Table 2.

The experimental evidence on interventions to increase HIV-testing uptake, antiretroviral therapy linkage and uptake, retention, and adherence in sub-Saharan Africa

Authors	Country (city or region)	Intervention	Unit of randomization	Sample size	Length of follow-up (months)	Effective	Results
HIV-testing uptake
Becker et al. [20]	Tanzania (Dar es Salaam)	Written and verbal invitation to ANC attendees to bring their male partner to the clinic for VCT for couples.	ANC attendee	1521	NR	Yes (negatively)	Women who received an invitation to bring their male partner to the clinic for couple VCT were significantly less likely to receive an HIV test than women who received standard individual VCT (39 vs. 71%, P<0.001).
Burnett et al. [21]	Swaziland (Manzini)	School-based HIV education programme 'It's our future tool' (13 weekly half-day sessions).	Individual student	312	NR	Yes	In the intervention group, the proportion of students who reported that they 'ever had HIV test' increased from 11 to 42% (P<0.001), whereas in the control group it increased from 5 to 9% (P=0.13).
Corbett et al. [22]	Zimbabwe (Harare)	On-site rapid testing at an occupational health clinic.	Business occupational health clinic	7482	2 years	Yes	HIV-testing uptake was significantly higher in the intervention group compared with the control group, which were offered vouchers for off-site HIV testing (RR=2.7; 95%, CI=1.8–3.9).
Ditekemena et al. [23]	Democratic Republic of Congo (Kinshasa)	Offer of VCT to ANC attendees' male partners at neighbourhood health centre vs. bar vs. church.	ANC attendee	2706	5 months	Yes	Male HIV-testing uptake higher in bars (26%, P<0.001) and churches (21%, P=0.163) compared with neighbourhood health centres (18%).
Fylkesnes and Siziya [24]	Zambia	VCT offered at an optional location (including home).	Participants in a population-based HIV survey	1445	NR	Yes	HIV-testing uptake was 4.7 times higher in the intervention group, which was offered VCT in an optional location (including homes if desired) compared with the control group, which was offered VCT in the local health clinic (56 vs. 12%).
Lugada et al. [25■]	Uganda (Jinja, Kamuli, Iganga, Mayuge, Mukono)	Home-based VCT for household members of ART patients.	ART patient	7184	NR	Yes	Household members in the intervention group were significantly more likely to receive VCT than household members in the control group who were offered clinic-based VCT (56 vs. 11%, P<0.001).
Mohlala et al. [26]	South Africa (Cape Town)	Written and verbal invitation to the male partner of ANC attendees to attend a VCT session.	ANC attendee	500	1 week	Yes	Women who received a written invitation to bring their male partner to the clinic for VCT were significantly more likely to do so than women who received standard pregnancy information without a written partner invitation (RR=1.36, P=0.002).
Sweat et al. [27]	Tanzania, Zimbabwe, Thailand	Community-based VCT.	Community	12 983 (in Tanzania) and 22 850 (in Zimbabwe)	3 years (in Tanzania) and 3.5 years (in Zimbabwe)	Yes	In the communities receiving community-based VCT, VCT clients were substantially more likely to receive their first HIV test than VCT clients in the control communities, which offered only standard clinic-based VCT (37 vs. 9% in Tanzania, and 51 vs. 5% in Zimbabwe).
Wanyenze et al. [28]	Uganda (Kampala)	In-patient HCT.	Hospital inpatient	500	6 months	Yes	Patients who were offered in-patient HCT were substantially more likely to receive HCT than patients in the control group, who were offered to return to the hospital for HCT 1 week after discharge (99 vs. 69%).
ART linkage and uptake
Wanyenze et al. [28]	Uganda (Kampala)	In-patient HCT.	Hospital in-patient	500	6 months	Yes	Patients who were offered in-patient HCT and found to be HIV-infected were significantly more likely to attend a HIV clinic than the patients in the control group who were offered to return for out-patient HCT after discharge and subsequently found to be HIV-infected (74 vs. 56%, P=0.032).
Zwarenstein et al. [29]	South Africa	On-site educational outreach programme in health clinics.	Health clinic	10136	23 weeks (intervention group) and 28 weeks (control group)	No	Enrolment in ART programme through new HIV testing was not significantly different in the intervention compared with the control group (53 vs. 50%, P=0.695).
ART retention
Sanne et al. [30■]	South Africa (Cape Town and Johannesburg)	Nurse vs. doctor-managed ART (a noninferiority trial).	ART patient	812	120 weeks	Yes	Patients who received nurse-managed ART were not significantly different in their ART retention than patients who received doctor-managed ART (HR=1.13, 95%CI=0.81–1.59).
ART adherence
Busari et al. [31]	Nigeria (Abuja)	Structured adherence education programme.	ART patient	420	8	Yes	99% mean adherence rate in intervention group vs. 88% in control group (P<0.001); 0.51 OI per patient per month in intervention group vs. 1.31 OI in control group (P=0.002); mortality significantly lower in intervention vs. control group (P=0.008); rate of hospital admission significantly lower in intervention vs. control group.
Busari et al. [32]	Nigeria (Abuja)	Structured adherence education programme.	ART patient	620	8	Yes	Patients in the intervention group had significantly higher mean adherence than patients in the control group (99 vs. 88%, P<0.001). They also had significantly higher CD4 cell count (238 vs. 141 cells/μl, P<0.001) and a significantly lower rate of hospital admissions.
