Safety of Tenofovir Disoproxil Fumarate for Pregnant Women Facing the Coronavirus Disease 2019 Pandemic

Abstract

We assessed the teratogenicity of tenofovir, a human immunodeficiency virus (HIV) drug similar to remdesivir that is currently being evaluated for the treatment of coronavirus disease 2019 (COVID-19). Using US Medicaid Analytic eXtract (MAX) claims data (2000–2014), we identified a population-based pregnancy cohort of women with HIV who filled at least 1 prescription for antiretroviral therapies (ART) during the first trimester. Women on tenofovir disoproxil fumarate (TDF) were compared with women receiving ART without TDF. Major malformations were identified by International Classification of Diseases, Ninth Revision, codes using validated algorithms. Relative risks and 95% confidence intervals were estimated using propensity score stratification to control for potential confounders. We incorporated the results into prior knowledge by conducting a systematic literature review and a meta-analysis. Major congenital malformations were diagnosed in 37 out of 866 (4.27%) infants exposed to TDF and 38 out of 1,020 (3.73%) infants exposed to ART other than TDF; the adjusted relative risk was 1.21 (95% confidence interval: 0.77, 1.90). Estimates for specific malformations were imprecise. The pooled relative risk from the meta-analysis with 6 prior studies was 0.88 (95% confidence interval: 0.75, 1.03). Based on evidence accumulated in patients with HIV, first-trimester TDF use does not increase the risk of major congenital malformations overall in the newborn compared with other ART.

Abbreviations

 
• ART

  antiretroviral therapies

 
• CI

  confidence interval

 
• COVID-19

  coronavirus disease 2019

 
• HIV

  human immunodeficiency virus

 
• LMP

  last menstrual period

 
• PS

  propensity score

 
• SARS-CoV-2

  severe acute respiratory syndrome coronavirus 2

 
• TDF

  tenofovir disoproxil fumarate

Worldwide, as of the end of 2020, over 100,000,000 women had been pregnant since the beginning of the coronavirus disease 2019 (COVID-19) pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (1). Pregnant women are susceptible to SARS-CoV-2 infection and, if infected, might have a higher risk of developing severe COVID-19 than nonpregnant women of the same age (2–5). Cases of maternal death in pregnant women with COVID-19 have been reported (6). Moreover, severe COVID-19 might increase the risk of prematurity and other adverse obstetric or neonatal outcomes (2, 4, 7, 8), and vertical transmission cannot be completely ruled out (2–4, 9–11). Therefore, many pregnant women and their physicians confront the decision of whether to use treatments for COVID-19.

By September 2020, tenofovir had emerged as a promising therapy for SARS-CoV-2 infection. Molecular docking and extension reaction studies suggest that remdesivir, tenofovir, and other nucleoside reverse transcriptase inhibitors might be effective against SARS-CoV-2 by inhibiting its RNA polymerase, as long as they attain high intracellular concentration at target tissues (12–14). Ferret infection models have shown that tenofovir-emtricitabine reduces SARS-CoV-2 titers in nasal washes (15). In observational studies, human immunodeficiency virus (HIV)-positive patients receiving tenofovir disoproxil fumarate (TDF)-emtricitabine had an up to 50% lower risk for COVID-19 diagnosis and related hospitalization (16–18). Tenofovir is currently being evaluated in controlled randomized clinical trials for both treatment and prophylaxis of COVID-19 (19). Given the positive clinical trials results for remdesivir (20, 21), tenofovir is expected to be effective because the 2 drugs are in the same family of medications. A key difference between these drugs is that tenofovir can be administered orally, whereas remdesivir requires daily intravenous infusion.

While there is no evidence regarding the safety profile in pregnancy for many of the treatments currently being evaluated for COVID-19, including remdesivir, some data exist for tenofovir based on its use for the treatment of HIV infection and, to lesser extent, for hepatitis B virus infection. It is therefore important to obtain and communicate reliable information about the safety of tenofovir, including teratogenicity, before pregnant populations are exposed on a broad scale. We reviewed the literature and evaluated the risk for major congenital malformations associated with early in pregnancy exposure to tenofovir as HIV treatment using a large health-care utilization database.

