Prioritizing high-risk women in HIV hyperepidemic settings is important for optimized introduction and scale of new prevention methods (e.g., oral PrEP). We assessed the performance of a risk score derived from the VOICE cohort using the CAPRISA 004 dataset and modified the score for age-specific prediction.
The initial risk score (IRS) included 7 items: age; HSV-2 serostatus; partner exclusivity; marital/co-habitation; partner’s financial support; alcohol use, and STIs. We developed an age-stratified modified risk score (MRS) and public health use risk score [(PHRS) excludes non-routine HSV serotesting] and evaluated risk prediction according to the area under the curve (AUC), sensitivity, specificity, and un/adjusted hazard ratios (aHR). Given the effect of the tenofovir gel, only participants in the CAPRISA 004 placebo arm (n=444) were used in this analysis.
There were 60 HIV infections over 660.7 person-years of follow-up (incidence=9.08/100 person-years (PY); 95%CI 7.05, 11.7). Women with IRS scores of <8 had a significantly lower HIV incidence of 5.95/100 PY (95%CI 3.88, 9.12) compared to those with scores ≥8 (12.83/100 PY; 95%CI 9.29, 17.7). This cut-off had a sensitivity of 64% and specificity of 57% (Figure 1). The AUC was 0.66. Among the 7 IRS items, only 3 were significantly associated with HIV acquisition in this dataset; age <25 years (aHR=2.47; 95%CI 1.24, 4.89), responding ‘don’t know’ and ‘yes’ that your partner has other partners (aHR=4.02; 95%CI 1.22, 13.3 and 3.77; 95%CI 1.07, 13.3), and being HSV-2 positive (aHR=2.10; 95%CI 1.19, 3.68). The age stratified MRS showed the most predictive score for women <25 years included HSV-2, casual partnerships and partner exclusivity (AUC: 0.70). The PHRS had an AUC of 0.62, and women <25 years with scores <3 vs. ≥3 had incidences of 1.57 (95%CI 0.22, 11.16) and 12.6/ 100 PY (95%CI 9.17, 16.44), respectively.
The ability of the IRS to predict incident HIV infections was comparable to previous validations. However, a cut-off where women were at a substantially low risk of HIV infection could not be determined. Exploratory analyses show the MRS and PHRS to be simpler tools that can be used for PrEP counselling, increasing risk perception and to focus interventions on modifiable risk factors. Yet, relatively lower risk women still had high HIV incidence suggesting further modification to the tool or the need to offer PrEP to any at-risk young woman in hyperepidemic settings.