These cross-sectional data from a large multi-country cohort of African WLHIV on life-time ART reveal a high prevalence (about half) of unintended last pregnancy and low rates of current use of effective contraception among those with contraception need. About one-in-five women with contraceptive need reported use of traditional methods which are known to be ineffective, and use of long-acting reversible contraceptives (LARCs) was very low (~ 20%). Based on multivariable regression analyses, marriage/stable relationship, sexual activity, future fertility desires, relative HIV control (≤ 1000 HIV copies/ml), and education were independently associated with use of reversible effective contraceptives (RECs); but no associations were observed with age, socio-economic factors (employment, household electricity), unintended last pregnancy, and clinic travel-time. LARC use was associated with relative economic independence (formal/self-employment) and HIV-control, but not the other factors. Compared to Zimbabwe, rates of REC use were similar in South Africa but lower in Malawi and Uganda; and LARC rates were lower in South Africa, Malawi, and Uganda.
The unacceptably high unintended pregnancy rates observed in this study, which varied considerably across countries, are comparable to high prevalence (35–70%) reports from previous African WLHIV studies in similar resource-limited settings in Rwanda (2007), Botswana (2010–2012), South Africa (2009–2010), Eswatini (2010), Zimbabwe (2012), and Kenya (2016). – However, the extremely high prevalence observed in this study at the Umlazi, Durban site in South Africa (81.9%) has not been reported previously. Moreover, it is likely that the prevalence observed in this urban/peri-urban study population is a conservative estimate of the respective country-level burdens since access to reproductive health services in many African countries is reportedly more than two-fold higher in urban versus rural settings where the majority of the population resides.  It has also been reported that WLHIV in these African settings who have not initiated lifelong ART or women who are not aware of their HIV infection status have a higher likelihood of reporting unintended pregnancy. , ,  All the women in this study were aware of their HIV infection status and the majority (> 97%) had initiated life-long ART by the time they conceived the reported unintended last pregnancy. Data on contraceptive use prior to the reported unintended pregnancy was not available for this analysis which precluded assessment of other risk factors of unintended pregnancy among WLHIV reported in these settings including unmet need for effective contraceptives or contraceptive failures. , , 
Regarding contraceptive use in this study population, traditional methods (breastfeeding, withdrawal, abstinence, calendar, or condoms only) which are potentially ineffective particularly when used incorrectly and/or inconsistently , were common (~ 20%), and LARC use was unacceptably low. These findings are concerning particularly because this cohort of WLHIV had recently transitioned from the PROMISE trial, a controlled five year follow-up study setting during which they had the benefit of routine counselling and referral/access to standard-of-care reproductive-health clinics. ,  Contemporaneous studies in these African settings reported comparable high rates of using potentially ineffective contraceptive methods and similarly low LARC use. , , – The high prevalence of a single contraceptive method at the respective sites is consistent with reports from other African settings, compared to other regions which tend to report relatively more diverse contraceptive method-mix.  ,  Since our study women received their contraceptives from standard-of-care/non-study providers, the observed skewed patterns may be an artifact of the policy, provider-factors (training, bias and attitudes) or available stocks at country, regional or facility level. Unfortunately, the predominantly reported methods in this study, both injectable- and oral-contraceptive methods are heavily user-action dependent, including 3-monthly provider-clinic visits for repeat injections, or require diligently swallowing daily oral contraceptives coupled with routine provider-clinic visits for refills. Intermittent non-adherence over a protracted period of self-perceived contraceptive use may affect contraceptive effectiveness to low levels potentially comparable to non-contraceptive use. Implants and IUDs, which are the most effective reversible contraceptive methods for an extended period without requiring user-action were rarely reported: only about one-in-ten of our study women reported implant use in South Africa and Uganda, and about one-in-five in Malawi and Zimbabwe; and less than 4% reported IUDs. In these settings where provision of safe sterilization services is challenging coupled with restrictive age eligibility criteria, LARCs are important alternatives for women of any age who particularly do not want to have more children. Unfortunately, implants and IUDs require more provider skill and training. It has been suggested that women in these settings who are using short-acting hormonal methods might prefer a LARC if they had the opportunity,  however, busy clinics may shun provision of LARCs since they are provider-labor-intensive. For example, the low enthusiasm towards implants by providers in South Africa is attributed to a policy change that was deemed rushed with inadequate provider-training during the roll-out in 2014, followed shortly after by a policy against implant provision to women on efavirenz-based ART.
