The Merck files, part 1: An inside look at the development of Propecia

In this series: Overview • Part 1Part 2Part 3

Merck first developed finasteride for benign prostatic hyperplasia, a condition typically found in men aged 55 and older. After receiving approval for the application in 1992 and bringing the drug to market as Proscar, Merck decided to seek approval of finasteride for male pattern hair loss—to be branded as Propecia. In a 1994 memo below, Merck’s Worldwide Marketing group set forth its “needs” regarding the development of Propecia. In November 1997, the New Drug Application (NDA) was submitted to FDA and approved the next month (not included below but available here). A one-year trial included in the NDA was extended for four additional years to assess long-term efficacy and safety. In 2001, Merck submitted a Supplemental NDA which included results from men who continued to take Propecia for five years. The Supplemental NDA is the last in this subset of documents.

In later documents, Merck acknowledged that a man who discontinued the trial due to a drug-related sexual adverse event (AE) did not experience resolution of the AE 6 months after discontinuation. In documents below, Merck explores different ways of reflecting this outcome, clearly aware that it would be unfavorable for approval and marketing purposes. Ultimately, in the official submission to FDA, Merck failed to mention this outcome, concluding that the 5-year trial “corroborat[es] the excellent safety and tolerability profile of the drug in this patient population.” Across all these documents, a pattern emerges:

  • The Marketing group determined the needed outcomes of trials before the trials were run.
  • Trial results were selectively described to regulators and doctors to overstate benefits and understate risks.
  • Merck failed to acknowledge biases in findings about risks of Propecia: dropouts in the long-term trial; participants’ hesitancy to report sexual AEs because of social stigma; and investigator bias in determining which AEs were drug-related.

In excerpts below, emphasis has been added in bold.

Worldwide Marketing Needs Report on Finasteride for Androgenic Alopecia in Men

Aug 29, 1994 • View document (7 pages)

This memo sets out the “needs” of Merck’s drug marketing division for the development of finasteride as a treatment for hair loss. They wished to pursue “the most rapid registration possible.” Sales were estimated to reach $580 million in the fifth year after approval.

Given its experience with Proscar, Merck was aware of the potential impact of sexual adverse effects on marketing efforts:

…it is also critical that the label does not include sexual adverse reactions…

The company sought to enlist the support of “opinion leaders”—a tactic also used in the marketing of Vioxx.

Even though the Phase III trials had not been completed, the memo specified that the safety profile will be the same as that of Phase II trials and that “we will be able to state that there was no difference in…the occurrence of sexual side-effects between the placebo and the finasteride treated group…”

…[A] statement similar to the following should be included in the
prescribing information for finasteride:
Finasteride is well tolerated; adverse reactions are mild when they
occur and not any different
from the patients in the placebo group.

Before a trial has taken place, Merck Worldwide Marketing states its “needs” for a trial of the effect of finasteride on semen production in men aged 20–50: “This [trial] will show a minimal effect on decrease of ejaculate volume of semen and that the effect is reversed within a short period of time when treatment is stopped.” (When the trial was eventually published, it was reported that one man’s finasteride-related sexual adverse event did not resolve while in the trial, with no follow-up reported.)

Email correspondence between Keith Kaufman and Bruno Gibelin

Apr 1, 1997 • View document (2 pages)

In this email, Bruno Gibelin asked the head of Propecia clinical development, Keith Kaufman, whether there was a statistically significant difference in sexual AEs between the treatment and placebo groups. Dr. Kaufman replied (emphasis added):

…for each of the sexual AE’s (decreased libido, erectile dysfunction, decreased ejaculate volume), the differences between treatment [and placebo] groups were NOT statistically significant. However, this is irrelevant for labeling purposes since, in general, all drug-related clinical AE’s that occur at an incidence of at least 1% and are greater in active vs placebo wind up in the label… If one summarizes the data differently such that we look at ALL patients who had ANY of sexual AE, then the differences between groups (3.8% for finasteride vs 2.1% for placebo) become statistically significant.

Dr. Gibelin’s reply to Dr. Kaufman shows Merck’s efforts to minimize this adverse effect in its communications with prescribing doctors. Gibelin replied to Kaufman: “Even if [lack of statistical significance between groups] could not be used with the agency [FDA] I’m sure sales forces would like to use it as an argument if the question was raised by the physician. Do you agree?”

