Study design
A prospective, observational study using data from ten Rwandan health centers that participate in the Central Africa International Epidemiologic Databases to Evaluate AIDS (www.iedea.org) was conducted. CA-IeDEA is a multi-country project collecting secondary data from patients receiving HIV care and treatment in the Central African region; one of seven regions that comprise a global IeDEA network (www.iedea.org).
Setting and population
In July 2016, Rwanda, an East African country with 12 million persons and an overall 3% HIV prevalence(11), became one of the first countries to implement Treat All nationally. Under this policy, all persons living with HIV (PLWH) are referred for ART immediately upon diagnosis after performing a routine serum creatinine test to determine regimen choice (12). Current ART guidelines include TDF as a component of the main first-line treatment and recommend a serum creatinine measurement prior to initiating ART, after one month of use, and then yearly while on TDF. For this analysis, we included adults (≥18 years receiving HIV care at CA-IeDEA affiliated health centers who initiated a first-line TDF-based regimen from 1 July 2016 through 30 July 2018, and whose serum creatinine was measured both before (within 12 months) and after (within 6 months) of TDF initiation.
Patients with diabetes and/or hypertension, those on second- or third-line ART, and those who did not have creatinine measured both before and after TDF initiation were excluded.
Among 1,536 patients who initiated TDF during the study period, i) 205 (13.3%) did not have any creatinine measurements within 12 months before or 6 months after starting TDF, ii) 456 (29.6%) had only pre-TDF creatinine measurements, and iii) 399 (22.0 %) had only post-TDF creatinine measurements. Thus iv) 476 (31.0 %) patients had both pre- and post-TDF creatinine measurements. A comparison of these above four groups demonstrated statistically significant differences in proportions who were women, and mean age, cluster of differentiation 4(CD4) count, creatinine and eGFR. Details are provided in the Supplemental Table and Supplemental Figure.
Outcomes and predictor variables
The primary outcome was change in eGFR after initiating TDF. We considered pre-TDF creatinine as the closet measure within one year prior to initiating TDF, and post-TDF as the first post-TDF initiation measurement within one to six months after initiation.
Serum creatinine was measured in mg/dl. The Chronic kidney disease (CKD)-EPI equation estimated GFR: eGFR = 141 x min (SCr/κ, 1) x max(SCr /κ, 1)-1.209 x 0.993Age x 1.018 [if female] where κ = 0.7 if female and 0.9 if male and = -0.329 if female and -0.411 if male . The North American adjustment for sub-Saharan African race was not used as this adjustment has been shown not to apply within Africa itself and moved the estimates in our population towards the non-normal range. We also examined the change in creatinine after initiating TDF. Additional baseline variables evaluated here included sex, age, body mass index (BMI), and CD4 count.
Analyses
Because creatinine had a right skewed distribution, it was log transformed to shift the distribution toward normality. Continuous variables were summarized as means and reported with standard deviations (SDs), skewness, kurtosis, medians, quartiles and ranges and categorical variables were reported as percentages. Time on TDF at the post-TDF creatinine measure was calculated as the difference between the date of the post-TDF initiation creatinine measurement and the date of TDF initiation. We evaluated eGFR, log-transformed creatinine, and pre- to post- TDF initiation changes in these values as continuous variables. T-tests and nonparametric tests compared continuous baseline characteristics while chi-square tests compared categorical characteristics among groups of patients. All adults initiating TDF in the study period defined, based on pre and post creatinine measurements they had have been recorded. Paired t-test and sign tests compared changes from baseline to follow up for continuous variables and McNemar’s discordant pairs tests did so for binary variables.