METHODS

Study population

We conducted a cohort study of pregnancies nested in the Medicaid Analytic eXtract (MAX; January 1, 2000, to December 31, 2014), which includes enrollment information and health-care claims for Medicaid beneficiaries nationwide in the United States. The development of the linked mother-infant pregnancy cohort has been described previously (22). Briefly, women with a code indicating delivery were identified and linked to live-born infants based on shared family case numbers. Women were required to be continuously enrolled in Medicaid, without supplementary private insurance or restricted benefits, for 3 months prior to the estimated date of last menstrual period (LMP) through 1 month after delivery, and infants were required to be insured from birth until 3 months thereafter, unless they died sooner. Pregnancies with exposure to a known teratogenic medication (i.e., warfarin, antineoplastic agents, lithium, isotretinoin, misoprostol, thalidomide) during the first trimester and pregnancies with a chromosomal abnormality were excluded.

The study population was restricted to women with a recorded diagnosis of HIV between 3 months prior to LMP and delivery (see Web Table 1, available at https://doi.org/10.1093/aje/kwab109, for diagnostic and procedure codes) (23). Because infants exposed to antiretroviral therapies (ART) in the first trimester have been reported to have slightly higher risk of malformations than those exposed later in pregnancy (24–27), we further restricted the cohort to pregnancies exposed to ART during the first trimester, defined by at least 1 dispensing of an antiretroviral medication between LMP and 3 months after LMP.

Exposure

Women were considered exposed if they filled at least 1 prescription for TDF (the tenofovir formulation available during the study period) during the first trimester of pregnancy. The reference group was composed of pregnant women who received ART that did not include any dispensings of TDF from 3 months before LMP through the end of the first trimester to reduce the probability of exposure during early pregnancy from use of TDF dispensed before the start of pregnancy.

Use of 3-drug ART regimes has been the standard of care during pregnancy; it reduces the risk of perinatal HIV transmission to less than 1% (28, 29). During the study period, HIV patients on TDF most often received it with emtricitabine plus an integrase inhibitor, a protease inhibitor, or a nonnucleoside reverse transcriptase inhibitor (30, 31). Therefore, we also present results for the TDF-emtricitabine combination (the combination for which the decrease in risk of COVID-19–related hospitalization had been reported in the observational study) (16, 17) and accounted for the other antiretrovirals dispensed during the first trimester in the analyses by balancing on the number of distinct ART drugs in the regime using the propensity score. Note that the combination TDF-emtricitabine has been recommended in the United States for pre-exposure prophylaxis in women only after the study period.

Outcomes

The outcome of interest was a composite measure of all major congenital malformations because the number of exposed women was too small to evaluate specific malformations. The presence of malformations was defined using validated algorithms based on inpatient or outpatient diagnoses and procedures, which have been shown to identify malformations with high positive predictive values (75%–98%) (32–35). See Web Table 2 for diagnostic and procedure codes.

Covariates

Potential confounders and proxies for confounders considered included maternal sociodemographic characteristics (e.g., state of residence, age, race/ethnicity), markers of HIV disease severity (e.g., infections), comorbid medical conditions (e.g., diabetes, hypertension), other prescription drugs dispensed (e.g., antimicrobials), and general markers of the burden of illness (see Table 1 for a comprehensive list).

Analyses

Propensity scores (PS) were calculated using a logistic regression model that estimated the probability of being dispensed TDF in the first trimester based on confounder values. A separate model was created for TDF-emtricitabine. All variables listed in Table 1 were included in the PS, except for the specific ART medications. After trimming observations from the nonoverlapping regions of the PS distribution, we created 50 equally sized PS strata based on the distribution among the treated women (36). In the outcome models, the reference observations were weighted using the distribution of the TDF-treated among PS strata. Baseline characteristics and total number of antiretrovirals prescribed during the first trimester were compared between exposed and reference groups, before and after PS stratification. The risk of major malformations was summarized in the full and stratified samples. Relative risks with their 95% confidence intervals were estimated using generalized linear models.

The research was approved by the institutional review board of Brigham and Women’s Hospital, which waived the need for informed consent.

Literature review

We conducted a systematic review and meta-analysis for studies that examined the relationship between use of TDF early in pregnancy and overall congenital malformations. We searched MEDLINE via PubMed and additionally did a Google-based web search for abstracts with terms related to “tenofovir,” “birth defect/malformation,” and “pregnancy.” The references cited in all included studies were reviewed to identify additional eligible studies. Search criteria are described in detail in Web Table 3.

