The use of antiretroviral drugs to prevent HIV infection through postexposure prophylaxis (PEP) has been recommended since zidovudine was first found to be associated with an 81% reduction in the odds of HIV infection following percutaneous exposure among healthcare workers in a case-control study over two decades ago . PEP is usually prescribed as a 28-day course of antiretroviral drugs that should be offered and initiated as early as possible to all individuals with an exposure that has the potential for HIV transmission, and ideally within 72 h. The efficacy and use of PEP is supported by data from animal studies  and trials demonstrating reduction of maternal–infant transmission , and indications for PEP have expanded over time to cover nonoccupational exposure, including accidental exposure, noncoercive sex, and sexual assault.
The prescription of PEP is far from uniform, and can differ by drug number (two drugs or more), drug class, and frequency (starter packs or full prescriptions), and according to different guidelines prescribing practices may vary according to the suspected level of risk. The extent to which these different approaches may influence completion rates is unclear.
Previous systematic reviews of adherence to PEP have highlighted poor rates of PEP completion for victims of sexual assault , with higher completion rates reported following nonoccupational nonforcible sexual exposure ; a comprehensive assessment of PEP completion rates across the cascade of care and across all populations and exposures has not yet been published. In order to inform of the development of WHO guidelines for the use of PEP irrespective of exposure type, and identify potential influences of poor outcomes, we undertook a systematic review of PEP prescription and completion rates across all populations at each step in the PEP cascade of care.
Randomized and nonrandomized studies reporting completion rates for PEP were included regardless of exposure type, age, or geographical location. No language restrictions were applied. Animal studies were excluded as were studies that included 10 or less patients offered PEP (initiations could be fewer), provided PEP as monotherapy, or described outcomes for preexposure prophylaxis or the prevention of mother-to-child transmission and related infant prophylaxis.
An initial search strategy was developed combining terms for postexposure prophylaxis and HIV. The results of this search were cross-checked against articles included in prior PEP reviews [4–6], and any missing studies were reviewed to identify additional terms to refine the search strategy. We then searched Medline via PubMed, EMBASE, the Cochrane Database of Systematic Reviews, and Lilacs from inception to 1 December 2013; this search was updated in PubMed on 01 June 2014. Conference abstracts of all Conferences of the International AIDS Society were searched from 2010 to 2013, and the Conference on Retroviruses and Opportunistic Infections for 2014 (past conference sites were not available online) to identify recent studies that may not yet have been published as full text.
Using a predefined protocol (Supplementary Web Appendix, http://links.lww.com/QAD/A598), two investigators (N.F., C.I.), working independently scanned all abstracts and independently assessed potentially eligible studies as full text. Consensus was sought prior to final inclusion and in case of disagreement a third investigator (Z.S.) was consulted.
Data were extracted independently by at least two reviewers (N.F., Z.S., and C.I.) using a piloted data extraction tool. Data abstractors collected information about the study country, study populations (age sex and key population status), sample size, exposure type, and regimens used. Where studies reported multiple exposure type, population, or regimen, studies were classified according to the group to which the majority (≥50%) of patients belonged; if this was not possible, studies were designated as ‘mixed’ cohorts for the covariate of interest. The following outcome data were extracted to ascertain where losses occurred along the cascade of care from initial eligibility up to post-PEP follow-up: number initiating PEP among those eligible, number completing 28 days treatment, and number attending follow-up. Where available, reasons for noncompletion of PEP were recorded. Study quality was assessed according to predefined criteria indicative of risk of bias criteria (see protocol in Supplementary Web Appendix, http://links.lww.com/QAD/A598). For randomized trials, risk of bias indicators were as follows: no/unclear sequence generation, allocation concealment, and blinding, selective patient recruitment or outcome reporting, results reported as intent-to-treat, and baseline group differences (imbalanced or not adjusted for at analysis); for observational cohorts the following items were considered indicative of risk of bias: retrospective study design, clear inclusion and exclusion criteria, refusal rate more than 20%, and selective patient recruitment and outcome reporting.
