Finasteride and the High-Grade Prostate Cancer Scare: A Detection-Bias Lesson

The short answer#

A single number from one large trial made finasteride look dangerous for nearly a decade, and that number was mostly an illusion. In the Prostate Cancer Prevention Trial (PCPT), the drug lowered the overall rate of prostate cancer by about a quarter and cut low-grade disease sharply. At the same time, a larger fraction of the cancers found in finasteride users were labeled high-grade, which is the type that actually shortens lives. Investigators eventually showed that the high-grade excess came largely from finding and grading tumors more accurately in a smaller prostate, not from finasteride making cancers more aggressive. Two decades of survival data confirmed it: the mortality curves for the two groups sat almost on top of each other.

Key points#

• Finasteride reduced total prostate cancer diagnoses in PCPT (18.4 percent versus 24.4 percent), driven mostly by fewer low-grade tumors.
• A higher share of the cancers found in the finasteride arm were high-grade, and that finding, not the benefit, dominated the headlines.
• Finasteride shrinks the prostate by roughly a quarter, which lets a biopsy sample and grade tissue more accurately, inflating the high-grade count without any change in biology.
• Fifteen-year survival was 78.0 percent with finasteride and 78.2 percent with placebo, with no meaningful difference in deaths.
• The episode is a clean teaching case in detection bias: when a treatment changes the organ you are measuring, it can change the measurement.

Two results that seemed to argue with each other#

PCPT was a randomized, placebo-controlled trial run through the NCI-supported trials network, enrolling nearly 19,000 men aged 55 and older. To join, a man needed a normal digital rectal exam and a PSA of 3.0 ng/mL or lower. Participants took finasteride, a 5-alpha reductase inhibitor, or placebo for seven years. Thompson and colleagues reported the primary results in the New England Journal of Medicine in 2003.

On its face, the headline finding was a success. Prostate cancer turned up in 18.4 percent of finasteride users versus 24.4 percent on placebo, a relative reduction near a quarter that later work put closer to 30 percent once measurement differences were accounted for. Most of that benefit landed on low-grade tumors, the ones least likely to ever threaten a man.

Then came the countercurrent. Among tumors graded Gleason 7 or higher, finasteride looked worse: 6.4 percent of that group versus 5.1 percent on placebo. Framed as a proportion of the cancers actually detected, roughly 37 percent in the finasteride arm were high-grade against about 22 percent on placebo. Because aggressive prostate cancer is what causes death, this single comparison eclipsed the larger drop in total cancers. The Food and Drug Administration later added labeling language noting the association, a neutral description of the label rather than a verdict on the drug, and many clinicians set finasteride aside for prevention.

The situation posed one sharp question. Was finasteride creating dangerous tumors, or was it changing how tumors were found and classified? Those answers point in opposite directions, and only one of them should alter what a doctor does.

Why a shrinking gland changes what a biopsy finds#

Several details of the trial favored the measurement explanation over a biological one. Picture searching a room for something small. A needle biopsy samples only a fraction of the prostate, so it is a bit like drawing a few cores from a large volume and hoping to hit anything abnormal. Finasteride reduces prostate size by about a quarter. In a smaller room the same handful of samples covers proportionally more of the space, so an existing tumor is more likely to be reached, and once reached it is more likely to be graded correctly rather than undersampled.

PSA behavior pushed in the same direction. Finasteride lowers PSA, and the protocol corrected for that shift, which sent relatively more finasteride-group men to biopsy in the first place. Each of these effects raises the tally of high-grade cancers found without touching the disease itself.

Pathology re-reviews and modeling studies that followed the main report reinforced this reading. Once analysts adjusted for the better sampling and grading that a smaller gland allows, much of the high-grade excess shrank or vanished. The pattern looked less like a drug that spawns lethal cancer and more like a drug that makes existing cancer easier to see and score. That is detection bias in a textbook form: the instrument, here the biopsy, simply behaved differently in the two arms.

The mortality data that closed the case#

Grade on a biopsy is an intermediate marker. The outcome patients care about is survival, and only time can measure it. Thompson and colleagues returned to the same participants and, using national death records, followed them for up to 18 years, publishing in the New England Journal of Medicine in 2013.

The result was reassuring where it counts. Fifteen-year survival was 78.0 percent with finasteride and 78.2 percent with placebo. The unadjusted hazard ratio for death was 1.02 (95 percent confidence interval 0.97 to 1.08). There was no significant difference in overall survival and none in survival after a prostate cancer diagnosis.

That flatness is the whole argument. If finasteride had truly generated an excess of deadly tumors, a mortality gap should have opened over 15 to 18 years. None appeared. The absence of a survival penalty is strong evidence that the high-grade signal was about detection, not about the drug driving fatal disease.

A reasoning template for the clinic#

The value of this story reaches past one medication. It is a compact model for reading any alarming intermediate result. An intermediate outcome, such as a biopsy-defined grade, can move for reasons unrelated to what a patient actually experiences. When a treatment alters the organ being measured, it can also alter the measurement. The disciplined move is to ask whether the measuring tool worked the same way in both groups before you decide that the underlying biology differed, and then to hold judgment until a hard endpoint reports.

None of this turns finasteride into something men should seek out for prevention. PCPT studied a narrow population of older men with low PSA, the trade-offs include well-documented sexual side effects, and prostate cancer screening has changed a great deal since. Whether a 5-alpha reductase inhibitor makes sense for a given person, whether for benign prostatic symptoms, hair loss, or another reason, is a conversation to have with a clinician who knows that person's history.

The durable lesson is narrower and steadier than the original scare. Finasteride reduced total and low-grade prostate cancer in PCPT, the high-grade alarm was largely a measurement artifact, and two decades of survival data show no penalty. The frightening number was real. The story built around it was not.