Chang et al. [33■]	Uganda (Rakai)	Treatment supporter.	ART patient	1336	27	Yes	Patients in the intervention group had significantly lower risk of virologic failure than patients in the control group at 96 weeks (RR 0.50, 95% CI 0.31–0.81), 120 weeks (RR 0.59, 95% CI 0.22–1.60), 144 weeks (RR 0.39, 95% CI 0.16–0.95), 168 weeks (RR 0.30, 95% CI 0.097–0.92), and 192 weeks (RR 0.067, 95% CI 0.0065–0.71). There were no significant effects on risk of virologic failure in earlier periods, pill count, self-reported adherence, or median CD4 cell count.
Chung et al. [34]	Kenya (Nairobi)	Educational adherence counselling vs. alarm device.	ART patient	400	18	Yes	Patients who received counselling were 29% less likely to have monthly adherence (assessed by pill count) <80% (hazard ratio (HR)=0.71, P=0.055) and 59% less likely to experience viral failure (HR=0.41, P=0.01) compared with those who received no counselling. The alarm device did not have a significant effect on poor adherence (HR=0.93, P=0.7) or viral failure (HR=0.99, P=1.0).
Lester et al. [35■]	Kenya (Nairobi)	Mobile-phone text messages (weekly text message asking 'How are you?', with follow-up calls in patients who said that they had a problem or who failed to respond within 48 h).	ART patient	538	12	Yes	Patients in the intervention group were significantly less likely to report nonadherence than patients in the control group (RR=0.81, P=0.006) and were significantly less likely to experience virologic failure (RR=0.85, P=0.04).
Mugusi et al. [36]	Tanzania (Dar es Salaam)	Treatment supporter (first intervention group), calendar, adherence education (second intervention group).	ART patient	621	12	No	There were no significant differences between the three groups in any of three assessed adherence outcomes.
Nachega et al. [37]	South Africa (Cape Town)	DOT.	ART patient	222	24	No	Patients in the intervention group had l arger median CD4 cell count increases than patients in the control group at 6 months (148 cells/μl vs. 111 cells/μl, P=0.02), but the increases were not significantly different at any other observation time-points. There were no significant effects of the intervention on viral suppression or adherence assessed by pill count.
Nachega et al. [38■]	South Africa (Cape Town)	DOT and educational adherence counselling.	ART patient	274	24	No	There were no significant differences in self-reported adherence, CD4 cell count, or viral load.
Ndekha et al. [39]	Malawi (Blantyre)	Lipid paste vs. flour supplement.	ART patient	336	9	(No)	There were no significant differences in self-reported adherence, CD4 cell count, or viral load.
Ndekha et al. [40]	Malawi (Blantyre)	Lipid paste vs. flour supplement.	ART patient	491	3	No	There were no significant differences in self-reported adherence, CD4 cell count, or viral load.
Pearson et al. [41]	South Africa (Western Cape)	Treatment supporter, DOT, and adherence education.	ART patient	350	12	Yes	Patients in the intervention group had significantly higher mean medication adherence than patients in the control group at 6 months (93 vs. 85%) and at 12 months (94 vs. 88%). There was no significant difference in mean CD4 change.
Pop-Eleches et al. [42■]	Kenya	Mobile-phone text messages in four intervention groups (short reminders every day, short reminders every week, long reminders every day, or long reminders every week).	ART patient	431	12	Yes	Patients in the groups receiving weekly reminders were significantly more likely to adhere than patients in the control group at 48 weeks (53 vs. 40%, P=0.03) at 48 weeks. Treatment interruptions were significantly less likely in the groups receiving weekly reminders at 48 weeks (81 vs. 90%, P=0.03).
Sarna et al. [43■]	Kenya (Mombasa)	DOT.	ART patient	234	18	Yes	According to pill count, patients in the DOT group were five times more likely to adhere to ART than patients in the control group (P<0.001) during the interventions, after adjustment for depression and HIV-related admission to hospital (P<0.001). There were no significant differences in adherence after the intervention, nor were there significant differences in CD4 cell count or viral load suppression. BMI increase was significantly greater in the DOT vs. the control group during the intervention (P=0.014).
Taiwo et al. [44]	Nigeria (Jos)	Treatment supporter and DOT.	ART patient	499	12	Yes	Patients in the intervention group had significantly higher odds of adherence than patients in the control group at 24 weeks (OR 3.06, P<0.01) and 48 weeks (OR 1.95, P<0.01). They were also significantly more likely to have undetectable viral load at week 24 (62 vs. 50%, P<0.05). There were no significant differences in either viral load suppression at week 48 or CD4 cell count increases at weeks 24 and 48.
van Loggernberg et al. [45]	South Africa (Durban)	Individual vs. group educational adherence counselling.	ART patient	297	9	No	There was no significant difference in viral load suppression between the intervention and the control group (demonstrating noninferiority of the group educational counselling in comparison to the individual educational counselling).

ART, antiretroviral therapy; CI, confidence interval; DOT, directly observed therapy; HCT, HIV counselling and testing; OI, opportunistic infections; RR, relative risk; VCT, voluntary counselling and testing.