This story was originally published by ProPublica.
By Amy Yurkanin, ProPublica
When Dr. Craig Towers moved to Knoxville, Tennessee, in 2010, the high-risk obstetrician encountered a crisis unlike anything he’d seen.
Towers, who had practiced for 25 years in California and Wisconsin, found himself caring for an unusual number of pregnant women addicted to painkillers — and delivering newborns who were suffering withdrawal symptoms. Those babies, born shaking and inconsolable and who required weeks of specialized care, were becoming tragic symbols of the region’s raging opioid epidemic. “We were inundated,” Towers later told the Knoxville News Sentinel.
In those initial years in Tennessee, Towers recognized a need — and a research opportunity.
Medical doctrine had long held that a pregnant patient with an opioid addiction could not be safely detoxed because withdrawal could harm the fetus, causing physical stress and even triggering miscarriage. Instead, doctors treated patients with methadone or buprenorphine, which are legal replacement opiates that control cravings and have a lower risk of overdose. Towers, who was hired to work at the University of Tennessee Medical Center’s clinic for high-risk pregnancies and later joined the medical school faculty, questioned that approach. He began studying the outcomes of women, most of them his patients, who stopped taking those medications. He ultimately collected data on more than 300 of them.
Towers’ research found that detoxing was much less harmful to the fetus than feared, and it prompted changes to how doctors treat pregnant opioid users. Across Tennessee — in Johnson City, Jellico and Sevierville — and in Daytona Beach, Florida, clinics followed his lead. The News Sentinel honored him as a “Health Care Hero.” CNN’s chief medical correspondent, Dr. Sanjay Gupta, interviewed him about his work and described babies born without withdrawal symptoms as “miracles.” In 2019, the medical school granted a request for additional funding for Towers’ work, praising “his many research efforts and leadership that has resulted in national recognition.”
And then in 2020, soon after the study was published, Towers abruptly stepped back from his research.
The study immediately caught the attention of specialists in addiction treatment. Some of them wondered how the results could be true. The study, the largest of its kind, enlisted 230 participants, 121 of whom took a daily dose of a drug called naltrexone, which, unlike methadone or buprenorphine, is not an opioid; instead it discourages using by blocking the high. Naltrexone was effective in keeping most patients off of opiates, the study found.
But in many other major studies, naltrexone had shown lackluster results. Rates of relapse among nonpregnant patients averaged 72%, according to a systematic review of 13 studies.
Eight specialists in the fields of addiction and pregnancy started sharing concerns with one another about Towers’ study and then drafted a letter to the University of Tennessee. Two of them agreed to speak with ProPublica. One was Dr. Mishka Terplan, a practicing OB-GYN board-certified in addiction medicine who’d served on several committees to develop federal and professional guidelines for treating opioid use during pregnancy.
“Dr Towers’ work has been impactful, especially in TN,” Terplan wrote to the university in July 2020, “and I am concerned that each day the article(s) is out there, potential harm to a vulnerable population accumulates through both ethically suspect clinical care and consequently misguided public health.”
By that time, a university audit had uncovered major problems with Towers’ research. The audit found that Towers failed to get university approval for the study, that he provided “inaccurate and/or unverifiable, contradictory source documentation” for a component of the study and that he did not provide research databases or data analysis as requested. Auditors also were unable to determine the extent to which he protected confidential patient health information.
For three years, the journal did nothing to publicly address those concerns. When the journal did act, in 2023, it published a 78-word correction stating that the study had been improperly listed as a “prospective,” a rigorous type of research that follows patients in real time.
ProPublica shared Towers’ study, the university’s audit and the journal’s correction with five experts on research ethics and substance abuse during pregnancy. All of them said that the documents raised major concerns about Towers’ conduct as a researcher and that the journal’s response didn’t do enough to address the issues the university unearthed.
The publisher of the American Journal of Obstetrics and Gynecology, the journal where the study appeared, disputes that. “We uphold the highest standards of rigor and ethics in our publishing to protect the quality and integrity of research,” Elsevier, the company that publishes AJOG, said in a statement.
Towers told ProPublica that the audit had reached erroneous conclusions, that the auditors did not give him the opportunity to fully defend his work and that the journal’s willingness to stand by the study is proof that his research is sound.
One of the experts ProPublica spoke to, Eugenie Reich, an attorney who has represented whistleblowers in several high-profile research fraud cases, said that the problems the university identified in Towers’ study were much deeper than the journal acknowledged. “The misrepresentation is not minor,” Reich said. “It runs throughout the piece.”
In 2018, Katie Carringer was working as an administrative assistant at a drug rehab facility and made an appointment with Towers at the encouragement of her boss. Carringer, who was about six weeks pregnant, had been addicted to heroin in the past but was not using drugs at the time. She told ProPublica she felt confident in her ability to remain drug-free and made the appointment with Towers because she wanted a doctor who specialized in high-risk pregnancies and could help her manage anxiety and depression.
At her second appointment, she expected a conversation about her mental health. She told ProPublica she was surprised when Towers recommended naltrexone. From her work at the clinic, she knew that the drug was meant for people battling cravings in the early stages of recovery. “With me being clean so long, it didn’t even make sense,” Carringer said.
According to her medical records, she had not used drugs during either of her previous pregnancies, in 2007 and 2009. Two years before her appointment with Towers, in 2016, she completed an inpatient drug treatment program, and she was “actually drug free for most of the time” since then, according to her medical notes. Her records showed two short relapses.
