17α-hydroxyprogesterone caproate is a synthetic progestogen initially approved in the 1950s to treat gynecologic and obstetrical conditions. Despite continued concerns about safety and short-term efficacy regarding the use of 17α-hydroxyprogesterone caproate for the prevention of preterm birth in pregnant women, little is known about the long-term effects of 17α-hydroxyprogesterone caproate on the health of the offsprings.
To examine the association between in utero exposure to 17α-hydroxyprogesterone caproate and the risk of cancer in the offspring.
The Child Health and Development Studies was a population-based cohort of >18,000 mother-child dyads receiving prenatal care in the Kaiser Foundation Health Plan (Oakland, CA) between 1959 and 1966. Clinical information was abstracted from the mothers’ medical records beginning 6 months before pregnancy through delivery. We identified the number and timing of 17α-hydroxyprogesterone caproate injections during pregnancy. Incident cancers diagnosed in the offspring were ascertained through 2019 by linkage to the California Cancer Registry. We used the Cox proportional hazard models to estimate the adjusted hazard ratios and their 95% confidence intervals, with the follow-up time accrued from the date of birth through the date of cancer diagnosis, death, or last contact.
A total of 1008 offspring were diagnosed with cancer over 730,817 person-years of follow-up. Approximately 1.0% of the offspring (n=234) were exposed in utero to 17α-hydroxyprogesterone caproate. Exposure in the first trimester was associated with an increased risk of any cancer (adjusted hazard ratio, 2.57; 95% confidence interval, 1.59–4.15), and the risk increased with the number of injections (1–2 injections: adjusted hazard ratio, 1.80; 95% confidence interval, 1.12–2.90; ≥3 injections: adjusted hazard ratio, 3.07; 95% confidence interval, 1.34–7.05). Exposure in the second or third trimester conferred an additional risk for the male (adjusted hazard ratio, 2.59; 95% confidence interval, 1.07–6.28) but not for the female (adjusted hazard ratio, 0.30; 95% confidence interval, 0.04–1.11) offspring. The risk of colorectal (adjusted hazard ratio, 5.51; 95% confidence interval, 1.73–17.59), prostate (adjusted hazard ratio, 5.10; 95% confidence interval, 1.24–21.00), and pediatric brain (adjusted hazard ratio, 34.72; 95% confidence interval, 7.29–164.33) cancer was higher in the offspring first exposed to 17α-hydroxyprogesterone caproate in the first trimester than the offspring not exposed.
Caution using 17α-hydroxyprogesterone caproate in early pregnancy is warranted, given the possible link with cancer in the offspring.
17α-hydroxyprogesterone caproate (17-OHPC) is a synthetic progestogen initially approved in 1956 to treat several gynecologic and obstetrical conditions, including habitual and threatened abortion in pregnant women. 17-OHPC was administered at a high dosage (250 mg/mL intramuscular injection) to millions of pregnant women in the US (Delalutin by Bristol-Myers Squibb) and Europe (Proluton by Schering) during the 1950s and 1960s. In October 1973, the US Food and Drug Administration (FDA) noted a lack of substantial evidence to support 17-OHPC for the prevention of habitual and threatened abortion and raised concerns of an association of 17-OHPC with congenital heart defects in the offspring. , They subsequently removed all pregnancy-related indications from its label, citing the possibility of teratogenic effects associated with systemic use. Although the labeling requirements of progestogens were later modified, the FDA withdrew their approval of 17-OHPC in September 2000 at the request of the manufacturer and because it was no longer being marketed.
Why was this study conducted?
Despite the continued use of 17α-hydroxyprogesterone caproate (17-OHPC) in pregnant women, little is known about its long-term effects on the health of the offspring.
The offspring exposed to 17-OHPC in utero had a higher risk of any cancer than those who are not exposed.
The risk was higher with exposure in the first trimester and with ≥3 injections.
Exposure in late pregnancy conferred an additional risk of cancer in male but not in female offspring.
What does this add to what is known?
Caution using 17-OHPC in early pregnancy is warranted, given the possible link with cancer in the offspring.
