Meta-analysis launders safety data from old pharma trials; blames patients for drug harms

Comment on: Zhang JJ, et al. Sexual, physical, and overall adverse effects in patients treated with 5α-reductase inhibitors: a systematic review and meta-analysis. Asian J Androl. 2022. doi:10.4103/aja202171 • PubMed

Key points

  • In a 2022 meta-analysis by Zhang et al., 26 of 28 included studies were sponsored by Merck or GlaxoSmithKline, making up more than 99% of the weighted data.
  • Merck-led studies have inadequate and selective safety data; most are more than 20 years old.
  • Authors use on-drug trial data as a proxy for a post-drug syndrome—a fatal flaw.
  • The authors attribute harms of finasteride and dutasteride to patients’ pre-existing conditions and irrational beliefs.

A specious premise

In a meta-analysis, Zhang et al.1 pooled data from dozens of clinical trials in an attempt to gain insight on symptoms of post-finasteride syndrome (PFS). Even though the syndrome occurs after discontinuation of the drug, the meta-analysis draws on trials of men who were taking finasteride or dutasteride. Post-drug experiences are not interchangeable with on-drug experiences. Although the authors acknowledge this limitation, they forge ahead anyway, speciously calling adverse events from trials “PFS-like.” Zhang et al. even tout this preposterous term as an original contribution: “To our knowledge, this is the first attempt to pool current evidence from placebo-controlled randomized clinical trials (RCTs) to evaluate PFS-like adverse effects.”

Laundering of pharma-sponsored trial data

The authors declared no competing interests in the paper. Although this superficially suggests freedom from industry influence, 26 of 28 included studies were sponsored by Merck or GlaxoSmithKline (GSK), the makers of finasteride and dutasteride, respectively (Table 1). These studies represented 99.6% of the weighted data. (The author of a small study without a declaration, D. Stough, had previously participated in studies of finasteride and dutasteride funded by Merck and GSK respectively.) The meta-analysis does not acknowledge the potential for industry bias, even though it has been confirmed to influence trial results.2

Table 1. The first column has been added to indicate study sponsorship. Of 28 included studies, all but 2 were funded by Merck or GSK or involved their researchers. Two studies whose funding was not confirmed carried a combined weight of 0.4% of the overall analysis (see Figure 2 in Zhang et al.). The author of one of these, Stough, had previously participated in studies of these drugs funded by Merck and GSK.

In Propecia trials, Merck co-opted academic researchers, conferences and journals to promote drugs

To understand the bias embedded in the meta-analysis, we will take a closer look at Merck’s trials which make up the majority of included studies. Merck’s strategy for Propecia resembled that of Vioxx, which came under scrutiny after tens of thousands of deaths linked to the drug.3 Merck recruited academic researchers to study groups, paid them consulting and lecture fees, offered co-authorship of papers, tilted safety data to support commercial interests, and presented biased results in medical conferences and influential journals to increase the drug’s profile.

From its experience developing Proscar, Merck knew that sexual adverse effects of Propecia could raise concerns of regulators, physicians and patients. The company had the opportunity to influence safety data in numerous ways. For example, they could reduce the sensitivity of sexual function questionnaires, pushing AE rates towards a floor so that differences between treatment and placebo groups were reduced. Investigators, not patients, determined whether an AE was drug-related. Basch has pointed out: “Current methods for detecting adverse events in clinical trials are acknowledged to lack sensitivity, and concerning symptoms might well come to light earlier in the drug-development cycle if patient reporting were standard practice.”4 Social stigma and embarassment over sexual adverse events (SAEs) could reduce patients’ willingness to report them, and might prevent dropouts from mentioning SAEs as a reason for discontinuing the trial.

Figure 1, an annotated excerpt of a paper on a finasteride trial, highlights several of Merck’s tactics. One was to co-opt academic researchers through authorship credits and payments. They were appointed to groups such as the Finasteride Male Pattern Hair Loss Study Group. To give the drug a global reach, trials were conducted in North America, Europe and Asia. Resulting studies were published in influential journals such as The New England Journal of Medicine, Urology, Journal of the American Academy of Dermatology and European Journal of Dermatology. Merck thus used its massive financial resources to buy allies and create an evidence base favorable to its business goals.