Articles were included if they were written in English and reported adequate information to calculate a relative risk for tenofovir exposure in the first trimester and congenital malformations. We excluded conference abstracts, animal studies, basic science research, case reports, case series, and commentaries. When reports were published on the same cohort, we included only the most recent publication to avoid duplication. We further restricted the meta-analysis to studies that had a comparison group that received ART.

Two authors (L.S., S.H.-D.) screened the titles and abstracts of all identified articles according to the inclusion and exclusion criteria listed above. For articles passing the initial screen, the 2 authors independently performed full text review, finalized inclusion decisions, and extracted the relevant information using a standardized form. All discrepancies were resolved through discussion until reaching a consensus. A fixed-effects meta-analysis was performed, and results are reported in a forest plot. The I² metric (which represents the percentage of variance in meta-analysis that is attributable to between-study heterogeneity) was computed to quantify between-study heterogeneity.

RESULTS

Study population

The cohort included 872 pregnancies exposed to TDF during the first trimester and 1,033 exposed to ART that did not include TDF in the first trimester (Table 1). Among TDF users, 743 received TDF-emtricitabine. Before PS stratification, there were some small differences in baseline characteristics between the exposure groups; women with first-trimester TDF exposures were slightly older, more likely to be from the Northeast than from the South, had more infections, had more trimethoprim prescriptions, had more health-care utilization, and received more antiretrovirals. The PS stratification procedure resulted in a sample of 866 women in the TDF group and 1,020 in the comparator group. After adjustment through PS stratification, all covariates were well balanced, including the number of distinct ART drugs (Table 1). Specific ARTs were not included in the PS and were not expected to be balanced.

Tenofovir and major congenital malformations

Before stratification, the prevalence of major congenital malformations was 4.24% for TDF-exposed infants (37 events out of 872), 4.17% for infants exposed to TDF-emtricitabine specifically (31 events out of 743) (Figure 1), and 3.68% for infants exposed to ART without TDF (38 events out of 1,033) (Table 2). The prevalence of major malformations after first-trimester exposure was similar for the most common ART (Figure 1). Compared with infants exposed to other ART as a group, the unadjusted relative risk was 1.15 (95% confidence interval (CI): 0.74, 1.80) for TDF and 1.13 (95% CI: 0.71, 1.81) for TDF-emtricitabine (Table 2). After PS stratification, the prevalence of major congenital malformations was 4.27% for TDF-exposed infants (37 events out of 866), 4.18% for infants exposed to TDF-emtricitabine specifically (31 events out of 741), and 3.73% for infants exposed to ART without TDF (38 events out of 1,020). Compared with infants exposed to other ART as a group, the adjusted relative risk for TDF was 1.21 (95% CI: 0.77, 1.90); it was 1.14 (95% CI: 0.71, 1.82) for TDF-emtricitabine. The most common malformations among TDF-exposed infants were limb defects and genital defects, while for other ART they were cardiac and gastrointestinal. The small numbers did not allow estimation of relative risks for specific defects.

Meta-analysis with previous studies

The PubMed search yielded 27 citations of interest, of which 14 were directly identified as search results, and 13 were identified based on sources cited in other search results. All 27 identified studies underwent a full text review, and results from 6 articles were included in the meta-analysis. Three additional citations were identified via the Google search strategy, of which none met the criteria described above to be included in the meta-analysis. While study designs varied between articles (Web Table 4), between-study heterogeneity was low (P value for heterogeneity = 0.346). Results from the meta-analysis indicated no increased prevalence of congenital malformations overall associated with TDF exposure during the first trimester compared with other ART without TDF. Since few studies provided adjusted estimates and the estimates did not change meaningfully with adjustment within those studies that did, we used crude relative risk estimates for the meta-analysis. The pooled relative risk not including the current study was 0.85 (95% CI 0.71, 1.00); upon including the current study it was 0.88 (95% CI: 0.75, 1.03; Figure 2).

DISCUSSION

Principal findings

Utilizing a large health-care database of publicly insured individuals in the United States from the years 2000–2014, we identified a nationwide cohort of pregnant women with HIV and estimated the relative prevalence of major congenital malformations in their newborns following exposure to TDF. Women who filled prescriptions for TDF, with or without emtricitabine, during the first trimester had a similar prevalence of malformations in their newborns to women with HIV who filled prescriptions for other ART that did not include TDF during the first trimester. However, based on this study alone, we could only deem as unlikely a 2-fold increased risk. The study size was also too small to consider specific birth defects.

