Secondary Logo

Journal Logo

Letters to the Editor


Yiannoutsos, Constantin T. PhD*; Wools-Kaloustian, Kara K. MD; Musick, Beverly S. MS*; Elul, Batya PhD, MSc‡,§

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: June 1, 2017 - Volume 75 - Issue 2 - p e56-e57
doi: 10.1097/QAI.0000000000001353
  • Free

To the Editors:

In response to the letter by Strassle et al, we would like to provide a rejoinder regarding some of the salient points raised by these authors. To do so, it would be useful to address each of these issues separately to clarify some of the confusion that may exist about the statistical analyses in our article.1

The first issue raised by Strassle et al concerned the treatment of loss to program in our analysis. Loss to program (also called loss to follow-up or patient dropout), particularly in the low- to middle-income settings where we work, prevents complete observation of a number of patient outcomes. Of most relevance to our article, loss to program limited observation of incident pregnancies but also other relevant factors such as re-engagement in HIV care at programs outside the initial clinical site and mortality after dropout. For this reason, we considered loss to program as a competing risk rather than an event censoring the time until the experience of an incident pregnancy, our endpoint of interest. It is critical here to recognize that this choice has no bearing on the ultimate numerical estimates of the impact of antiretroviral therapy (ART) on the likelihood (hazard) of incident pregnancy but rather affects solely the interpretation of the results. The reason is that, operationally, loss to program abbreviated the follow-up time in all analyses presented and thus resulted in censoring along with all other events limiting complete follow-up such as death and known transfers, in addition to administrative censoring (end of observation due to database closure) resulting in the end of follow-up. When loss to program is modeled as a competing event, ART is considered a risk factor for the cause-specific hazard of experiencing an incident pregnancy versus the cause-specific hazards of experiencing other events such as death, loss to program, or transfer to another care and treatment facility. So, absent some correction on the current estimates resulting from additional data obtained from some or all the dropouts,2,3 or sensitivity analyses based on historical or otherwise external data, our estimates of the effect of ART on incidence pregnancy do not change whether loss to program is viewed as a competing risk or a censoring event.

Strassle et al additionally noted that loss to program may have downwardly biased estimates of the impact of ART on incident pregnancy. Although we concur with this assessment, we do not agree that attenuation of the estimates is necessarily the result of informative censoring induced by loss to program. As is well known,4 even noninformative censoring can attenuate associations between a risk factor and the outcome of interest. Informative censoring, however, may result in underestimation or overestimation of the true effect,4,5 as the authors correctly state. However, their proposed remedy, inverse probability of censoring weighting methods,6 is unlikely to fully address these biases because data in patients lost to program in this setting are, most likely, not missing at random,7 a key assumption in inverse probability of censoring weighting methods. In the presence of data missing not at random, the effect of ART could have been underestimated or overestimated in ways that are not obvious. For example, it could have been underestimated given the propensity of pregnant women to drop out of care at higher rates compared with the general patient population (see for example the references cited by Strassle et al8–11). As these women are more likely to experience an incident (repeat) pregnancy, as per our analysis, dropping out of care may reduce the ranks of patients on ART who are more likely to experience an incident pregnancy (since their probability of accessing ART after leaving an HIV care and treatment program is likely much lower). However, the effect of ART on pregnancy, which is based on our analysis of patients under observation, may also have been overestimated, if a significant portion of loss to program were because of unreported death, particularly among the sickest of patients who may also be less likely to get pregnant11 (resulting in the attrition, among patients receiving ART, of those who might be less likely to experience an incident pregnancy while on treatment). Unfortunately, the proportion of patients who belong to each of these groups and HIV-care seeking behaviors after leaving the current program can only partially be estimated based on data obtained while these patients were in care (the very essence of data missing not at random).