Descriptive statistics were used to compare the number of studies published over time and proportions compared using the χ2 test and, in the case of small numbers, the Fisher's exact test. Point estimates and 95% CI were calculated for the proportion of individuals reaching each step in the cascade of care from eligibility determination to attendance of follow-up visit (up to 6 months post 28-day treatment course for HIV test). In instances where PEP was discontinued because it was subsequently found not to be needed (either because baseline HIV test was positive or because the source was confirmed to be HIV negative), these cases were excluded from the denominator for assessing PEP completion rates and these outcomes are reported separately. Data were pooled using the DerSimonian and Laird random effects method , with proportions transformed prior to pooling using the Freeman–Tukey double arcsine transformation  and then back transformed to the original scale . We calculated the τ2 statistic using DerSimonian and Laird method-of-moments estimator to assess between-study heterogeneity . Preplanned subgroup analyses were undertaken to determine the potential influence of the following covariates: exposure type, age, key population status, regimen, study design, level of economic development (as defined by the World Bank http://data.worldbank.org/about/country-classifications), and with differences assessed through interaction test for subgroup effects. All P values are two sided and P < 0.05 was considered significant. All analyses were conducted using Stata version 12.0 (StataCorp, College Station, Texas, USA).
From the initial screen of 3184 titles and 75 conference abstracts, 94 full-text articles and three conference abstracts were included in the final review (Fig. 1), reporting outcomes on 21 462 PEP initiations; one article reported data from two separate studies . Study size ranged from eight to 2278 PEP initiations (median = 112) and studies were published between 1998 and 2014 (median = 2007). The majority of studies (68 studies, 17 293 PEP initiations) were conducted in high-income countries. There were five randomized trials, 45 prospective cohorts, and 47 retrospective cohorts. Nonoccupational exposure to HIV was the main reason for PEP in 34 studies (n = 11 840), occupational exposure for 22 studies (n = 3058), sexual assault for 26 studies (n = 3093), and in the remainder of studies (15 studies, n = 3471) a mix of different exposures were reported. Fifteen studies reported that PEP was provided multiple times, most frequently to MSM (13 studies), but information on duration between prescriptions was not consistently reported.
Study quality and reporting overall was determined to be low to moderate. Among the observational studies, the majority (88%) reported inclusion and exclusion criteria, and there was no evidence of bias with respect to patient selection (83%) or outcome reporting (87%); however, almost half (49%) were retrospective cohorts, and only 48% reported adverse drug event data disaggregated by drug. Among the trials, all reported inclusion criteria and there was no evidence of bias with respect to patient selection and outcome reporting; however, none of the trials reported allocation concealment, and only one reported results as intent-to-treat.
Characteristics of included studies are summarized in Table 1, and individual studies are detailed in Supplementary Web Table 1, http://links.lww.com/QAD/A598.
PEP regimens varied considerably in terms of type of antiretroviral drugs, number of drugs prescribed, and the use of starter packs. Two-drug regimens were prescribed in 33 studies (7053 prescriptions) and three-drug regimens in 49 studies (10 348 prescriptions). In the majority of cases, the third drug provided was a protease inhibitor, although in four studies, the third drug was an integrase inhibitor or a nonnucleoside reverse transcriptase inhibitor. There was no clear relationship between prescribing practices (i.e. number of drugs or use of starter pack) and exposure type (Supplementary Web Table 1, http://links.lww.com/QAD/A598).