Literature review

We incorporated the results from our analysis into existing knowledge by conducting a systematic literature review. Both the pooled relative risk from the meta-analysis and that estimated from the Medicaid cohort are consistent with a null finding for the impact of TDF on malformations, yet point estimates and upper limits of the 95% confidence intervals were quite different. The largest study was the Antiretroviral Pregnancy Registry, which reported a prevalence of congenital malformations for first-trimester exposure of 2.8% for ART overall (n = 10,120), 2.4% for TDF (n = 4,005), and 2.6% for emtricitabine (n = 3,345). Their most recent report also included tenofovir alafenamide (n = 233), approved for HIV treatment in the United States since November 2015, with an estimated prevalence of malformations of 5.2% (2.7%–8.8%) (37). Among the other studies, some estimated a slightly lower prevalence of malformations in infants with first-trimester exposure to TDF (26, 27, 38), while others found a slightly higher prevalence of malformations for TDF (24, 25) than for other ART (Table 3). Relative risks for emtricitabine were uniformly close to the null across studies. TDF has also been used for chronic hepatitis B virus infection during pregnancy. Although the accumulated evidence is more limited than for HIV, it also suggests no significant increased risk of malformations associated with first-trimester TDF exposure (37, 39).

We included only studies with internal active reference groups, hence minimizing confounding by HIV or associated comorbidities. Although the publications did not report adjusted relative risk estimates for the specific comparison of TDF versus other ART, residual confounding is expected to be modest given that our own estimate moved half a decimal point upon adjustment. Similarly, although some of the studies had substantial losses to follow-up and others allowed retrospective enrollment, differential ascertainment of malformations by ART would be unlikely. Only 2 studies included stillbirths and elective terminations of pregnancy for fetal anomaly (37, 38). However, most terminations were for chromosomal anomalies, and the relatively low number of stillbirths with malformations had little impact in the results.

Although statistical testing did not suggest significant heterogeneity in relative risk estimates, when interpreting the results of the meta-analysis it is important to be aware of substantial diversity in the designs of the included studies. In addition to random error, heterogeneity in results might be due to a number of important differences, including outcome definitions (e.g., Metropolitan Atlanta Congenital Defects Program vs. EUROCAT), internal study validity (e.g., amount of confounding control, although all studies restricted to women with HIV and ART therapy during the first trimester), or geographic differences (e.g., Europe vs. United States). For example, the estimated prevalence of malformations for TDF exposure ranged from 2.4% to 8.9% across studies. Despite these differences, all relative risk estimates consistently fluctuated around the null.

Other pregnancy outcomes

Results must be interpreted alongside other safety findings related to TDF use in pregnancy. In utero TDF exposure has been linked to reduced infant bone mineral content (40), although this effect has been refuted in a recent clinical trial (41). Also, the PROMISE (Promoting Maternal and Infant Survival Everywhere) trial, conducted across multiple sites in sub-Saharan Africa and India, identified a potential safety concern for women randomized to receive TDF, emtricitabine, and ritonavir-boosted lopinavir. These participants were twice as likely to have infants born very prematurely (<34 weeks) or at very low birth weight (<1,500 g), compared with women concurrently randomized to receive zidovudine, lamivudine, and ritonavir-boosted lopinavir (29). Infants with in utero exposure to the TDF regime also had substantially greater risk of death within 14 days postpartum. However, these adverse birth outcomes were not elevated in 2 US-based cohort studies (the Surveillance Monitoring for ART Toxicities and International Maternal Pediatric Adolescent AIDS Clinical Trial Network P1025 studies) (31). Overall, the World Health Organization recommends TDF-based ART regimens as first-line therapy for all HIV-infected adults, including pregnant women (42), and it is one of the most commonly used regimens among HIV-infected pregnant women in the United States (43).

Strengths and limitations

This study is, to our knowledge, the second largest in terms of TDF-exposed pregnancies, thus meaningfully expanding the evidence. Moreover, it used real-world data from a population-based sample (vs. self-selected volunteers), used prospectively collected exposure information with respect to outcomes (vs. inclusion of participants after prenatal screening), included an internal reference group (vs. using as a reference for the risk of malformations estimations from external populations), and carefully reduced the possibilities of confounding inherent to observational studies by restricting to women on ART for HIV during the first trimester and further balancing characteristics through PS stratification.

