When questions arise about tighter government controls on mifepristone, abortion-pill advocates routinely point to studies reporting that medication abortion is highly effective and that serious complications are “extremely rare.” But a closer look at one widely cited study from the University of California, San Francisco (UCSF) raises serious questions about how those conclusions are reached and how valid they are.
The study, “Effectiveness and safety of telehealth medication abortion in the USA,” by Ushma Upadhyay and UCSF colleagues, appeared in Nature Medicine in 2024. It was cited as recently as July 3, 2026, by journalist Julie Ivanova of telehealth.org as proof that “studies have generally supported the safety of medication abortion provided through telehealth and at-home use.”
Upadhyay’s latest study examined 6,034 telehealth abortions managed by three virtual clinics serving women in 20 states and the District of Columbia between April 2021 and January 2022. The researchers reported that 97.7% of abortions were complete without known intervention or continuing pregnancy, while 99.8% were not followed by a “serious adverse event.”
SUPPORT LIFENEWS! If you want to help fight abortion, please donate to LifeNews.com!
Only 0.25% were classified as serious abortion-related adverse events, and 1.3% resulted in an emergency-department visit.
Those figures sound reassuring. The methodology, however, deserves scrutiny.
Significantly, the study did not follow all 6,034 women. Researchers had complete follow-up information for only 4,454. They had no follow-up contact with 1,421 women and did not know the outcome of another 159.
Rather than determining what happened to those women, the researchers effectively assumed their outcomes resembled those of the women who reported back. That assumption is particularly problematic for a telehealth model, where patients may have little or no ongoing relationship with the prescribing clinic and may more conveniently and comfortably seek care locally if something goes wrong.
The missing data could contain uneventful cases—or complications and failed abortions that were never reported to the virtual clinics. Without knowing outcomes for sure, researchers cannot confidently establish the true complication or success rates for the entire group.
The potential impact can be illustrated with a hypothetical example. If 100 of 200 patients reported back and 98 successfully aborted, researchers could, under this methodology, report a 98% success rate among known cases and then project for the rest. But if several women among the 100 lost to follow-up subsequently required emergency treatment or surgical procedures to complete failed abortions, the rate for the entire group would be substantially lower.
If, of the other 100 who did not report back, 21 showed up at the ER hemorrhaging or with an infection, and another 38 went to the ER to complete what they said was a stalled miscarriage, this would yield a far lower overall efficacy rate of 80% and a serious complication rate of 11%. These are numbers closer to what large national studies in Canada, Finland, and independent analyses in the U.S. have found.
Screenings and reported results also relied heavily on patient reports rather than objective medical confirmation. Gestational age—the number of days from a woman’s last menstrual period (LMP) —and abortion completion was generally not established through physical examinations, ultrasound, or laboratory testing, but patient self-reports. Some women supplied pregnancy-test or other medical information, but others simply reported that the abortion was complete without any medical confirmation. Such reports can provide useful information, but they cannot establish outcomes as reliably as objective medical evidence.
Another reason for concern involves the definition of “serious” complications. The study reported a serious adverse-event rate of just 0.25%, yet 2.3% of abortions were described as “not initially complete” and required additional medication or surgical aspiration. Nearly 1% (0.94%) involved confirmed or suspected continuing pregnancy, and 1.3% of patients with known outcomes visited an emergency department.
Those figures are all substantially higher than the reported serious-complication rate, indicating that many treatment failures, emergency visits, and continuing pregnancies were simply classified as not “serious.”
This is not an isolated methodological concern. In an earlier 2015 study, Upadhyay reported complications in 5.19% of chemical-abortion patients but classified only 0.31% as serious. Noticeably she excluded some hemorrhages, infections, and abortion failures requiring aspiration because they did not meet the study’s more restrictive criteria.
The larger issue is not that every woman lost to follow-up necessarily experienced a complication, nor that every patient self-report is inaccurate. It is that researchers cannot know the true outcome of patients they failed to track who may have sought treatment elsewhere. Yet the resulting study is frequently cited as evidence that telehealth medication abortion is exceptionally safe and effective.
That limitation matters for a drug increasingly distributed without an in-person examination. A safety study is only as reliable as its ability to identify adverse outcomes—including those occurring outside the system that dispensed the drug.
The UCSF study may provide some useful information about women whose outcomes were known. But its unusually high safety and efficacy claims should be interpreted cautiously when more than one-quarter of the included patients had no follow-up, many outcomes were based on medically unverified self-reports, and the definition of “serious” complications excluded a substantial number of significant adverse outcomes.
LifeNews.com Note: Randall O’Bannon, Ph.D., is the director of education and research for the National Right to Life Committee.