Kaufman replies on April 2, 1997 (emphasis added):

This [sexual adverse effects] is almost certainly a real effect of the drug, and is supported both by the PROSCAR experience and by the statistically significant, though numerically small, changes seen in the sexual function questionnaire in the PROPECIA trials. Moreover, if the PROPECIA trials were larger (say 2000 patients/group), then we would have likely reached p<0.05; i.e., this is more of a sample size/statistical power issue than a physiological one. If asked, of course, you can respond that there is no statistically significant difference between treatment group for each of the 3 AE’s.

Propecia 5-Year End-of-Study Phase III Controlled Data

Aug 8, 2000 • View document (30 pages)

Slides in this document summarize a 5-year Phase III clinical trial assessing the long-term efficacy and safety of Propecia. A timeline on p. 3 shows that the marketing application was withdrawn in 6 of 8 countries in the European Union because of their concerns about long-term safety and efficacy.

P. 4 shows that only 41.5% of the Year 1 finasteride group stayed in the trial through the 5th year (323 out of 779). If dropouts had experienced a higher rate of sexual AEs than those continuing in the trial, results in later years would understate the risks of finasteride. The following slide summarizes sexual adverse experiences throughout the 5-year trial:

The percentage of participants in the finasteride group with drug-related sexual AEs went from 4.4% in the 1st year to 0.5% in year 4 and 0.6% in year 5. The last row summarizes dropouts due to sexual AEs. There are two important limitations of this data. First, investigators determined whether an AE was drug-related. Investigators have a conflict of interest: they are employed by Merck and would have understood that drug-related AEs would be unfavorable for drug approval.1 Second, social stigma might discourage the reporting of sexual AEs.2

Study attrition could also overstate long-term efficacy (effectiveness in treating hair loss). Merck concluded (p. 21): “Patient satisfaction with appearance of hair overall increased with continued therapy.” This could merely reflect dropouts by participants who did not see improvements. With no mention of these limiting factors, Merck concluded: “Safety data support excellent long-term tolerability.

P. 24 shows that Austria, Belgium, Greece, Holland, Ireland and Luxembourg declined to approve Propecia because of concerns over long-term safety, long-term efficacy and a negative risk/benefit ratio.

P. 27 shows Merck’s strategic approach to approval in the European Union: “need to decrease risk of repeat MR [mutual recognition] by minimizing influence of outstanding countries.” (Mutual recognition is a process whereby countries may rely on regulatory determinations by other countries.)

Email correspondence between Prita Pillai and Keith Kaufman regarding presentation of new clinical data on Propecia use

Aug 16, 2000 • View document (4 pages)

In this email exchange, Merck Marketing staffer Prita Pillai asked the director of Propecia clinical development, Keith Kaufman, MD, for feedback on an abstract to be submitted to a major European dermatology conference. Kaufman replies:

Several errors, starting with the first sentence… You are also using the ‘no change’ category for global photos as a positive for finasteride and a negative for placebo – definitely not appropriate. And how did we get 99.4% of patients on
finasteride with no sexual adverse experiences?

When a revision was sent the next day, Kaufman replied:

Still misleading, and the safety section is deceptive. The fact that we have lost most of the patients by Year 5 and that most of the patients with sexual AEs have dropped out of the study does not really permit you from making the statement that ‘99.4%’ of patients at Year 5 had not [sic] sexual AEs.

How did Jerry write this? Where did he get the data from?

Pillai’s reply, the last in this exchange, asked: “Is this any better?”

In what appear to be Kaufman’s comments on the abstract, his response to the statement “No patients in the finasteride group had any decrease in hair counts” was “Not true – correct answer is 34%.”

The original draft stated: “The drug was well tolerated with 99.5% of the finasteride group experiencing no sexual adverse effects.” Kaufman commented: “this is totally misleading, as you have weeded out the dropouts with the sexual AEs (N=323).

Email distribution re: Merck Task Force Meeting held 10/23/2000

Oct 20, 2000 • View document (2 pages)

This document is unremarkable except that it shows who attended a meeting to discuss changes to Propecia labeling based on results of the 5-year study.