With Carringer’s permission, ProPublica asked Towers to review her medical file and answer questions about why he recommended naltrexone. Towers, who said Carringer was not included in the study, noted that her medical record stated she was interested in naltrexone.
Carringer denied to ProPublica that she expressed interest in the drug, adding that she ended up trying naltrexone because she felt it was what Towers wanted her to do. She said, and her records confirm, that it caused severe nausea and she stopped taking it after the first dose.
Terplan said Carringer’s experience points to the salience of a question the letter posed: Did Towers explain to his patients that curbing addiction with naltrexone was an experimental treatment in pregnant women? According to their letter, the specialists were concerned that “informed consent was not obtained or not properly obtained” from patients who showed up in the naltrexone research: “This uncertainty leads us to question whether ethical standards that promote and ensure respect for human subjects and protect their health and rights were followed,” the letter said.
“You need to explain to people that there may be additional risk,” said Columbia University professor Dr. Robert Klitzman, an expert in bioethics. “That’s why informed consent is so important.”
Towers said that concerns about consent and explanation of risk are baseless. All study participants received a “notice of research” and chose to go forward, he said. “I am not aware of any patient that felt pressured to participate,” he wrote. “I always presented the patients with the risks and benefits of each course of treatment so that they could make an informed decision.”
University of North Carolina researcher Hendrée Jones, one of the specialists who drafted and signed the letter, told ProPublica that she was surprised by another aspect of Towers’ study: that it included reports of only minor health problems among the mothers, including nausea and headache, during the course of their pregnancies. Jones said it is unusual for there to be no serious medical events among that many pregnant patients, especially ones who’d struggled with opioid addiction. “That made my hair on the back of my neck stand up,” Jones said.
In its audit, the university also had flagged problems with Towers’ data. The audit said Towers supplied a printout of his entire clinical practice database, but it did not designate which patients were involved in the research study. Towers told ProPublica that he obtained broad consent for his research: “Every patient that was seen in the clinic received a notice regarding the research, consented to participation in that research and understood that their data would be prospectively collected but deidentified.”
Klitzman said researchers should be able to provide details about the actual patients in a study, not just everyone at a clinic. “That to me is a red flag,” he said.
Researchers are required to run all proposed studies involving human participants through institutional review boards — panels of scientists and other compliance specialists who ensure patients’ rights and confidentiality are protected. In its audit, the university determined that Towers hadn’t sought IRB approval for the study. Towers told ProPublica that the university was mistaken and that the journal itself had “concluded,” contrary to the audit, that he did have IRB approval. A spokesperson from Elsevier did not respond to repeated questions about whether the journal had concluded that the IRB had approved the study.
In February 2020, about two weeks after the audit got underway, Towers retired from his faculty position. Months later, he retired from clinical practice as well. In his email to ProPublica, he said he stepped down for health and family reasons.
According to a determination letter from the chair of the IRB dated days after the audit’s completion, members of the board voted to suspend all of Towers’ open studies for safety reviews and file a report with the federal Office for Human Research Protections. The audit’s findings, the letter said, constituted “serious noncompliance.” Towers said he disputes that finding and said the board did not make an effort to get his side of the story.
A spokesperson for the University of Tennessee Health Science Center said she could not find anyone to answer questions from ProPublica “due to time and a change in leadership.”
Soon after Towers resigned from his faculty position, he reached out to the journal and sent IRB approval letters that he said confirmed the study had been approved. The audit, though, had found that the IRB approvals he cited did not apply to the study.
The University of Tennessee also contacted the journal, requesting the paper be retracted — but as of ProPublica’s reporting, no action had been taken by the journal. Less than 1 in 2,500 published studies are retracted, according to the journal Science. Ivan Oransky, co-founder of Retraction Watch, said he has encountered other cases where journals failed to retract a study after a university made such a request. “You have a situation where the fox is guarding the henhouse,” he said.
During the years that the journal spent considering how to respond to the university’s request for a retraction, more doctors across the country began prescribing naltrexone. Researchers cited the work dozens of times in other studies.
Towers said he stands by naltrexone as a treatment option: “I believe it is crucial in medicine to revisit old assumptions and consider all options.”
Dr. Geogy Thomas, medical director of Dayspring Health in east Tennessee, said he was inspired by Towers’ research, which improved care for women in the area. “Our community is indebted to Dr. Towers and his research on substance abuse in pregnancy,” Thomas wrote in an emailed statement.
But another obstetrician, Dr. Cathleen Suto, who worked at a clinic with an approach inspired by Towers’ research, found detox wasn’t effective. Many of her patients couldn’t make it through the weaning process and often stopped coming to the clinic. “Detox would be a great answer,” Suto has said. “The problem is that my patients were not successful at doing it.”
When the journal finally acted on the university’s concerns in 2023, more than three years after the study came out, it did not retract the study. Instead, it published a brief correction: “The authors wish to clarify that the study is listed as a prospective cohort study — which is not the case. We did a retrospective analysis of prospectively collected data.”
Towers views the journal’s actions as vindication. “AJOG would not keep a study in print if they did not believe it was IRB-approved after their review,” Towers said.
Jones, one of the letter’s signatories, said the journal failed in its duty to hold its research to a higher standard. “When you make a mistake, you own up to that mistake and you fix it,” she said. “And that paper is out there, and continues to be cited, and I see it.”
Sign up for Dispatches, a ProPublica newsletter that spotlights wrongdoing around the country, to receive stories like this in your inbox every week.