HIV Risk-Reducing Behavior Interventions for Heterosexuals
HIV Risk-Reducing Behavior Interventions for Heterosexuals
Twenty-two reports (see supplementary web content for references of included reports) containing evaluations of 29 separate couple-based interventions met the selection criteria and were included (see table S1, supplementary web content). Interventions were conducted in the USA (k=15), Africa (k=10), the Caribbean (k=3), Asia (k=2) and Europe (k=1). Two interventions were conducted in multiple countries simultaneously. Participants were most commonly recruited through clinical contact, and two-thirds of those recruited agreed to participate. The average length of time from the last intervention session until first available follow-up was 7.64 months.
Overall, 5168 couples were enrolled in the 29 interventions. Across the participant samples, the mean age was 31 years old, average relationship length was 5.46 years, and 61% were married. Forty percent of individuals were living with HIV, and 18% reported having a concurrent partner. Ten interventions enrolled only serodiscordant couples and nine enrolled seroconcordant and serodiscordant couples. USA samples were, on average, 53% Latino/a, 36% African–American and 20% Caucasian. Nearly all participants (90%) attended all intervention sessions, and 78% of those present at baseline were present at the first available follow-up.
Seven of the 22 included reports simultaneously evaluated two separate couple-based HIV/AIDS prevention interventions. Only six reports incorporated a control group that either delivered intervention content to the female partner only or delivered general health and wellness information to couples. Thirteen interventions reported using a theory to guide intervention development. Interventions consisted of an average of about four sessions for a total of nearly 8 h of contact. Seventeen interventions reported providing condoms, 12 provided VCT, 7 reported addressing concurrent partners, 6 reported addressing gender, 6 reported addressing power, and 4 eroticised safer sex. Thirteen interventions included active condom use skills training (5 of which included practice with partner), and 11 interventions including active communication skills training (5 of which included practice with partner). Nine interventions reported a total of 71 seroconversions between baseline and last follow-up.
Analyses indicated significant increases in condom use with study (or unspecified) partner from baseline to first follow-up (d+=0.94; 95% CI 0.56 to 1.33). After removing one extreme outlier that had an ES more than 3 SDs larger than the mean, the mean ES for condom use remained significant (d+=0.78; 95% CI 0.48 to 1.09; figure 1; Table 1). The hypothesis of homogeneity was rejected, and a large I value indicated large heterogeneity within study effects (I=97.80; 95% CI 97.37 to 98.15). Trim-and-fill results showed no asymmetry for condom use with study partners; no missing studies were estimated. Begg's and Egger's tests were non-significant (Begg's z=1.76, p=0.08, and Egger's bias=4.32, t=1.70, p=0.10), suggesting no publication bias.
(Enlarge Image)
Figure 1.
Effect size estimates for condom use with main partner. Effect sizes appear in temporal order of studies collecting earlier data first. Positive effect sizes indicate increased condom use with study partner at follow-up compared with baseline.
Analysis also indicated significant reductions in reported partner concurrency from baseline to first follow-up (d+=0.38; 95% CI 0.13 to 0.60; figure 2), with heterogeneity present among the ESs (I=95.33; 95% CI 93.46 to 96.66). For reports of concurrent partners, trim-and-fill results estimated six missing studies; Begg's and Egger's tests were both significant (Begg's z=2.84, p<0.01, and Egger's bias=7.59, t=4.11, p<0.01), suggesting publication bias. Yet, the fact that there was heterogeneity implies that more complex models are necessary (whereas, these statistics assume only a single population ES).
(Enlarge Image)
Figure 2.
Effect size estimates for partner concurrency. Effect sizes appear in temporal order of studies collecting earlier data first. Positive effect sizes indicate decreased reports of concurrent partners at follow-up compared with baseline.
ESs were not calculated for condom use with concurrent partners, as only one study reported this outcome. Moderator analyses were conducted for condom use with study partner, but not reported partner concurrency, due to the small number of studies reporting a concurrency outcome and the suggested publication bias with regard to concurrency outcomes. Due to heterogeneity in study outcomes, further analyses used only random-effects assumptions.
Several study dimensions related to ES magnitude (see Table 2). Studies that were conducted toward the beginning of the epidemic and in lower HDI countries showed larger increases in condom use. Interventions that included couples who had been together for a longer period of time, included participants with STIs other than HIV, enrolled serodiscordant couples, provided VCT, had longer follow-up periods, or used face-to-face interviews to collect outcome data, all had larger ESs. When interventions included condom skills training, ES were larger to the extent that condom skills were practiced with study partner, and to the extent that interventions presented content in a variety of contexts. Interventions delivered to individuals or individual couples also showed significant increases in condom use, but interventions delivered only to groups did not. Interventions that delivered intervention content in more sessions tended to have larger ES, but this relationship did not reach statistical significance. No other coded sample characteristics (eg, age, proportion African–American, Latino, married, unemployed, reporting illegal drug use or concurrent partners) or intervention dimensions (eg, theoretical background, condom provision, or inclusion of information about gender, power, culture, or concurrent partners) were significant predictors of ES magnitude.
