Drug resistance patterns following pharmacy stock shortage in Nigerian Antiretroviral Treatment Program

We conducted a cross-sectional evaluation of samples from ART patients that experienced VF between December 2004 and August 2005 at the Jos University Teaching Hospital (JUTH), a Harvard/AIDS Prevention Initiative Nigeria (APIN) Presidents Emergency Plan for AIDS Relief (PEPFAR) Program-supported GoN ART site; all patients had enrolled for ART between June 2004 and June 2005. Interrupted patients were those that did not receive their regular medications between November 2003 and January 2004, while uninterrupted patients were those that were able to receive their regular medication during the same time period. Inclusion criteria for this evaluation consisted of: enrollment in National Program prior to January 2004, ≥ 12 months on ART, age ≥ 18 years, and viral load (VL) count ≥ 10,000 copies (cp)/mL at or after VF.

Plasma virus from patients in VF was genotyped in the reverse transcriptase (RT) and protease (PR) genes using the ViroSeq HIV-1 Genotyping System 2.0 Assay (Abbott, Chicago, IL). All sequence data were edited and analyzed to generate lists of mutations and polymorphisms. Mutations were categorized according to the International Antiviral Society (IAS)-USA recommendations [2]. Drug susceptibility was based on the Stanford University HIV Drug Resistance Database (HIVdb) algorithm [3, 4].

Following enrollment in the Harvard/APIN PEPFAR-supported GoN program, patient data were collected and stored using an electronic medical records system (EMRS) [5]. For the statistical evaluations, patient data, including ART regimen, date of birth (for age calculation), sex, occupation, education, marital status, WHO stage at entry, HIV subtype, CD4+ cell count at program entry and at sample dates, VL at entry and sample dates, and adherence were abstracted from the EMRS. Time of ART initiation was defined as the date the patient initially started taking ART in the GoN program, time of program entry was defined as the date the patient enrolled in the Harvard/APIN PEPFAR-supported ART program at JUTH, and sample date was defined as the date the patient’s sample was collected or DRM testing (following program entry). VL was stratified based commonly accepted clinical categories: < 400; 400–999; 1000–9999; 10,000–99,999; and ≥ 100,000 cp/mL (i.e., high VL).

Patient characteristics at the time of program enrollment were compared for interrupted versus uninterrupted patients using bivariate methods, including Wilcoxon rank-sum test for continuous variables and Chi square or Fisher’s exact tests for categorical variables, as relevant. We examined the potential association between HIV-1 subtype to DRMs and resistance using the Fisher’s exact test; for statistical analyses, A1, A3, CRF06_cpx and unknown were combined to make an “other” category. To assess if adherence patterns could potentially explain the difference in VL values at sample date, we evaluated the association using Fisher’s exact test. To evaluate the predicted impact of substituting the d4T+3TC+NVP regimen to the recommended alternative regimen of zidovudine (AZT)+3TC+efavirenz (EFV), for patients failing at that time, a genotype susceptibility score (GSS) to the alternative recommended regimen was calculated. GSS was computed based on the drug resistance scores extracted from the Stanford HIVdb. Each ARV drug in the alternative recommended regimen was assigned a score according to the five-level Stanford HIVdb interpretation: 1.00 for susceptible, 0.75 for potential low-level resistance, 0.50 for low-level resistance, 0.25 for intermediate resistance and 0.0 for high-level resistance [6]. The GSS was the sum of all scores. Percentage of patients with a low GSS were compared by drug interruption status using the Wilcoxon rank-sum test. Due to the nature of the small sample size and the fact that we were underpowered to generate a multivariate model on the data, we also conducted a stratified analysis to compare GSS by drug interruption status in only those patients who had < 12 months between entry and their sample date. All statistical analyses were performed using Stata 13.1 (College Station, TX).

All patients provided written informed consent for inclusion in the cohort. This study was approved by the Institutional Review Boards at JUTH and the Harvard T. H. Chan School of Public Health.