In addition to these strengths, our study is also subject to certain limitations. First, we included only women with a live-born delivery because the diagnosis of malformations in abortions and stillbirths is incompletely recorded in health-care utilization data. This approach might have resulted in the exclusion of pregnancies with the outcome, given that fetuses with malformations are more likely to experience fetal death or termination. Therefore, the prevalence of major malformations at birth reported in this study could underestimate the risk in pregnancy. However, fetal losses would need to be unrealistically differential between women on different ART strategies to be a meaningful source of selection bias (44). Second, identification of major congenital malformations was based on diagnosis and/or procedure codes recorded in claims. While our algorithms to identify malformations have shown a high positive predictive value (32–34), the potential for some misclassification remains. Misclassification would tend to bias relative risks toward the null unless it were differential among different ART strategies, which is unlikely. Third, information on TDF-exposed pregnancies was obtained from claims of filled prescriptions. Because some women might fill prescriptions for medications but not use them, our study might misclassify TDF-unexposed pregnancies into the exposed group, thus underestimating any potential effect. However, women with HIV that filled their prescription are likely to use their drug to avoid vertical transmission. Fourth, while we required a diagnosis of HIV, if such codes were misapplied, then it would be possible for some patients in our cohort to be using TDF as HIV pre-exposure prophylaxis. However, prophylaxis with TDF was approved by the Food and Drug Administration in 2012 and recommendations for use in women came later; thus, this indication is unlikely for women who delivered before 2014. Moreover, our finding of increased infections and more antiretrovirals received by women in the TDF cohorts suggests that this was not a frequent occurrence. Fifth, we provide evidence for the safety of patterns of use in women with HIV, which could be different from that expected for COVID-19. However, at the same TDF doses, which virus triggered the indication is unlikely to modify teratogenic effects on the fetus. Sixth, the Medicaid Analytic eXtract data set, along with most of the studies included in the meta-analysis, is an observational cohort, and there is always some potential for residual confounding, although this should be limited by the restriction to women with the indication, the use of PS stratification, and an active comparator group. Finally, the numbers were too small to assess specific malformation groups.

Conclusions

In conclusion, accumulated evidence suggests that first-trimester exposure to TDF in women with HIV does not increase the risk of major congenital malformations overall in the newborn compared with other antiretroviral treatment strategies. These findings are reassuring for pregnant women who might be eligible for TDF treatments in the context of the COVID-19 pandemic.

ACKNOWLEDGMENTS

Authors affiliations: Department of Epidemiology, Harvard T. H. Chan School of Public Health, Boston, Massachusetts, United States (Sonia Hernández-Díaz); Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, United States (Brian T. Bateman, Loreen Straub, Yanmin Zhu, Helen Mogun, Michael Fischer, Krista F. Huybrechts); and Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, United States (Brian T. Bateman).

The study was supported by internal funds of the Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, Harvard Medical School. B.T.B, S.H.-D., and K.F.H. are supported by the National Institute of Mental Health (grant R01 MH116194), the National Institute on Drug Abuse (grants R01 DA044293, R01 DA049822), and the Eunice Kennedy Shriver National Institute of Child Health and Human Development (grant R01 HD097778).

The data underlying this article are available through Centers for Medicare and Medicaid Services (CMS) conditional on data use agreements and fees.

S.H.-D. reports being an investigator on grants to her institution from Takeda for unrelated studies; personal fees from UCB and Roche outside the submitted work; and having served as an epidemiologist with the North America antiepileptic drugs (AED) pregnancy registry, which is funded by multiple companies. B.T.B. received research grants to Brigham and Women’s Hospital from Eli Lilly, Baxalta, and Pacira for unrelated studies; received personal fees from Aetion and from Alosa Foundation outside the submitted work; and served on an expert panel for a postpartum hemorrhage quality improvement project that was conducted by the Association of Women’s Health, Obstetric, and Neonatal Nurses and funded by a grant from Merck for Mothers. M.F. received a research grant to Brigham and Women’s Hospital from Surescripts for unrelated studies and personal fees from Alosa Foundation outside the submitted work. K.F.H. received research grants to Brigham and Women’s Hospital from Eli Lilly for unrelated studies. The other authors report no conflicts.

REFERENCES