As part of its Accelerated Approval Program, the FDA again approved 17-OHPC in February 2011 (Makena by Covis Pharma, Waltham, MA) for pregnant women with a history of spontaneous preterm birth on the basis of a randomized trial that demonstrated reductions in the incidence of preterm birth at 37 weeks. Detailed analyses of that trial (reviewed by Calda ) raised concerns regarding fetal toxicity, noting a small, though not statistically significant, increase in fetal deaths and stillbirths among the women who received 17-OHPC. Two large trials of 17-OHPC in multiple gestations similarly showed an excess of serious adverse fetal or neonatal events. The signals for embryo-fetal toxicity associated with 17-OHPC were later confirmed in rhesus monkeys and rodents in a review of experimental studies.
The FDA required a confirmatory trial as part of their accelerated approval of 17-OHPC; the Progestin’s Role in Optimizing Neonatal Gestation trial was completed in March 2019 and demonstrated no reduction in the incidence of preterm birth at 35 weeks or neonatal morbidity and mortality. Therefore, the Center for Drug Evaluation and Research recommended in October 2020 to withdraw approval and maintained this position after the Evaluating Progestogens for Preventing Preterm birth International Collaborative (EPPPIC) meta-analysis was published. Conflicting perspectives , highlight the ongoing controversy surrounding 17-OHPC.
Despite repeated concerns of safety and short-term efficacy, little is known about the long-term effects of 17-OHPC on the health of offspring. The potential for synthetic hormones to disrupt embryologic development and manifest adverse health outcomes is well-established by epidemiologic studies of the synthetic estrogen diethylstilbestrol (DES) and a large literature of experimental studies. In utero exposure to DES increases the risk of cancer in offspring across the life course. , Similarly, exposure to synthetic progestogens during fetal development may permanently alter organ morphology and function. This is consistent with evidence that 17-OHPC crosses the placental barrier , and the fetus and placenta are capable of metabolizing 17-OHPC , ; embryo-fetal toxicity signals are also identified in trials and experimental studies. Furthermore, like DES, early exposure to 17-OHPC may lead to cellular, molecular, and epigenetic changes that play a role in carcinogenic processes later in life. ,
Here, we examine the association between in utero exposure to 17-OHPC and cancer in offspring in the Child Health and Development Studies (CHDS), which was a population-based cohort of >18,000 mother-child dyads receiving care in the Kaiser Foundation Health Plan (Oakland, CA) in the 1960s who were followed for 60 years.
Materials and Methods
Established in 1959, the CHDS enrolled nearly all (98%) the pregnant women receiving prenatal care from the Kaiser Foundation Health Plan (Oakland, CA) between June 1959 and September 1966, with deliveries through June 1967 (n=18,751 live births excluding neonatal deaths among 14,507 mothers). Additional details of the CHDS and the methodology are available elsewhere.
We monitored the CHDS participants by annual linkage to the California Department of Motor Vehicles, California Department of Vital Statistics, and California Cancer Registry. Mothers and their families were matched to these sources using an accumulated name and address history, routinely identifying >80% of families.
We ascertained the incident cases of cancer in offspring through 2019 by linkage with the California Cancer Registry. The California Cancer Registry is one of the largest cancer registries in the US and meets the highest quality data standards set by the National Program of Cancer Registries and the US Centers for Disease Control and Prevention. , We used a rigorous protocol to verify the cases, comparing the fixed (eg, birth date, sex, race) and changeable (eg, address) identifiers by manual review.
In utero exposure to 17α-hydroxyprogesterone caproate
Clinical information including prenatal visits, diagnosed conditions, and prescribed medications was abstracted from the mothers’ medical records, beginning 6 months before pregnancy through delivery. All medications were linked to the date and the conditions for which they were prescribed. We identified the mothers who received 17-OHPC during pregnancy and measured the in utero exposure at the trimester of the first exposure (first trimester: 0–90 days; second trimester: 91–180 days; third trimester: ≥181 days). We also measured the total number of 17-OHPC injections (1–2 or ≥3 injections).