Figure 1. An excerpt of a finasteride study5 showing Merck’s tactics of co-opting research institutions, physicians and the dermatology specialty to increase the drug’s profile. The company paid academic researchers to join as co-authors, presented results at specialty conferences and published them in leading specialty journals, thereby increasing the drug’s profile among prescribers. The trials thus formed the leading edge of a marketing strategy.

Opaque methodologies and selective reporting of safety concerns

The safety sections of published studies are typically perfunctory, as shown in Figure 2.

Figure 2. This paragraph is the entire safety section of Leyden, 1999.4 The study does not specify how sexual adverse events were assessed. The highlighted phrase (“for most patients”) directs attention away from the bad news that some patients did not see resolution of sexual adverse events (SAEs) during the study. The number of these patients is not provided, and there is no indication whether the SAEs resolved after the study ended.

In contrast with detailed information about efficacy in Leyden 1999,4 safety reporting is limited and leaves questions unanswered. No information is provided on how adverse events were assessed. If they were assessed in an interview with an investigator, stigma and embarrassment could suppress reports of SAEs.6 Study participants might also understate AEs because of hopes that a new drug will help them, and a desire to comply with expectations of the sponsor.

The number of patients who did not see resolution of SAEs during the trial is not provided. Nor is there follow-up information about whether SAEs resolved after the study ended. This study provides a representative example of Merck’s lackluster, slanted approach to safety reporting. Since the Zhang meta-analysis pools studies with limited and biased data, it reproduces the biases in the underlying studies, while conferring the superficial authority of a meta-analysis. Given varying safety assessment methods across 28 studies, pooled risk ratios are diluted by less sensitive studies. This will tend to understate risks in a meta-analysis.

Additional limitations

The meta-analysis is subject to additional limitations of included studies.

Prior determination of AEs to be assessed. Safety data in trials is limited by the drug sponsor’s a priori decisions for what adverse events to measure. In Zhang et al., relative risk of neuropsychiatric events could not be calculated because only one study counted this class of AEs. More than 20 years after approval of Propecia, a study identified a signal for psychological adverse events associated with finasteride, including suicidality.7 In 2022 FDA required the drug sponsor to add suicidal ideation and behavior to the postmarketing events section of the label.8

Limited characterization of adverse events. Information is lacking on the duration and resolution of adverse events in trials. The severity of adverse events is determined by investigators, who may discount severity. Seruga et al. noted:

Increasing evidence shows that, compared with patients, clinicians (physicians and nurses) under-detect and under-report symptomatic adverse events of anticancer drugs, both in everyday clinical practice and clinical trials […]. Several reasons might explain this effect: insufficient time during patients’ visits to fully discuss symptoms, under-reporting of symptoms by patients because of their desire to remain on therapy, and downgrading of symptoms by clinicians to justify continuation of treatment.9

Selective recruitment of participants. Belknap et al.10 pointed out that participants in finasteride trials were selected to be generally healthier than the broader population who will eventually use the drug. This practice tends to understate safety concerns for a general population.

Short duration of trials. The median length of included studies was 36 weeks, or just over 8 months. Many young men taking finasteride for hair loss use the drug for years. Kiguradze et al. found that among young men, risk of persistent erectile dysfunction was associated with longer duration using finasteride.11 To the extent that risk of AEs from finasteride or dutasteride increases over time, studies of shorter duration will understate safety concerns.

Authors blame patients for lasting harms of finasteride and dutasteride

Despite the numerous weaknesses of the meta-analysis, the authors nevertheless take the opportunity to blame lasting dysfunctions after 5-ARIs on patients’ shortcomings. They imply such dysfunctions are due to pre-existing “mental health issues,” sexual dysfunction, “inquiries on safety issues” and an irrational “belief system.”