Redlined Propecia product insert

Oct 23, 2000 • View document (15 pages)

The draft product insert appears to show changes to the US product label requested by Marketing, to be approved by the clinical development team. Certain edits accentuate drug efficacy while downplaying adverse effects. It is suggested to add this sentence on p. 13: “The incidence of each of the above side effects decreased to ≤ 0.3% by the fifth year of treatment with PROPECIA.” In a previous document, Keith Kaufman had pointed out that this low percentage reflected a large number of participants who dropped out with sexual AEs.

Redlined Propecia product insert

Nov 3, 2000 • View document (15 pages)

This draft product insert appears to reflect revisions by the clinical development team, with another round of comments by Marketing. P. 13 shows discussion of a crucial passage regarding outcomes of sexual AEs in trial participants:

In the phrase “Resolution occurred in all men who discontinued therapy,” Marketing proposes to replace “all men” with “many men.” This acknowledges that not all men who dropped out of the trial because of drug-related sexual AEs experienced resolution of those AEs (to be discussed in Part 2 of this series).

Among men who experienced sexual AEs on Propecia and stayed in the trial, 58% experienced resolution of these AEs. This means sexual AEs did not resolve in 42% of men who stayed in the trial. Marketing proposed to replace “58%” with “most men,” a vague term that obscures the proportion of men who had this adverse experience.

Supplemental New Drug Application for Propecia: 5-Year Data

Apr 6, 2001 • View document (132 pages)

This Supplemental NDA submitted to FDA adds data from the four-year extension of the one-year trial submitted with the first NDA. The following shows Merck’s changes from the previously approved label, with deletions struck through and additions underlined (pp. 101–102):

In the phrase “Resolution occurred in all men who discontinued therapy with PROPECIA,” the word “all” is removed. An earlier internal suggestion to replace “all men” with “many men” would have (at least indirectly) acknowledged the participant whose drug-related sexual AEs were unresolved 6 months after discontinuing the trial. Replacing “all men” with “men” fails to acknowledge this case of persistent unresolved drug-related sexual AEs.

As in the draft product insert above, 58% is replaced with “most men.” This vague term obscures the significant percentage, 42%, which did not see resolution of drug-related sexual AEs while in the trial.

Merck added to the label: “The incidence of each of the above side effects decreased to ≤0.3% by the fifth year of treatment with PROPECIA.” There is no mention of the effect of dropouts. Moreover, this refers only to each sexual AE, but does not mention the aggregate percentage which was 0.6% (see slide in August 8, 2000 document above).

The Patient Information sheet includes this text:

These side effects went away in men who stopped taking PROPECIA. They also disappeared in most men who continued taking PROPECIA.

Again, there is no mention of the participant who discontinued the trial due to drug-related sexual AEs, whose sexual AEs were still unresolved 6 months after discontinuation of Propecia.

Despite this concerning outcome, p. 124 states:

Consistent with the experience with finasteride 5 mg, a low incidence of adverse events was observed in the clinical studies with finasteride 1 mg in the treatment of men with male pattern hair loss. Potential pharmacological effects of finasteride 1 mg were also examined, and the results were reassuring, corroborating the excellent safety and tolerability profile of the drug in this patient population.

From p. 125:

4.5 Safety Conclusions

1. In men with male pattern hair loss, finasteride 1 mg/day is generally safe and well tolerated.
2. The safety profile for finasteride-treated patients over 5 years is consistent with the observations on safety presented in the original New Drug Application.

The last page of this document (p. 132) states that the application for Finasteride 1 mg had not been rejected in any country; however, this omits the fact that Merck had withdrawn its initial application to six European countries which had concerns about efficacy, safety and risk-benefit profile: Austria, Belgium, Greece, Holland, Ireland and Luxembourg. Once again, Merck omitted information unfavorable to its business goals.


  1. See a 2010 article by Ethan Basch, The Missing Voice of Patients in Drug-Safety Reporting.
  2. See also a 2016 meta-analysis of Propecia clinical trials by Belknap et al.

In this series: Overview • Part 1Part 2Part 3