We enrolled 692 participants all of whom initiated cART with non-nucleoside reverse transcriptase-based therapy at least 6 months prior to enrollment (median age=11.9 years (IQR 10.2–13.6, range 8–16.9); median treatment duration=56 months (IQR 37–72, range 6–134); 50.3% female) from Francistown and Maun, Botswana. More than 90% of subjects were perinatally HIV-infected with 12 (1.7%) reportedly infected via breastfeeding and 1 (0.1%) infected via rape. Orphan status was common (52.6%) with 102 (28%) being double-orphans, 261 (72%) being single-orphans, but only 18 (5%) residing in an orphanage. Most children (98.7%) had been enrolled in school with 96% being within 2 years of the age-appropriate grade level. The median age at initiation of cART was 7.4 (IQR 5.4–9.3) years. At baseline 48% were CDC immune category 3 and 6.8% were CDC immune category 3 at the time of enrollment in our study. Virologic failure had occurred in 161 (23.3%) patients at some time in their treatment history.
A score cut-off of 20 on the PSC was chosen based on the ROC curve analysis. The chosen cut-off gave us an accuracy of 92% for correctly classifying subjects according to our standard measure. Using the cut-off score of 20, 120 (17.3%) patients had high scores. The median PSC score for those with virologic failure was 12 (IQR 7–19) and the median for those without virologic failure was 10 (IQR 5–16, p<0.01). A scatter plot with a fitted line showing the proportion of subjects with each PSC score who had virologic failure demonstrates increasing proportions with virologic failure among those with higher PSC scores (). Virologic failure was more common among those with a PSC score ≥20 (31.7%) than among those with a PSC score of <20 (21.5%, unadjusted OR 1.7, 95% CI 1.1–2.6; sensitivity 23%, specificity 85%, PPV 32%, NPV 78%). The results were not significantly altered by controlling for demographic or clinical factors, including age, age at initiation of cART, sex, nadir CD4, orphan status, primary caregiver, and HIV disclosure status.
Proportion of subjects with virologic failure by PSC score
From the full-length (35 question) PSC, 9 questions had a statistically significant test of trend for a higher likelihood of failure in those with higher scores at the 0.05 level and 6 questions had a statistically significant test of trend at the 0.01 level. shows the distribution of test scores and results from the test of trend for the 17 questions that make up the brief-report version of the PSC. Questions which tended to be scored higher among patients with virologic failure were mainly related to attention/executive dysfunction, and depressive symptoms.
Distribution of PSC brief report item scores