Proportion completing postexposure prophylaxis
Overall, the proportion of people considered eligible for PEP who were reported to have completed the full 28-day course was 56.6% (95% CI 50.9–62.2%; τ2 0.25). Completion rates were highest in studies reporting PEP use in nonoccupational exposures (65.6%, 95% CI 55.6–75.6%) or a mix of occupational and nonoccupational exposures (62.6%, 95%CI 50.1–75.2%) and lowest for sexual assault (40.2%, 95% CI 31.2–49.2%). Higher rates of PEP completion were reported by MSM (67.2%, 95% CI 59.5–74.9%). By age, completion rates appeared to be lower for adolescents (36.6%, 95% CI 4.0–69.2%) compared with adults (59.1%, 95% CI 53.9–64.2%) or children (64.0%, 95% CI 41.2–86.8%). There was no observable difference in completion rates comparing two-drug or three-drug regimens, or protease inhibitor or nonprotease inhibitor regimens. Completion rates by subgroup are summarized in Fig. 2.
Postexposure prophylaxis discontinuations
Sixty studies (64%) reported that 1033 people discontinued PEP because of an adverse event; there were fewer such discontinuations for two-drug regimens (1.9%, 95% CI 0.7–3.1%) compared with three-drug regimens (9.1%, 95% CI 7.1–11.0%). Forty-three studies reported that 815 people (8.7%, 95% CI 7.4–9.9%) discontinued PEP because the source case was subsequently considered to be low or no risk of transmitting HIV; such discontinuations were highest for occupational exposure (21.9%, 15.3–28.6%) and lowest for sexual assault (0.1%, 95% CI 0–0.3%). Twenty-five studies reported that 87 people (0.7%, 95% CI 0.3–1.1%) discontinued PEP because they were subsequently found to be HIV positive. When starter packs were provided (37 studies, n = 5997), 28.0% (95% CI 21.4–34.5) failed to return to pick up the rest of their PEP course.
Among all individuals considered eligible for PEP (30 studies reporting data on 6028 participants), 14% of people refused PEP (95% CI 12.0–16%) with no clear differences according to exposure type. Finally, of those who completed PEP, 30.6% (95% CI 23.1–38.1%) failed to attended a follow-up visit. Figure 3 summarizes losses along the PEP cascade of care.
Among all people who were considered eligible for PEP (i.e. excluding people who were reported to already be HIV positive or where the source exposure was established to be HIV negative; n = 18 807), there were 37 seroconversions among 8007 participants completing PEP. The majority of these seroconversions were reported to have occurred among nonoccupational exposures who had ongoing high-risk behaviors following PEP completion; only three seroconversions were ascribed as PEP failures. Seroconversions are described in Supplementary Web Table 2, http://links.lww.com/QAD/A598.
Interventions to support adherence
Three studies assessed enhanced counseling interventions to support adherence. A randomized trial from South Africa assessed telephonic psychosocial support, a leaflet and an adherence diary to support victims of sexual assault ; the two other randomized trials, conducted in France  and the United States , assessed counseling interventions to support nonoccupational-exposed cases. All three studies showed a tendency towards improved adherence to PEP, with an overall pooled odds ratio for improved adherence of 1.4 (95% CI 0.9–2.3%).
Despite over two decades of use, the findings of this review suggest that adherence remains a challenge across populations, settings, and exposure types. There is a need to improve retention in care along the cascade of care from PEP eligibility assessment, to drug prescription, and adherence to a full-course regimen and acceptance to post PEP follow-up.
A fuller understanding is needed of reasons for refusal and noncompletion of PEP across different populations and exposures. Qualitative surveys have identified some potential areas for programme improvements, for example, provision of counseling, and prescribing of all PEP medication at the first visit to the health clinic is preferred by victims of sexual assault in South Africa . This and other studies  point to the need to simplify PEP access, prescribing practices and provide adherence support in order to improve uptake and completion rates.
Recent guidelines issued by the United States Centers for Disease Control in 2013 recommend a minimum of three drugs for PEP . Although this review suggested lower rates of PEP discontinuation with two-drug regimens compared with three-drug regimens, completion rates overall are similar. The relative efficacy of two-drug and three-drug regimens for PEP is challenging to determine in terms of ethics and study design (sample size required to show a difference) and in the absence of such evidence physicians are likely to continue to prefer three-drug regimens, particular for high-risk exposures. The identification of a single PEP regimen that is simple, safe, well tolerated, and cost-effective would simplify prescribing practices and may improve acceptance and completion rates.