We used the Cox proportional hazards model to estimate the hazard ratios (HRs) and their 95% confidence intervals (CIs) for the association between in utero exposure to 17-OHPC and any cancer in the offspring; this was done overall and by the trimester of the first exposure and the number of injections. To account for the correlation between observations from siblings (n=4244), we used robust sandwich estimators. The follow-up time was accrued from the date of birth through the date of cancer diagnosis, the date of death, or the date of last contact. Because the participants were regularly monitored for residence and vital status, we used the year of last contact from all sources to create the date of last contact. We assessed the proportional hazards assumption in all models by visually examining the plots of the survival function vs survival time and log(−log[survival]) vs log (survival time). The assumption was not violated in any model.
Because the distribution of the cancer types differed in the offspring exposed and those not exposed to 17-OHPC, we explored some of these cancers in more detail, including prostate, colon and rectum, and pediatric brain cancers. We selected these cancers because there were multiple diagnoses of these in the exposed offspring. We used the Cox proportional hazards model to estimate the HRs and their 95% CIs for the associations between in utero exposure to 17-OHPC and cancer in the offspring, overall and in the first trimester. The follow-up time was accrued from the date of birth through the date of cancer diagnosis, date of death, or the date of last contact (or age 18 years for the model of pediatric brain cancer).
We examined the interaction with respect to in utero exposure to 17-OHPC and the offspring sex. For purposes of this analysis, we defined in utero exposure as the first exposure in early (first trimester) or late (second or third trimester) pregnancy. We compared the nested models with and without early pregnancy×sex and late pregnancy×sex product terms using a likelihood ratio test; we calculated the contrasts from linear combinations of the product terms to estimate the associations of exposure in early pregnancy vs no exposure and exposure in late pregnancy vs no exposure, jointly for male and female offspring. We also estimated the stratum-specific HRs.
Across all the models, the following were evaluated a priori as confounders, individually and simultaneously: the year of birth, sex, maternal age at pregnancy, race or ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Asian, other), maternal education (less than high school, high school or trade school, some college, or more), parity at pregnancy (primiparous, multiparous), total family income (above or below the median, adjusted for 1960 dollars), gestational age (<37 weeks, ≥37 weeks), maternal body mass index (underweight or normal, overweight, obese), and birthweight. We selected these confounders because they may be directly or indirectly related to the mothers’ use of 17-OHPC and the offsprings’ risk of cancer. We used the height and weight reported by the mothers during in-person interviews at enrollment or recorded at the first prenatal visit to measure the maternal body mass index. The gestational age was calculated by subtracting the date of the last menstrual period from the date of delivery (range 20–42 weeks). To select the most parsimonious model, we retained the potential confounders that, if removed from the model, changed the effect estimate by >10%. ,
We also estimated the incidence rates (of any cancer) and 95% CIs on the basis of the discrete probability distribution for a binomial parameter, separately by the trimester of the first exposure to 17-OHPC and the number of injections.
We conducted several sensitivity analyses to enhance the rigor of our approach, which is detailed below.
Confounding by indication
We examined the association between any cancer in the offspring and the conditions indicating 17-OHPC in mothers, such as threatened abortion.
Using age as the underlying time parameter, we estimated the HRs and their 95% CIs from Cox proportional hazards regression models. We included the product terms with the age at follow-up (±50 years) and the first exposure to 17-OHPC in the first trimester and compared the models with and without product terms using a likelihood ratio test.
Probabilistic bias analysis
The association between in utero exposure to 17-OHPC and cancer in the offspring may be confounded by shared factors between mother and offspring and that were not measured in the CHDS. We conducted a probabilistic bias analysis , to model errors from unmeasured confounding.
Missingness ranged from 0.0% (birthweight, year of birth) to 13.3% (maternal body mass index). We used multiple imputation by fully conditional specification to estimate the associations between in utero exposure to 17-OHPC and any cancer in the offspring. Fully conditional specification relaxes the assumptions of joint multivariate normality and linearity and is well-suited for the imputation of both categorical and continuous variables.
