To our knowledge, the current study is the largest analysis to date investigating the natural progression of HIV disease in sub-Saharan African adults not initially receiving antiretroviral therapy. As such, we could investigate with high confidence multiple predictors of both eventual need for cART and overall mortality.
Our primary goal was to evaluate the utility of renal function to predict HIV-related outcomes. We found that lower renal function, defined either as estimated CrCl < 60 mL/min or as estimated eGFR < 60 mL/min/1.73 m2
, at enrolment was independently associated with an increased risk of HIV disease progression. Our results differ from the only other study to assess renal abnormalities as predictors for AIDS progression in patients not receiving cART [5
In analyses of the Women's Interagency HIV Study (WIHS) cohort, Szczech et al
] found that dipstick proteinuria, but not inverse creatinine, was significantly associated with the development of a new AIDS-defining illness. Several reasons may explain the differences in results. The WIHS cohort included only women, whereas our study included both men and women. Differences in diet and environmental conditions may also have contributed to the discrepant results. The definitions of renal function also differed between our analyses. Szczech et al
used inverse creatinine as a continuous predictor variable, while we used categorical definitions of both estimated creatinine clearances and glomerular filtration rates. Perhaps most importantly, the WIHS cohort analysis could adjust for multiple other potentially confounding variables, including HIV-1 RNA levels, proteinuria, albuminuria and presence of other co-morbidities (hepatitis C co-infection, diabetes, hypertension), which we did not have available in our study cohort.
We did not find in weight-adjusted analyses that renal function was associated with overall mortality. Again, our results conflict somewhat with those from the WIHS analyses, in which inverse creatinine predicted mortality in women who did not receive cART. In addition, Gardner et al
] found that American women enrolled in the HIV Epidemiology Research Study (HERS) before the availability of cART with either a serum creatinine ≥1.4 mg/dL or proteinuria ≥2+ on urine dipstick had a significantly greater risk of death.
The differences between our study and the HERS study may have occurred for similar reasons as noted already between our African cohort and the WIHS cohort. However, in follow-up analyses from the WIHS cohort, Estrella et al
] found that having an eGFR < 60 mL/min/1.73 m2
prior to initiation of cART was associated with higher mortality. In addition, a large Zambian study of nearly 26,000 patients initiating cART [6
] found that 90-day mortality rates after cART initiation were significantly higher in patients with reduced baseline renal function. The lack of association between reduced renal function and mortality in those initiating cART in our study may have occurred due to a relative lack of power since only 1946 subjects eventually received cART in our cohort. In our experience, the mortality rates in the proportion of patients who are lost to follow up are significantly higher than those observed among patients retained in care; as such, high rates of loss to follow up may have impacted this outcome [18
The mechanisms by which reduced renal function may lead to faster HIV disease progression are not completely clear. The most likely explanation is that the observed relationships may be confounded by the lack of adjustment for HIV-1 RNA levels and increased systemic inflammation, both of which are related to HIV disease progression and renal function [20
]. Additional studies that incorporate these HIV disease progression markers are needed to better understand the relationships between renal dysfunction and outcomes in both resource-limited and resource-rich environments.
In patients with low muscle mass, low serum creatinine values may more likely reflect reduced creatinine generation even in the face of renal function impairment. Thus, the use of serum creatinine alone to estimate renal function would not be appropriate for the current study cohort. Given the presence of patients with protein malnutrition and HIV wasting in our cohort (both etiologies of muscle wasting), we chose to use estimated renal function using the two most common equations currently in practice, namely the Cockcroft-Gault equation and the 4-variable MDRD equation, which incorporate variables that should adjust for variability in muscle mass. As such, both equations include not only serum creatinine, but also age and sex. The Cockcroft-Gault equation, in contrast with the 4-variable MDRD equation, also includes weight. Our results demonstrate that the specific inclusion of weight in the Cockcroft-Gault equation greatly influenced the prevalence estimates of reduced renal function estimates in this Kenyan population not yet receiving cART.
Our results corroborate those from another HIV-infected African cohort [14
] in which the prevalence of renal dysfunction was much greater when using the Cockcroft-Gault equation compared with the simplified MDRD equation. In addition, adjustment for weight in the CrCl prediction models reduced the association between reduced CrCl and HIV disease progression and completely negated the relationship between lower CrCl and mortality in our study. The importance of weight in our analyses should not be surprising given that lower weight has long been known to be associated with decreased survival in those infected with HIV [24
]. In addition, it should also be noted that the lack of associations between renal function and outcomes in our models using CrCl and eGFR as continuous variables suggest that the renal function may only be associated with outcomes once a critically low threshold is met and not at higher values.
Several limitations should be acknowledged. As mentioned earlier, the retrospective design relied on using existing data, so several other potential predictors of clinical outcomes, such as HIV-1 viral loads, proteinuria, C-reactive protein, metabolic abnormalities and viral hepatitis co-infection status, could not be studied. Because serum creatinine was not calibrated to the MDRD reference laboratory, bias may have occurred and would limit comparisons with other populations [26
]. We acknowledge that missing data, including serum creatinine values, in a substantial number of the USAID-AMPATH enrollees, may limit generalizability. However, the very large sample size of the analysis cohort and its similarity to the excluded patients greatly mitigates this limitation. Also, the results of this study may not extend to those groups who were excluded from these analyses, namely women who became pregnant during the study period. However, we believe our results may be generalizable to other sub-Saharan African cohorts.
In our study, approximately 20% had CrCl < 60 mL/min and 9.4% had stage 3 chronic kidney disease, as defined by the National Kidney Foundation as an estimated eGFR < 60 mL/min/1.73 m2
. These proportions are similar to published reports of the frequency of renal dysfunction in patients in Zambia, Uganda, and Zimbabwe [6
]. In addition, our cumulative probability of 22% for meeting cART initiation criteria over the first year is similar to a previous Ugandan study [27
] investigating the natural progression of HIV infection to WHO stage 4 disease (26%) for those who had either stage 1 or 2 disease at initial diagnosis. The relatively short follow-up period may have also limited our ability to find significant associations between reduced renal function and mortality in several of our models. Finally, we acknowledge that neither the Cockcroft-Gault equation to estimate CrCl nor the simplified MDRD equation to estimate eGFR has been fully validated in an antiretroviral-naïve HIV-infected population. Thus the accuracy of these estimating equations to reflect true renal function in sub-Saharan African patients is not known.