Strengths of this review include an inclusive approach that allowed for the identification of outcomes for over 21 000 PEP initiations across a range of exposures. In anticipation of large heterogeneity across studies, we used random effects models to pool proportions and undertook a priori subgroup analyses to explain differences. Publication bias is inevitable for reviews of outcomes from routine programme settings as not all programmes publish outcomes. The relatively poor outcomes reported overall suggest that bias towards the publication of positive results is not a major concern for this review. The main limitation of this review is the quality of the studies, in particular the inconsistency in reporting of important study variables and outcomes such as adverse events and reasons for discontinuation. Seroconversion rates need to be read with particular caution in light of the poor follow-up rates.
Notwithstanding these limitations, our review points to several considerations for future PEP research, policy, and practice. First, high-quality studies are needed to identify interventions to improve adherence to PEP, in particular for vulnerable groups such as adolescents (who are increasingly recognized as a priority group for target support across all aspects of HIV programming)  and victims of sexual assault. There are substantial ethical and methodological challenges in prospectively evaluating interventions for PEP, and there is a need to standardize the reporting of data from routine programmes, which could be assisted through the establishment of an international PEP registry. Although further studies are needed to evaluate adherence support interventions specifically for PEP, promising interventions may be found in the HIV treatment adherence literature, including treatment supporters, mobile phone messaging, and the use of fixed-dose combinations [19,20]. Second, the use of starter packs deserves more careful consideration, considering the high rates of defaulting between first and second prescription [21,22]. Third, the choice of antiretroviral drugs for PEP needs to consider the need to maximize completion rates. Adverse drug reactions , pill burden, drug–drug interactions, cost and daily dosing are all important drivers of nonadherence to antiretrovirals and need to be carefully balanced .
In conclusion, the potential for antiretroviral drugs to prevent HIV infection has received renewed attention in recent years, both through the provision of antiretrovirals to high-risk HIV-negative individuals prior to exposure to HIV to prevent acquisition  and the immediate treatment of HIV-positive individuals to reduce the risk of onward transmission . This review highlights the need to also include PEP as part of the broader agenda when considering how to improve antiretroviral delivery to prevent new HIV infections.
N.F., C.I., and Z.S. designed the study and acquired and analysed the data. All authors assisted in the interpretation of the data and the drafting of the manuscript, provided critical revision, and approved the final manuscript.
Source of funding: This work was in part supported by funds from the Bill and Melinda Gates Foundation. The funders had no role in the design and conduct of the study, collection, analysis, and interpretation of the data, or the preparation, review, or approval of the manuscript.
Conflicts of interest
There are no conflicts of interest.
1. Centers for Disease C, Prevention. Case-control study of HIV seroconversion in health-care workers after percutaneous exposure to HIV-infected blood-France, United Kingdom, and United States, January 1988-August 1994
. MMWR Morb Mort Wkly Rep
2. Black RJ. Animal studies of prophylaxis
. Am J Med
1997; 102 (5B):39–44.