To date, there are no predictive factors for the risk of developing PFS. Nevertheless, it would seem appropriate to evaluate the history of preexisting mental health issues and the possibility of potential sexual dysfunction. Additionally, a stricter selection of patients is needed before the initiation of 5ARI treatment, such as patients with repeated inquiries on safety issues of the drug or holding on to their belief system tenaciously despite any rational argumentation against it. Administration of 5ARIs may more easily put these patients at an increased risk of developing sexual or emotional disorders.1


Despite serious methodological shortcomings—including the preposterous use of on-drug safety data as a proxy for a post-drug condition—the authors nevertheless blame the post-drug syndrome on patient shortcomings. Zhang et al. have yoked bad logic to biased data in order to deflect attention away from drug harms. The article is ill-conceived and blind to industry bias in almost all included trials. It propagates the mistaken view that post-5-ARI dysfunctions are “side effects” which somehow persist after the drug has been metabolized. This is a straw man requiring a supernatural explanation—and therefore beloved by physicians seeking to deny that a drug may have caused irreversible harm to patients. Since this low-quality article passed peer review, reviewers must share the authors’ medical hubris, blindness to industry influence, methodological incompetence, and contempt for patients. Peer review has yielded a science-like product which rubber-stamps the status quo.


1. Zhang JJ, Shi X, Wu T, et al. Sexual, physical, and overall adverse effects in patients treated with 5α-reductase inhibitors: a systematic review and meta-analysis. Asian J Androl. 2022. doi:10.4103/aja202171 • PubMed

2. Ahn R, Woodbridge A, Abraham A, et al. Financial ties of principal investigators and randomized controlled trial outcomes: cross sectional study. BMJ. 2017. doi:10.1136/bmj.i6770PubMed

3. Grifo F. Vioxx shows what happens when the drug safety system breaks down. The Equation blog hosted by Union of Concerned Scientists. March 19, 2012.

4. Union of Concerned Scientists. Merck manipulated the science about the drug Vioxx. Union of Concerned Scientists blog. October 12, 2017.

5. Basch E. The missing voice of patients in drug-safety reporting. N Engl J Med. 2010. doi:10.1056/NEJMp0911494PubMed

6. Leyden J, Dunlap F, Miller B, et al. Finasteride in the treatment of men with frontal male pattern hair loss. J Am Acad Dermatol. 1999 Jun;40(6 Pt 1):930–7. doi:10.1016/s0190-9622(99)70081-2 • PubMed

7. Moore TJ. Finasteride and the uncertainties of establishing harms. JAMA Dermatol. 2015 Jun. doi:10.1001/jamadermatol.2015.37 • PubMed

8. Nguyen DD, Marchese M, Cone EB, et al. Investigation of suicidality and psychological adverse events in patients treated with finasteride. JAMA Dermatol. 2020 Nov 11. doi:10.1001/jamadermatol.2020.3385 • PubMed

9. U. S. Food and Drug Administration. Final Response Letter from FDA CDER to Post-Finasteride Syndrome Foundation. June 8, 2022. Accessed August 18, 2022.

10. Seruga B, Templeton AJ, Badillo FE, Ocana A, Amir E, Tannock IF. Under-reporting of harm in clinical trials. Lancet Oncol. 2016 May. doi:10.1016/S1470-2045(16)00152-2 • PubMed

11. Belknap SM, Aslam I, Kiguradze T, et al. Adverse Event Reporting in Clinical Trials of Finasteride for Androgenic Alopecia: A Meta-analysis. JAMA Dermatol. 2015. doi:10.1001/jamadermatol.2015.36 • PubMed

12. Kiguradze T, Temps WH, Yarnold PR, et al. Persistent erectile dysfunction in men exposed to the 5α-reductase inhibitors, finasteride, or dutasteride. PeerJ. 2017. doi:10.7717/peerj.3020 • PubMed

Further reading

Context matters: a rebuttal to yet another analysis of adverse events of finasteride. Finasteride Watch. Published June 3, 2022.

Elliott C. White Coat, Black Hat: Adventures on the Dark Side of Medicine. Beacon Press: 2011. (Amazon)

Grifo F. Vioxx shows what happens when the drug safety system breaks down. The Equation, a blog from Union of Concerned Scientists. Published March 19, 2012.

Responding to a literature of doubt: limitations of three studies of adverse events of finasteride and dutasteride. Finasteride Watch. Published February 24, 2022.

Stegenga J. Is meta-analysis the platinum standard of evidence? Stud Hist Philos Biol Biomed Sci. 2011. doi:10.1016/j.shpsc.2011.07.003PubMed