Table 1.

Summary of evidence on intrauterine device use and risk of HIV acquisition in women

Author, year, location, funding	Study design, purpose, period of data collection	Population	Number seroconverted/number analysed, number of seroconverters by exposure group, overall HIV incidence	Exposure	Comparison	Results (point estimate, CI)	Strengths	Weaknesses	Quality
Saracco 1993,11 Italy Ministry of Health and National Research Council of Italy	Cohort; determine incidence and risk factors for male to female sexual HIV transmission, 1987–1991	343 seronegative women with HIV-infected male partners, recruited from 16 hospital and outpatient clinic locations; 529.6 person-years follow-up	19/343 after mean 14.1 months 1/1 IUD and no condom use seroconverted 0/1 IUD and condom use seroconverted 0/22 OC users seroconverted (11 reported always using condom) 3.6 per 100 person-years	IUD (type not specified)	No clear comparison group; seroconversions reported for OC users	No comparative estimates given	Serodiscordant couples Study interval 6 months	No comparative estimates provided and no clear comparison group No adjustment for condom use No time-varying analyses Follow-up rates not reported or unclear	Unlikely to inform the primary question
Sinei 1996,12 Kenya US NIH and USAID, WHO, Ortho Pharmaceutical Corp.	Cohort; pilot study to examine relationship between contraceptive methods and HIV transmission, 1990–1992	1537 seronegative women recruited from a family planning clinic	16 seroconversions by 12 months: IUD users: 7 DMPA users: 2 OC users: 5 12-month cumulative incidence rates (95% CI) per 100: Overall: 2.1 (95% CI 1.1 to 3.2): IUD: 1.66 (95% CI 0.4 to 2.9) DMPA: 1.92 (95% CI 0.0 to 4.6) OC: 4.45 (95% CI 0.5 to 8.4)	IUD (type not specified)	No clear comparison group, but separate seroconversion rates provided for DMPA and OC users	No comparative estimates given	Study interval of 3 months	No comparative estimates provided and no clear comparison group No adjustment for condom use No time-varying analyses Low follow-up rates (29% overall)	Unlikely to inform the primary question
Kapiga 1998,13 Tanzania Rockefeller Foundation, US NIH	Cohort, to study the incidence of HIV associated with contraceptive methods, 1992–1995	1370 (of an original cohort of 2471) women recruited from family planning clinics; 2236 person-years follow-up	75/1370 seroconverted 3.4 per 100 person-years (95% CI 2.6 to 4.1) IUD: 8/162	IUD (copper T loop specified); ever-use during the follow-up period	No IUD use (included those using hormonal methods, non-hormonal methods or no method)	IUD use vs no use: adjRR 0.80 (95% CI 0.38 to 1.69) Duration of IUD use 1–12 months vs 0 months: adjRR 1.22 (95% CI 0.44 to 3.37) >13 months vs 0 months: adjRR 0.59 (95% CI 0.21 to 1.66)		No adjustment for condom use in the main model; authors state in text that results were not materially altered when excluding condom users or adjusting for condom use No time-varying analysis of IUD exposure (exposure measured as ever-use during the follow-up period) Comparison group included all non-IUD users Study interval unclear; HIV testing only occurred at baseline and final interview 44.5% loss to follow-up	Unlikely to inform the primary question
Lavreys 2004,14 Kenya US NIH	Cohort, to assess the association between contraceptive use and HIV acquisition, 1993–2003	1498 seronegative women, recruited from a clinic for sex workers, Mombasa, 1272 returned for follow-up; providing 2931 person-years follow-up; 15 428 follow-up visits	248/1272 Number seroconverted by group not reported	IUD (type not specified)	No contraceptive method or tubal ligation	adjHR 1.1 (95% CI 0.4 to 3.0)	Adjusted for condom use Time-varying analysis of contraception exposure and confounders Clearly defined comparison group Study interval 1 month	Follow-up rates not reported or unclear Potential for residual confounding	Informative but with important limitations