3. Connor EM, Sperling RS, Gelber R, Kiselev P, Scott G, O'Sullivan MJ, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. Pediatric AIDS Clinical Trials Group Protocol 076 Study Group
. N Engl J Med
4. Chacko L, Ford N, Sbaiti M, Siddiqui R. Adherence to HIV postexposure prophylaxis in victims of sexual assault: a systematic review and meta-analysis
. Sex Transm Infect
5. Oldenburg CE, Barnighausen T, Harling G, Mimiaga MJ, Mayer KH. Adherence to post-exposure prophylaxis for nonforcible sexual exposure to HIV: a systematic review and meta-analysis
. AIDS Behav
6. Young TN, Arens FJ, Kennedy GE, Laurie JW, Rutherford G. Antiretroviral postexposure prophylaxis (PEP) for occupational HIV exposure
. Cochrane Database Syst Rev
7. DerSimonian R, Laird N. Meta-analysis in clinical trials
. Control Clin Trials
8. Freeman MFTJ. Transformations Related to the Angular and the Square Root
. Ann Math Stat
9. Miller J. The inverse of the Freeman-Tukey double arcsine transformation
. Am Stat
10. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses
11. Olshen E, Hsu K, Woods ER, Harper M, Harnisch B, Samples CL. Use of human immunodeficiency virus postexposure prophylaxis in adolescent sexual assault victims
. Arch Pediatr Adoles Med
12. Abrahams N, Jewkes R, Lombard C, Mathews S, Campbell J, Meel B. Impact of telephonic psycho-social support on adherence to postexposure prophylaxis (PEP) after rape
. AIDS Care
13. Bentz L, Enel P, Dunais B, Durant J, Poizot-Martin I, Tourette-Turgis C, et al. Evaluating counseling outcome on adherence to prophylaxis and follow-up after sexual HIV-risk exposure: a randomized controlled trial
. AIDS Care
14. Roland ME, Neilands TB, Krone MR, Coates TJ, Franses K, Chesney MA, et al. A randomized noninferiority trial of standard versus enhanced risk reduction and adherence counseling for individuals receiving postexposure prophylaxis following sexual exposures to HIV
. Clin Infect Dis
15. Christofides NJ, Muirhead D, Jewkes RK, Penn-Kekana L, Conco DN. Women's experiences of and preferences for services after rape in South Africa: interview study
16. Ko NY, Yeh SH, Tsay SL, Ma HJ, Chen CH, Pan SM, et al. Intention to comply with postexposure management among nurses exposed to blood and body fluids in Taiwan: application of the theory of planned behaviour
. J Hosp Infect
17. Kuhar DT, Henderson DK, Struble KA, Heneine W, Thomas V, Cheever LW, et al. Updated US Public Health Service guidelines for the management of occupational exposures to human immunodeficiency virus and recommendations for postexposure prophylaxis
. Infect Control Hosp Epidemiol
18. Lowenthal E, Bakeera-Kitaka S, Marukutira T, Chapman J, Goldrath K, Ferrand R. Perinatally acquired HIV infection in adolescents from sub-Saharan Africa: a review of emerging challenges
. Lancet Infect Dis
19. Barnighausen T, Chaiyachati K, Chimbindi N, Peoples A, Haberer J, Newell ML. Interventions to increase antiretroviral adherence in sub-Saharan Africa: a systematic review of evaluation studies
. Lancet Infect Dis
20. Ramjam RCA, Vitoria M, Mills EJ, Hill A, Cooke G, Ford N. Patient and programme impact of fixed-dose combination antiretroviral therapy: a systematic review and meta-analysis
. Trop Med Int Health
21. Greub G, Gallant S, Zurn P, Vannotti M, Bürgisser P, Francioli P, et al. Spare nonoccupational HIV postexposure prophylaxis by active contacting and testing of the source person
22. Kim JC, Askew I, Muvhango L, Dwane N, Abramsky T, Jan S, et al. Comprehensive care and HIV prophylaxis after sexual assault in rural South Africa: the Refentse intervention study
23. Al-Dakkak I, Patel S, McCann E, Gadkari A, Prajapati G, Maiese EM. The impact of specific HIV treatment-related adverse events on adherence to antiretroviral therapy: a systematic review and meta-analysis
. AIDS Care
24. Nachega JB, Parienti JJ, Uthman OA, Gross R, Dowdy DW, Sax PE, et al. Lower pill burden and once-daily dosing antiretroviral treatment regimens for HIV infection: a meta-analysis of randomized controlled trials
. Clin Infect Dis
25. Baeten JM, Donnell D, Ndase P, Mugo NR, Campbell JD, Wangisi J, et al. Antiretroviral prophylaxis for HIV prevention in heterosexual men and women
. N Engl J Med
26. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al. Prevention of HIV-1 infection with early antiretroviral therapy
. N Engl J Med