Tuesday, April 21, 2020

Don't miss these important prostate and lung cancer screening studies

In a time when the most common cause of death in older Americans is COVID-19, it seems nearly old-fashioned to write about screening for prostate and lung cancers, which, respectively, account for a small proportion of overall deaths in men (prostate) and only in smokers rivals mortality from heart disease (lung). But while the pandemic has riveted media attention on COVID-19 studies of questionable value, including non-peer reviewed preprints and uncontrolled case series, I think it's still important to highlight two notable recent publications on pitfalls of cancer screening.

I generally advise my average-risk patients not to have prostate cancer screening. Even viewing the evidence in the most optimistic light, few men are helped and many more are harmed. Notably, for this linked graphic the U.S. Preventive Services Task Force used the prostate cancer mortality risk reduction estimated in the European screening trial, the only one of three major PSA screening trials (the others were conducted in the U.S. and United Kingdom) to show any benefit whatsoever. And, as I explained in an earlier blog post, relying on observational trends in prostate cancer mortality to deduce benefits from screening is fraught with problems.

In a new analysis in the New England Journal of Medicine, Drs. Gil Welch and Peter Albertsen dive deeper into the prostate cancer mortality data and its implications for screening. They propose an explanation for a trend that I had always found puzzling: before starting to decline in the early 1990s (which some experts attribute to the effect of screening, but Welch and Albertsen argue convincingly must also be the result of improved treatment), mortality rose between 1970 and 1990 by more than 30 percent. This earlier mortality increase likely resulted from "sticky diagnosis bias," which occurred because transuretral resections of the prostate (TURP) procedures were widely used to treat benign prostatic hyperplasia (BPH) in older men, leading to many more incidental diagnoses (or overdiagnoses) of prostate cancer that ended up on death certificates. Welch and Albertsen go on to explain why "decreasing prostate cancer mortality may be a misleading metric in evaluating PSA screening":

Screening may more easily change the distribution of causes of death (trading off one cause for another) than extend life (as implied by promises to “save lives”). This issue is particularly relevant to PSA screening, since the median age at death due to prostate cancer is so high — 80 years (as compared with 72 for lung cancer and 68 for breast cancer). For the elderly, the combination of a high burden of competing risks for death and high rates of intervention-related complications conspires to limit any reduction in all-cause mortality offered by screening.

Of course, length of life is not the only relevant outcome; quality of life is equally important. If screening helped avert the pain that can be associated with metastatic disease, that would change the calculus, but it is not clear that it often does. Furthermore, the quality-of-life question has two sides. Prostate cancer treatment itself results in substantial morbidity: surgery and radiation can produce impotence and bowel and bladder problems; antiandrogen therapy leads to hot flashes, decreased stamina, and metabolic syndrome. Which group of men — the treated or the untreated — feels a bigger effect on quality of life can be debated.

From a statistical perspective, screening for lung cancer has always held more promise than screening for prostate cancer, since many more people (men and women) die prematurely from the former. Based on consistent results from several randomized trials, the medical consensus is that low-dose CT screening in carefully selected older smokers (or former smokers) reduces lung cancer mortality, and, possibly, all-cause mortality. However, as I've argued in the past, the potential harms of LDCT make it essential that clinicians discuss the pros and cons of screening with patients first, rather than universally recommending it.

A study in JAMA Internal Medicine assessed the informational content of 162 lung cancer screening program websites, half at academic medical centers and half at community medical centers. The authors found that 98 percent of web sites described potential benefits, but only 48 percent described potential harms, with community centers being even less likely (40 percent) to mention any harms. Overdiagnosis, which is arguably the most substantial harm since it usually leads to unnecessary invasive treatment, was mentioned by only 14% of academic centers and none of the community centers. Granted, going to a website is only the first step toward actually being screened for lung cancer, but it's discouraging that this content is so unbalanced.

On the Lown Institute blog,  Judith Garber pointed out one reason for the "big benefit, little or no harm" messages that these websites send:

It is important to acknowledge the profit motive behind screening. If we reimbursed hospitals not for the volume of procedures done, but the outcomes for patients, we would likely a much greater effort to target individuals who would most benefit from screening, rather than hospitals trying to persuade as many people as possible to get screened.