Hofmeyr 2017,15 South Africa Effective Care Research Unit, East London, South Africa	RCT; to compare rates of incident HIV among women using progestogen-only injectables and those using Cu-IUDs; 2009–2012	2493 enrolled, 1246 in injectable arm and 1247 in IUD arm, recruited from women attending pregnancy termination services at two hospitals in South Africa	ITT analysis 20/656 (3%) injectable arm 22/634 (3.5%) IUD arm Median follow-up time: 19–20 months	Cu-IUD	DMPA and NET-EN together and separately	ITT analysis Injectables vs IUD RR 0.88 (95% CI 0.48 to 1.59) Separate ITT results for different injectables not given Per protocol analysis Injectables vs IUD RR 0.94 (95% CI 0.52 to 1.71) DMPA vs IUD RR 1.01 (95% CI 0.55 to 1.86) NET-EN vs IUD RR 0.58 (95% CI 0.14 to 2.42)	Appropriate randomisation procedures, with good concealment of allocations at point of assignment Clearly defined comparison group	No measurement or adjustment for condom use (or other potential confounders) No information on discontinuation or switching, or time-varying analysis of contraceptive use 19–20 month median interval for HIV testing 37% of follow-up HIV test results by self-report	Unlikely to inform the primary question
Palanee-Phillips 2019,9 South Africa NIH	Cohort; from RCT that examined effectiveness of dapivirine ring to prevent HIV; 2012–2015	1136 HIV-seronegative women from South African clinical study sites; 1771 person-years follow-up	95/1136 seroconverted Median follow-up: 1.6 years HIV incidence/100 woman-years: Overall: 5.6 DMPA: 5.79 NET-EN: 6.22 Implant: 1.93 Cu-IUD: 4.48	Cu-IUD	DMPA NET-EN Implant	Cu-IUD vs DMPA:adjHR 1.10 (95% CI 0.53 to 2.27)* Cu-IUD vs NET-EN: adjHR 0.98 (95% CI 0.47 to 2.04)* Cu-IUD vs implant: adjHR 2.17 (95% CI 0.59 to 7.69)* Cu-IUD vs all three hormonal methods together: adjHR 1.11 (95% CI 0.57 to 2.22)*	Adjusted for condom use. Time-varying analysis of contraception exposure, condom use, and other relevant confounders Clearly defined comparison groups Monthly intersurvey interval	No information on follow-up time or attrition by study group Potential for residual/ unmeasured confounding	Informative but with important limitations
Evidence for Contraceptive Options and HIV Outcomes (ECHO) Trial Consortium, 2019,1 12 sites in Eswatini, Kenya, South Africa, Zambia Bill & Melinda Gates Foundation, USAID and PEPFAR, Swedish International Development Cooperation Agency, South African Medical Research Council, UNFPA, Government of South Africa	RCT, 2015–2019	7830 HIV-seronegative women seeking contraception (7829 randomised)	397/7715 seroconverted HIV incidence/100 woman-years Overall: 3.81 DMPA-IM: 4.19 LNG implant: 3.31 Cu-IUD: 3.94 Follow-up time: up to 18 months	Cu-IUD	DMPA-IM LNG-implant	ITT analysis, Cu-IUD vs DMPA-IM: HR 0.96 (96% CI 0.75 to 1.22)* Cu-IUD vs LNG implant: HR1.18 (96% CI 0.91 to 1.53) Continuous use analysis, Cu-IUD vs DMPA-IM: adjHR 0.91 (96% CI 0.69 to 1.19)* Cu-IUD vs LNG implant: adjHR 1.18 (96% CI 0.90 to 1.55) No effect modification by age or HSV-2 status, and no substantial differences by several other factors	Publication of full trial protocol16 Clear randomisation and allocation procedures Good adherence to allocated treatments (99% uptake; continuation 92% of follow-up time), with limited discontinuation or change of contraceptive method	Patients and clinicians not blinded; however, study team made concerted efforts to not provide different information/counselling to women in DMPA-IM group vs other groups	Informative with few limitations

*Estimates are for the inverse associations of those reported by the authors.

adjHR, adjusted hazard ratio; adjRR, adjusted relative risk; CI, confidence interval; Cu, copper; DMPA, depot medroxyprogesterone acetate; HIV, human immunodeficiency virus;HSV, herpes simplex virus; IM, intramuscular; ITT, intention-to-treat;IUD, intrauterine device; LNG, levonorgestrel; NET-EN, norethisterone ethanate; OC, oral contraceptive;PEPFAR, President's Emergency Plan for AIDS Relief; RCT, randomised controlled trial; UNFPA, United Nations Population Fund; USAID, United States Agency for International Development; US NIH, United States National Institutes of Health; WHO, World Health Organization.