A study can acquire authority long before the documents beneath it receive the same attention. That is the pattern ProPublica reconstructed around Dr. Craig Towers, a high-risk obstetrician whose work on opioid use disorder in pregnancy received national and local recognition. The reporting also documents Towers’ rejection of the audit’s conclusions, an essential part of the record.
For years, standard practice generally involved stabilizing pregnant patients with opioid use disorder on methadone or buprenorphine. Doctors were concerned that detoxification could endanger the fetus and that a subsequent return to opioid use could expose the patient to serious risks. Towers questioned whether detoxification followed by naltrexone maintenance could offer another option for some patients.
His study appeared in the January 2020 issue of the American Journal of Obstetrics and Gynecology. It included 230 patients, with 121 receiving naltrexone and 109 receiving methadone or buprenorphine. The paper reported markedly lower neonatal abstinence syndrome among infants in the naltrexone group and concluded that naltrexone could be a viable option for some pregnant patients who had chosen to detoxify.
The findings attracted attention because they challenged longstanding caution within obstetric and addiction medicine. CNN chief medical correspondent Dr. Sanjay Gupta interviewed Towers, the Knoxville News Sentinel recognized his work, and Towers discussed the findings at a university science forum. ProPublica reported that clinics in several Tennessee communities and in Daytona Beach, Florida, later adopted approaches influenced by his work.
Behind that public recognition was a University of Tennessee audit signed in February 2020. According to ProPublica’s account of the audit, reviewers concluded that Towers had not obtained university approval covering the study and that documentation for one component was inaccurate, contradictory, or impossible to verify. The audit also said the requested research databases and data analysis were not provided.
Towers did provide auditors with a printout from his wider clinical-practice database. However, the audit said that printout did not identify which patients were included in the research. Auditors also said they could not determine the extent to which confidential patient health information had been protected, which is narrower than concluding that no protections existed.
Eight specialists in addiction and pregnancy subsequently sent a letter to the university describing their concerns. One was Dr. Mishka Terplan, a practicing OB-GYN who is board-certified in addiction medicine and has helped develop professional and federal guidance on opioid use during pregnancy. The letter raised questions about study design, participant selection, informed consent, and the possible effect of the paper on a vulnerable patient population.
A separate determination letter from the university’s institutional review board described the audit findings as “serious noncompliance.” According to ProPublica, the board voted to suspend Towers’ open studies for safety reviews and to report the matter to the federal Office for Human Research Protections. The university also asked the journal to retract the paper.
Towers disputed those conclusions. He told ProPublica that the audit was mistaken, that the approvals he supplied covered his work, and that he was not given an adequate opportunity to present his position. He also said he retired from his faculty and clinical roles for health and family reasons, and that the journal’s decision to leave the study in print supported his view that the research remained sound.
The journal did not retract the study. In March 2023, it issued a 78-word correction clarifying that the work was not a prospective cohort study, as the paper had described it, but a retrospective analysis of prospectively collected data. The correction did not publicly address the university’s approval, database, confidentiality, or informed-consent findings.
ProPublica provided the study, audit, and correction to five experts in research ethics and substance use during pregnancy. All five said the documents raised serious concerns and that the journal’s response did not adequately address the audit findings. Eugenie Reich, an attorney who has represented whistleblowers in research-fraud cases, said the methodological misrepresentation extended throughout the paper.
Elsevier, which publishes the American Journal of Obstetrics and Gynecology, defended the decision. The company said the journal concluded that a correction was the appropriate response based on the information available at the time. That position belongs in the record alongside the university’s request for retraction and the specialists’ objections.
The episode exposes a practical gap in scientific correction. Audits, institutional-review-board letters, journal notices, and media profiles live in different places and reach different audiences. A clinician who encounters the original paper may not automatically encounter the university audit or the correspondence requesting retraction.
Older publicity also remains searchable after a study is challenged. Television segments, local awards, university appearances, and the paper itself do not automatically acquire warnings when new information emerges. Unless those records are clearly connected, the original prestige can continue working while the correction remains difficult to find.
The study’s influence was not merely theoretical. ProPublica reported that clinics in Johnson City, Jellico, and Sevierville in Tennessee, as well as Daytona Beach in Florida, followed approaches associated with Towers’ work. It also reported that the paper continued to be cited by other researchers after the university had raised its concerns.
The clinical evidence surrounding naltrexone adds another layer of caution. A Cochrane systematic review identified 13 randomized studies involving 1,158 nonpregnant outpatients who received oral naltrexone after detoxification. It found low treatment retention and concluded that oral naltrexone had not been scientifically shown to be superior to the alternatives examined.
That review does not directly determine whether naltrexone is appropriate during pregnancy, nor does it prove that Towers’ conclusions were wrong. It does show why unusually positive findings in a vulnerable population required especially clear documentation and scrutiny. The university audit’s reported inability to obtain the requested research database and analysis therefore matters independently of the ultimate clinical answer.
The central failure is not that disagreement occurred. Scientific disagreement is expected, and a contested audit is not the same thing as a final adjudication of research misconduct. The problem is that the disagreement, the audit, the university’s retraction request, and the journal’s limited correction were not presented together in the public publication record.
That distinction also matters for fairness. Towers continues to defend the research, while the university’s audit, eight specialists, and five experts consulted by ProPublica raised substantial concerns. Readers should see both positions, along with the documentary findings that produced the dispute.
Terplan warned the university about potential harm in July 2020. Six years later, the original paper remains available with a 78-word correction, while the audit and retraction request remain separate from the article most clinicians would encounter. That distance between the celebrated finding and the documents challenging it is the real story.