The institutional review board at the Public Health Institute and the University of Texas Health Science Center at Houston approved this study. All the analyses were conducted in SAS version 9.4 (SAS Institute, Cary, NC).
Table 1 shows the characteristics of 18,751 offspring. Most (48.2%) were born in the early 1960s. Approximately one-fourth (n=4332, 23.1%) were non-Hispanic Black, and half (52.1%) were in families with an annual income less than the median. The median follow-up was 49.5 years (interquartile range [IQR], 25.5–53.5 years).
|Demographic||In utero exposure to 17-OHPC (n=234)||Not exposed to 17-OHPC (n=18,517)|
|Year of birth|
|1959 – 61||98||41.9||5505||29.7|
|1962 – 64||116||49.6||8929||48.2|
|1965 – 67||20||8.6||4083||22.1|
|Gestational age (wk)|
|2500 – 3999||200||85.5||15,847||85.6|
|Maternal characteristics b|
|Maternal age at pregnancy (y)|
|20 – 24||61||26.2||5587||30.5|
|25 – 29||63||27.0||5317||29.0|
|30 – 34||59||25.3||3257||17.8|
|35 – 39||29||12.5||1895||10.3|
|Parity at pregnancy|
|Body mass index (kg/m 2 ) c|
|Underweight or normal (<25)||167||79.2||12,056||75.2|
|Overweight (25 – 29.9)||35||16.6||2970||18.5|
|Less than high school||34||16.0||2865||18.2|
|High school or trade school||82||38.5||6121||38.8|
|Some college or college degree||97||45.5||6796||43.1|
|Annual family income d|
Approximately 1.0% of the offspring (n=234) were exposed in utero to 17-OHPC. 17-OHPC was most commonly indicated for threatened abortion (41.0%); the first 17-OHPC injection occurred at a mean of 12 weeks’ gestation (median: 10 weeks; IQR, 7–15 weeks), and there was a mean of 2.4 injections (median: 1 injection; IQR, 1–2 injections). Most (n=165 of 234, 70.5%) of the offspring were first exposed in the first trimester. There was no difference in the median follow-up between the offspring exposed (50.5 years) and those not exposed (49.5 years) to 17-OHPC.
Table 2 shows the types of cancers (n=1008) diagnosed in the offspring by sex and the gestational day of the first exposure to 17-OHPC. Among the exposed offsprings (n=234), 23 were diagnosed with cancer, including 2 diagnoses in childhood (age <18 years) and 21 in adulthood (age ≥18 years). The cancer types included the following: melanoma (n=2), lymphoma (n=2), leukemia (n=1), polycythemia vera (n=1), colon and rectum (n=3), prostate (n=3), brain (n=2, both in childhood), breast (n=2), thyroid (n=1), oral cavity (n=1), lung and pleura (n=1), cervix (n=1), uterus (n=1), kidney (n=1), and testis (n=1). The median age at diagnosis was similar for the offspring exposed (45 years, IQR, 37–51 years) and those not exposed (45 years, IQR, 34–51 years) to 17-OHPC.
|Cancer type||Male offspring (n=391)||Female offspring (n=617)|
|Not exposed||Early pregnancy||Late pregnancy||Not exposed||Early pregnancy||Late pregnancy|
|Oral cavity||18||1 (115)||4|
|Colon and rectum||30||2 (34, 71)||35||1 (46)|
|Nose or nasal cavity||3||3|
|Lung and pleura||15||1 (50)||24|
|Bone and joint||3||3|
|Melanoma or other nonepithelial skin||57||45||2 (38, 53)|
|Breast||—||—||—||197||2 (35, 67)|
|Vagina or vulva||—||—||—||9|
|Prostate||53||2 (45, 77)||1 (236)||—||—||—|
|Brain||21||1 (62)||10||1 (76)|
|Central nervous system||7||19|
|Thyroid or other endocrine||8||34||1 (96)|
|Lymphoma||29||1 (114)||22||1 (73)|
|Miscellaneous b||7||1 (149)||8|