The effect of a conditional cash transfer on HIV incidence in young women in rural South Africa (HPTN 068): a phase 3, randomised controlled trial

This study investigated whether a "conditional cash transfer" could reduce the rate of new HIV infections among young women.

The **intervention** group received a monthly cash payment. This payment was "conditional" on the participant attending at least 80% of school days each month. The exact amount of the cash transfer is not specified in the abstract, but it was designed to be a meaningful incentive in the local context.

The **comparator** group received no cash transfer and continued with their lives as usual, without any specific financial incentive linked to school attendance.

The **primary outcome measure** was HIV incidence, which means the rate at which new HIV infections occurred in each group over the study period. A secondary outcome was the incidence of herpes simplex virus type 2 (HSV-2), though results for this are not detailed in the abstract. The study also explored the relationship between school attendance and HIV risk.

The study enrolled **2,537 young women** and their parents or guardians. Initially, 10,134 young women were screened, but 2,537 met the strict eligibility criteria and were enrolled.

The participants were **girls aged 13-20 years old** at the start of the study. The median age was 15 years, with an interquartile range (IQR) of 14-17 years, meaning half of the girls were between 14 and 17 years old.

All participants were:

Enrolled in school, specifically in grades 8-11.

Not married or pregnant at the time of enrollment.

Able to read.

Residing in the study area (rural Bushbuckridge subdistrict in Mpumalanga province, South Africa) and intending to remain there until the trial was completed.

Both the young woman and her parent or guardian had the necessary documentation to open a bank account, which was required for receiving the cash transfers.

At baseline, 672 (27%) of the enrolled participants reported having had sex at least once.

The study was conducted in a **rural setting in South Africa**, a region with a high burden of HIV.

The researchers used a combination of methods to collect data:

**HIV and HSV-2 testing:** At baseline (the start of the study) and then annually at 12, 24, and 36 months, participants underwent HIV and herpes simplex virus type 2 (HSV-2) testing. These were objective biological tests, likely blood tests, to determine infection status. This is the most direct and reliable way to measure the primary outcome.

**Audio Computer-Assisted Self-Interview (ACASI):** Before and after HIV testing and counseling, participants completed an ACASI. This method involves participants listening to questions through headphones and entering their answers directly into a computer. This approach is often used for sensitive topics like sexual behavior, as it can increase privacy and reduce social desirability bias (where participants might give answers they think are expected rather than truthful ones) compared to face-to-face interviews. The specific questions asked in the ACASI are not detailed in the abstract but would typically cover demographics, sexual history, risk behaviors, and other relevant social and behavioral factors.

**Computer-Assisted Personal Interview (CAPI):** Parents or guardians of the participants completed a CAPI at baseline and each follow-up visit. In a CAPI, an interviewer uses a computer to administer the questionnaire, recording responses directly. This would have gathered information from the parents about household circumstances, school attendance, and potentially other factors relevant to the young women's lives.

**School attendance records:** Although not explicitly stated as a "measurement instrument" in the abstract, the conditionality of the cash transfer (≥80% school attendance) implies that school attendance was rigorously tracked, likely through school records, to determine eligibility for the monthly payments.

This study was a **phase 3, randomised controlled trial (RCT)**, known as HPTN 068. A phase 3 trial is typically a large-scale study designed to confirm the effectiveness of an intervention and monitor side effects, often comparing it to standard treatment or a placebo. In this case, it compared a conditional cash transfer to no cash transfer.

**Study Design:**

**Randomised Controlled Trial (RCT):** This is considered the gold standard for evaluating the effectiveness of interventions. Participants were randomly assigned to either the intervention group (cash transfer) or the control group (no cash transfer). This random assignment is crucial because it helps ensure that, on average, the two groups are similar in all characteristics (known and unknown) at the start of the study, except for the intervention they receive. This minimizes bias and allows researchers to attribute any observed differences in outcomes directly to the intervention.

**Parallel-group design:** Participants remained in their assigned group for the entire duration of the study.

**Randomisation:**

Participants (and their parents/guardians) were assigned in a **1:1 ratio** to either the cash transfer program or the control group. This means an equal number of participants were allocated to each group.

The randomisation process used **numbered sealed envelopes containing a randomisation assignment card**. These envelopes were numerically ordered, and **block randomisation** was employed. Block randomisation ensures that an equal number of participants are assigned to each group within predefined "blocks" of participants. For example, if the block size was 4, then within every 4 participants, 2 would be assigned to the intervention and 2 to the control. This helps maintain balance between the groups throughout the enrollment period, especially if enrollment is stopped early or if there are changes in the study population over time. The use of sealed envelopes helps prevent researchers from knowing the next assignment, reducing selection bias.

**Blinding:**

The abstract does not explicitly state whether participants, researchers, or outcome assessors were blinded. For an intervention like a cash transfer, it is generally **not possible to blind participants** to whether they are receiving money or not. Similarly, the parents/guardians would know if they were receiving the transfer.

However, the **primary outcome (HIV incidence) is an objective biological measure** (a blood test). This means that even if participants and researchers were aware of group assignments, the actual measurement of HIV status is not subject to subjective interpretation or participant bias. This reduces the risk of detection bias for the primary outcome. It is possible that the laboratory personnel conducting the HIV tests were blinded to the participants' group assignments, which would further strengthen the objectivity of the results.

**Duration:**

The study followed participants for a total of **36 months (3 years)**. Data was collected at baseline, and then at annual follow-up visits at 12, 24, and 36 months. This long duration is important for studying HIV incidence, as seroconversion (the time it takes to develop detectable antibodies to HIV after infection) can vary, and behavioral changes take time to manifest in health outcomes.

**Statistical Approach:**

An **intention-to-treat (ITT) analysis** was used for the primary outcome (HIV incidence). In an ITT analysis, all participants are analyzed in the group to which they were originally assigned, regardless of whether they actually received the intervention or adhered to it. For example, if a participant in the cash transfer group stopped attending school and thus didn't receive payments, they would still be analyzed as part of the cash transfer group. This approach reflects real-world effectiveness, as not everyone adheres perfectly to an intervention, and it preserves the benefits of randomisation, preventing bias that could arise from participants dropping out or switching groups.

A **stratified proportional hazards model** was used to compare HIV incidence between the intervention and control groups. This statistical model is appropriate for analyzing time-to-event data (like time to HIV infection) and can account for potential confounding variables by stratifying