In total, there were 56 patients included in this evaluation: 28 that experienced interruption and 28 that did not. We found that patients in the interrupted group only differed from the uninterrupted group in gender breakdown, time from ART initiation to sample date, and time from program entry to time of sample date. There was a higher percentage of males in the interrupted group as compared to the uninterrupted group (57.1% vs. 21.4%; p = 0.006; Table 1). We found that the time from both ART initiation (AI) and program entry to sample date was longer in the interrupted patients versus uninterrupted patients (AI to sample date: 365 vs. 213 days; entry to sample: 1036 vs. 601 days; p < 0.0001). At the time of sampling, 14.3% of the interrupted patients versus 10.7% of the uninterrupted patients had VL ≥ 100,000 cp/mL (p < 0.001). While 100% of the patients in the interrupted group versus 35.7% of those in the uninterrupted group had VL ≥ 10,000 at the sampling date, we found no association between VL at sample date and adherence patterns during the time period of program entry to VL sampling date; the data indicated that 72.7% of patients with an average adherence of < 95% had VL ≥ 10,000 cp/mL compared to 64.7% of patients with an average adherence of ≥ 95% (p = 0.57; data not shown). Median CD4+ cell count at sample date for the interrupted patients was 123 cells/mL as compared to 257 cells/mL for the uninterrupted patients (p = 0.004). The majority of the patients in the cohort did not achieve an undetectable VL following enrollment in the program and there was no difference by drug interruption status.

In the interrupted arm, patients had a median of five DRMs [interquartile ratio (IQR): 3–6] compared to 3 (IQR: 0–4) in the uninterrupted arm (p = 0.001). The most common nucleoside reverse transcriptase inhibitor (NRTI) mutation for the cohort was M184I/V (76.8%), where 25 (89.3%) of the interrupted patients and 18 (64.3%) of the uninterrupted patients harbored the mutation at the time of sampling (p = 0.06; Fig. 1a). Thymidine analog mutations (TAMs) were only found in patients that harbored M184I/V (62.8% of those with M184I/V and 0% of those with no M184I/V; p < 0.001; data not shown). The only NRTI mutation for which we detected a statistically significant difference by drug interruption status was T215F, a TAM-2 (35.7% for interrupted vs. 3.6% for uninterrupted; p = 0.005; Fig. 1a). We also found that interrupted patients were more likely than uninterrupted patients to have one or more TAM-2 mutations (57.1% interrupted vs. 21.3% uninterrupted; p = 0.04). The most common non-NRTI (NNRTI) mutation was Y181C, but the difference between patient groups was not statistically significant. The only NNRTI for which we found a difference in proportion harboring the DRM by drug interruption status was V108I (25.0% interrupted vs. 0% uninterrupted; p = 0.01). The only mutation for which we detected a statistically significant difference by subtype was A98G, where patients infected with subtype G or CRF02_AG were more likely than those with G-prime or any other mutation to have the A98G mutation (CRF02_AG: 26.1%; G: 33.3%; G-prime: 0.0%; Other (all others combined): 0.0%; p = 0.03; data not shown); however, the association did not remain once we stratified by drug interruption status (data not shown).

While the study was not powered to detect differences by drug interruption status for the majority of the individual mutations, we found that a higher percentage of patients overall were resistant to the NVP and 3TC components of their ART regimen as compared to their d4T component (Fig. 1b). The majority of patients retained susceptibility to tenofovir (TDF). We were not able to show a statistically significant difference in proportion of patients with resistance to 3TC, FTC, NVP or EFV by drug interruption status. However, we did find a statistically significant difference by drug interruption status for resistance to both d4T (53.7% interrupted vs. 17.9 uninterrupted; p = 0.011) and AZT (64.3% interrupted vs. 25.0% uninterrupted; p = 0.003). We were also not able to detect a statistically significant difference in drug resistance by HIV-1 subtype (data not shown).

The overall median GSS to alternate regimen for the study population was 1.00 (IQR: 0.25, 3.00), with the median being 1.25 (IQR: 0.75–3.00) for the interrupted group and 0.50 (IQR: 0.25–1.00) for the uninterrupted group. Using a cut-off of GSS < 1, we found that 19 (67.9%) of the interrupted patients, as compared to 7 (25.0%) of the uninterrupted patients, did not have full susceptibility to one drug in the regimen to which guidelines recommended they be switched (p = 0.001). When we limited the GSS comparison to only those patients that had < 12 months between entry into the PEPFAR program and their sample date, GSS remained 1.25 (IQR: 0.75–3.00) for the interrupted group and dropped to 0.38 (IQR: 0.25–1.25) for the uninterrupted group, where 9 (56.3%) of the interrupted patients versus 7 (25.0%) of the uninterrupted patients had susceptibility to none of the recommended drugs (p = 0.04).