a reduced risk to CKD of about 0.34 times, Treatment EFV+AZT+3TC relative to treatment NVP+D4T+3TC has a reduced risk to CKD of about 0.1 times while treatment EFV+3TC+TDF relative to NVP+D4T+3TC has a reduced risk to CKD of about 0.4 times. The effect of treatment NVP+3TC+TDF relative to NVP+D4T+3TC is not statistically significant in experiencing the risk of CKD since the graphs cross ( Figure 4) and the 95% confidence internal includes a value of no effect '1' ( Table 5).      The baseline findings from Figure 1 and Table 2 show that patients with baseline creatinine which is above normal level are about 2.5 times likely to experience CKD relative to patients with baseline creatinine which is at normal level. In addition, patients with baseline creatinine which is below normal level are about 0.6 times likely to experience CKD relative to patients with baseline creatinine which is at normal level. These findings reinforce the need for accuracy in creatinine testing and for the inclusion of creatinine in GFR equation.
Gender is significant in modelling (Table 1) but not significant in the testing of the hypothesis of no difference of homogeneity using the Log-rank test and Kaplan-Meier functions ( Figure   2, Table 3). This difference is because Cox modelling accounts for the effects of other factors whereas Kaplan-Meier is a univariate analysis which does not consider the effects of other variables. Research study in Albert Luthuli found out that being a male increases the hazard of having CKD by about 1.6 times. The significance of gender as a risk factor to CKD is reinforced by its inclusion in the MDRD formula.
Another significant researched risk factor to CKD and which happens to be included in the GFR formula is age. The fitting of age in the Cox regression in this study reinforces its inclusion in GFR formula. The effect of age on CKD is due to glomerular filtration rate which declines at 6.3 ml/min/1.73m 2 per decade as cited by Denic et al. (2016). Findings on age in Table 4 indicate that patients aged above 50 years relative to patients aged 16-30 years are about 2.7 times likely to experience CKD.
The finding that CKD is associated with baseline creatinine, age and gender is not novel as these are the components of the GFR formula (Iseki, 2008;Omuse et al., 2017). Therefore, the concurrence of the regression results with the existing formula confirms the appropriateness and accuracy of the research methods used.
On the management of CKD; Pedro et.al (2016) supported the use of TDF but hinted on its nephrotoxic potential and pointed out the need for the adjustment of the dose when baseline creatinine clearance is below 50ml/min. Boswell and Rossouw (2017) pointed out that current local guidelines recommend that TDF be substituted by an alternative Nucleoside reverse Transcriptase Inhibitors (NRTI) when a patient's GFR is less than 50ml/min and when using MDRD method. Moosa et al. (2015) gave some guidelines on dose adjustments for ART in CKD cases for Lamivudine (3TC), Stavudine (d4T) and Tenofovir (TDF). Some other points to note on the management of CKD as cited by several studies are: • All individuals should be assessed for kidney disease at the time of diagnosis of HIV/AIDS and annually thereafter.
• Prevention of CKD at the population level through interventions that lead to a reduction in BP, obesity, type 2 diabetes mellitus, smoking and salt ingestion.
• Careful screening and monitoring of high-risk patients.
• Increased awareness of CKD and knowledge on how to manage kidney disease should be emphasized for general medical practioners.

Conclusion
This study, through the application of Cox regression, found the factors associated with CKD as: gender, age, baseline creatinine, hospital, treatment (regimen 1) and Baseline alanine transaminase. Given the HIV/AIDS background and the diversity of the worsening risk factors to CKD in Africa and in South Africa, there is need for some concerted effort to alleviate health and economic burden caused by CKD. This research has presented risk factors associated with CKD in Albert Luthuli Municipality and the expected action is to educate the nation on prevention, early detection and on informed management of CKD. The study established diverse baseline statistics against which future research may be based.

Ethics declarations Ethics approval and consent to participate
This study was conducted in accordance with the South African local and national research guidelines. Ethics approval for this project was obtained from UNISA Ethics Review Committee (ERC) with the approval number being (2017/SSR ERC/005). The permission to conduct the study at Carolina and Embhuleni hospitals was obtained from Mpumalanga Department of Health with the permission number being (MP_201708_013). The consent to participate does not apply to a patient since no reference to an individual respondent was made, all results were handled in aggregate format.

Availability of data and materials
The data that support the findings of this study are available from the Department of health, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the Department of health.

Consent for publication
Not applicable.