Instead of speculating about these disparate sources of potential bias, we opted to account for biases which are addressable based on the data available to us. The largest and most obvious of these was bias arising from “time-dependent confounding” of the role of ART on pregnancy.12 Such confounding was produced by factors associated both with the probability of an incident pregnancy and with the likelihood of ART initiation. These included CD4 count, weight, and World Health Organization stage, contextual factors of care plus, most notably, previous pregnancies. In particular, measures of disease severity are associated with the likelihood of initiating therapy and, once therapy commences, are affected by patient response to ART. To minimize the effect of such biases, we used inverse probability weighting. The corrected estimates transformed an apparently detrimental impact of ART on pregnancy to a marginally beneficial one, albeit much lower than what has previously been reported. In the discussion of our article, we provided a number of possible explanations for this departure from the previous literature. This included the fact that most prior studies did not fully account for the role of previous pregnancies in their cohort selection and, in one case, explicitly recruited women who were either pregnant or had recently become pregnant.13 Given policies to prevent mother-to-child transmission of HIV, women who experience prior pregnancies constitute a large subgroup of women starting therapy with less advanced disease (the subgroup of women most likely to experience a subsequent pregnancy).

In conclusion, although we cannot discount the possibility that biases arising from loss to program may have resulted in an underestimation of the impact of ART on incident pregnancy, absent additional information on the dynamics between loss to program, pregnancy, and HIV disease, our analyses exhaust what can be gleaned from observational data collected as part of routine clinical care in this setting.


1. Elul B, Wools-Kaloustian KK, Wu Y, et al. Untangling the relationship between antiretroviral therapy use and incident pregnancy: a marginal structural model analysis using data from 47,313 HIV-positive women in East Africa. J Acquir Immune Defic Syndr. 2016;72:324–332.
2. Geng EH, Emenyonu N, Bwana MB, et al. Sampling-based approach to determining outcomes of patients lost to follow-up in antiretroviral therapy scale-up programs in Africa. JAMA. 2008;300:506–507.
3. Yiannoutsos CT, An MW, Frangakis CE, et al. Sampling-based approaches to improve estimation of mortality among patient dropouts: experience from a large PEPFAR-funded program in Western Kenya. PLoS One. 2008;3:e3843.
4. Copeland KT, Checkoway H, McMichael AJ, et al. Bias due to misclassification in the estimation of relative risk. Am J Epidemiol. 1977;105:488–495.
5. Bakoyannis G, Siannis F, Touloumi G. Modelling competing risks data with missing cause of failure. Stat Med. 2010;29:3172–3185.
6. Robins J, Rotnitzky A, Zhao LP. Analysis of semiparametric regression models for repeated outcomes in the presence of missing data. J Am Stat Assoc. 1995;90:106–121.
7. Geng EH, Glidden DV, Bangsberg DR, et al. A causal framework for understanding the effect of losses to follow-up on epidemiologic analyses in clinic-based cohorts: the case of HIV-infected patients on antiretroviral therapy in Africa. Am J Epidemiol. 2012;175:1080–1087.
8. Kaplan R, Orreel C, Zwane E, et al. Loss to follow-up and mortality among pregnant women referred to a community clinic for antiretroviral treatment. AIDS. 2008;22:1679–1681.
9. Phillips T, Thebus E, Bekker LG, et al. Disengagement of HIV-positive pregnant and postpartum women from antiretroviral therapy services: a cohort study. J Int AIDS Soc. 2014;17:19242.
10. Wang B, Losina E, Stark R, et al. Loss to follow-up in a community clinic in South Africa—roles of gender, pregnancy and CD4 count. S Afr Med J. 2011;101:253–257.
11. Makumbi FE, Nakigozi G, Reynolds SJ, et al. Associations between HIV antiretroviral therapy and the prevalence and incidence of pregnancy in rakai, Uganda. AIDS Res Treat. 2011;2011:519492.
12. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11:550–560.
13. Myer L, Carter RJ, Katyal M, et al. Impact of antiretroviral therapy on incidence of pregnancy among HIV-infected women in Sub-Saharan Africa: a cohort study. PLoS Med. 2010;7:e1000229.
Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.