Results
Description of Studies
Twenty-two reports (see supplementary web content for references of included reports) containing evaluations of 29 separate couple-based interventions met the selection criteria and were included (see table S1, supplementary web content). Interventions were conducted in the USA (k=15), Africa (k=10), the Caribbean (k=3), Asia (k=2) and Europe (k=1). Two interventions were conducted in multiple countries simultaneously. Participants were most commonly recruited through clinical contact, and two-thirds of those recruited agreed to participate. The average length of time from the last intervention session until first available follow-up was 7.64 months.
Overall, 5168 couples were enrolled in the 29 interventions. Across the participant samples, the mean age was 31 years old, average relationship length was 5.46 years, and 61% were married. Forty percent of individuals were living with HIV, and 18% reported having a concurrent partner. Ten interventions enrolled only serodiscordant couples and nine enrolled seroconcordant and serodiscordant couples. USA samples were, on average, 53% Latino/a, 36% African–American and 20% Caucasian. Nearly all participants (90%) attended all intervention sessions, and 78% of those present at baseline were present at the first available follow-up.
Seven of the 22 included reports simultaneously evaluated two separate couple-based HIV/AIDS prevention interventions. Only six reports incorporated a control group that either delivered intervention content to the female partner only or delivered general health and wellness information to couples. Thirteen interventions reported using a theory to guide intervention development. Interventions consisted of an average of about four sessions for a total of nearly 8 h of contact. Seventeen interventions reported providing condoms, 12 provided VCT, 7 reported addressing concurrent partners, 6 reported addressing gender, 6 reported addressing power, and 4 eroticised safer sex. Thirteen interventions included active condom use skills training (5 of which included practice with partner), and 11 interventions including active communication skills training (5 of which included practice with partner). Nine interventions reported a total of 71 seroconversions between baseline and last follow-up.
Overall Efficacy of the Interventions
Analyses indicated significant increases in condom use with study (or unspecified) partner from baseline to first follow-up (d+=0.94; 95% CI 0.56 to 1.33). After removing one extreme outlier that had an ES more than 3 SDs larger than the mean, the mean ES for condom use remained significant (d+=0.78; 95% CI 0.48 to 1.09; figure 1; Table 1). The hypothesis of homogeneity was rejected, and a large I value indicated large heterogeneity within study effects (I=97.80; 95% CI 97.37 to 98.15). Trim-and-fill results showed no asymmetry for condom use with study partners; no missing studies were estimated. Begg's and Egger's tests were non-significant (Begg's z=1.76, p=0.08, and Egger's bias=4.32, t=1.70, p=0.10), suggesting no publication bias.
(Enlarge Image)
Figure 1.
Effect size estimates for condom use with main partner. Effect sizes appear in temporal order of studies collecting earlier data first. Positive effect sizes indicate increased condom use with study partner at follow-up compared with baseline.
Analysis also indicated significant reductions in reported partner concurrency from baseline to first follow-up (d+=0.38; 95% CI 0.13 to 0.60; figure 2), with heterogeneity present among the ESs (I=95.33; 95% CI 93.46 to 96.66). For reports of concurrent partners, trim-and-fill results estimated six missing studies; Begg's and Egger's tests were both significant (Begg's z=2.84, p<0.01, and Egger's bias=7.59, t=4.11, p<0.01), suggesting publication bias. Yet, the fact that there was heterogeneity implies that more complex models are necessary (whereas, these statistics assume only a single population ES).
(Enlarge Image)
Figure 2.
Effect size estimates for partner concurrency. Effect sizes appear in temporal order of studies collecting earlier data first. Positive effect sizes indicate decreased reports of concurrent partners at follow-up compared with baseline.
ESs were not calculated for condom use with concurrent partners, as only one study reported this outcome. Moderator analyses were conducted for condom use with study partner, but not reported partner concurrency, due to the small number of studies reporting a concurrency outcome and the suggested publication bias with regard to concurrency outcomes. Due to heterogeneity in study outcomes, further analyses used only random-effects assumptions.
Condom Use With Study Partner
Several study dimensions related to ES magnitude (see Table 2). Studies that were conducted toward the beginning of the epidemic and in lower HDI countries showed larger increases in condom use. Interventions that included couples who had been together for a longer period of time, included participants with STIs other than HIV, enrolled serodiscordant couples, provided VCT, had longer follow-up periods, or used face-to-face interviews to collect outcome data, all had larger ESs. When interventions included condom skills training, ES were larger to the extent that condom skills were practiced with study partner, and to the extent that interventions presented content in a variety of contexts. Interventions delivered to individuals or individual couples also showed significant increases in condom use, but interventions delivered only to groups did not. Interventions that delivered intervention content in more sessions tended to have larger ES, but this relationship did not reach statistical significance. No other coded sample characteristics (eg, age, proportion African–American, Latino, married, unemployed, reporting illegal drug use or concurrent partners) or intervention dimensions (eg, theoretical background, condom provision, or inclusion of information about gender, power, culture, or concurrent partners) were significant predictors of ES magnitude.