Collectively, our findings of high prevalence of unintended last pregnancy coupled with subsequent low REC usage rates, particularly LARCs, underscores the need to better understand individual-level, as well as community- and facility-level factors that may influence utilization of effective contraception by WLHIV in these settings. Concerningly, unintended last pregnancy did not appear to incentivize subsequent use of RECs, an indicator of perennial contextual factors that preclude utilization. Understandably, and consistent with previous reports, factors that were independently associated with REC use included future fertility desires; ,  correlates of self-perceived risk of pregnancy including sexual activity and stable/sustained sexual partnerships; ,  as well as relative HIV control,  a proxy for good health. However, the negative association between REC use and higher education in this study is counterintuitive, similar to previous African studies linking higher education with unmet contraceptive need, – and another suggesting education had no influence on effective contraceptive choice.  A plausible explanation is that WLHIV in our study with higher education delayed childbearing since higher education was correlated with desire for another child. The considerable variations in the contraception options reported, including REC as well as LARC use, across different countries in our study, is characteristic in the region. This heterogeneity reflects the existing substantial contextual dissimilarities, including level of political will and governments’ investment and commitment to contraceptive services across African countries.  It is not surprising that age was not associated with REC or LARC use in this study considering the relatively homogenous age group with a median (IQR) 31 (28–35) years of age. This is consistent with previous studies with similar age band comparisons., ,  Studies that reported associations between age and contraceptive use included younger women and adolescents, a group associated with higher unmet contraception needs , –. However, these studies did not specify use of effective contraception, and while they were conducted in high HIV-burden settings, individual HIV status was not ascertained. An Ethiopian study of WLHIV reported lower use of modern contraceptives among older women above 35 years compared to their 15–24 year old counterparts. 
This study had some limitations. A recall bias inherent to cross-sectional surveys was likely in these analyses based on individuals thinking back and reporting their pregnancy intent, although the magnitude was likely minimal since as has been argued, pregnancy related events are relatively easier to recall given their significance in a woman’s lifetime. A woman’s desire for a given pregnancy is fluid and intervening factors before, during and after pregnancy such as partner and/or other social-economic support, access to care, loss of the pregnancy or demise of an older child, etc., will likely influence the woman’s attitude towards the pregnancy.– A tendency towards “retrospective rationalization” of previously undesired pregnancies has been proposed ,  suggestive of underreporting of unintended pregnancy rates; although the reverse has been reported where women changed from intended to unintended pregnancy reports, in a repeated measure analysis. Also, the lack of data on contraceptive-choices immediately preceding the reported unintended pregnancies precluded the ascertainment of whether the reported unintended pregnancies resulted from non-use of effective contraceptives or contraceptive failure including user-related (inconsistent or incorrect use) or method-related failures. Also, contraceptive data was not available for about 30.4% of the women enrolled before the relevant reproductive health questions were added to the baseline questionnaire. The exclusion of these women from these analyses may have potentially resulted in a selection bias. Reassuringly, the distribution of baseline characteristics was similar among the excluded women compared to those who contributed to these analyses. Nonetheless, these study findings contribute to the growing body of empirical data suggestive of reproductive health challenges that persist in resource-limited African settings, which also have a heavy HIV burden. This analysis was based on a large sample of WLHIV on life-long ART from very high burden countries. Data was well characterized including laboratory confirmation of HIV status, a challenge in previous studies which used simpler self-report data.