A recent study published in Health Affairs either proves the superiority of U.S. medical care for cancer, or illustrates again how ignorance of basic statistical principles can lead to wrong conclusions.
The study found that U.S. cancer patients who were diagnosed between 1995 and 1999 lived, on average, 11.1 years after their diagnosis. Similar patients from 10 European countries lived an average 9.3 years. By 1999 (the last year the researchers analyzed), the average U.S. expenditure per cancer case was $70,000. That was nearly 50 percent higher than the cost in 1983. The cost in Europe was $44,000-16 percent higher for the same interval. Using standard figures for each extra year of life, the researchers concluded that the value of the U.S. survival gains outweighed the cost by an average $61,000 per case. They pronounced the additional spending on cancer care in the United States “worth it.”
But a Reuters story analyzing the research begged to differ. “This study is pure folly,” Dr. Don Berry, a biostatistician at MD Anderson Cancer Center in Houston, told Reuters. “It’s completely misguided and it’s dangerous. Not only are the authors’ analyses flawed but their conclusions are also wrong.”
Reuters also found the credentials of the study’s lead author, Tomas Philipson of the University of Chicago, wanting. He’s a health economist who served in the administration of President George W. Bush and advised the McCain presidential campaign on health-care issues. The point? The research might not be as unbiased as science demands.
Thirteen common cancers were examined in the study. Researchers analyzed survival-how long a patient lived after being diagnosed-in the period from 1983-1999. They looked at survival gains, or how long patients diagnosed in later years lived compared with those diagnosed earlier in the period. Those gains, they said, demonstrate the progress countries made in treating cancer.
Sounds reasonable. But survival data are tricky; they’re not cold, irrefutable numbers that can quantify success, thanks to something called lead-time bias.
Take two hypothetical people who both get the same kind of cancer on the same day. One of them gets an immediate diagnosis and lives another two years. The other is diagnosed eighteen months after the first, and lives only six months. Patient No. 1 had a “better” survival rate — two years compared to the six months of Patient No. 2, but she didn’t live any longer. She just knew she was sick earlier. Both patients lived two years. That’s “lead-time bias” at work.
That’s why, as the Reuters analysis says, “Crediting medical care with ‘improving survival’ is therefore misleading, cancer experts have long argued. Lead-time bias makes it seem patients live longer, but the only thing that is longer is the number of years they know they have cancer…”
But Philipson’s team based its conclusions on survival data, arguing that because U.S. cancer mortality rates declined faster than those of Europe, they’re evidence of survival gains.
Berry articulated a related point with which readers of this blog will be familiar: that overdiagnosis is a problem. Cancer screening, particularly for breast and prostate cancers, is more common in the U.S. than in Europe, and the more testing, the more cancer will be found. But as Berry noted for Reuters, “These are cancers that tend to be slowly growing and many would never kill anyone.”
If, in a diagnostic procedure, you find what you’re looking for, does that denote a successful test? Not if it makes a healthy person a cancer patient if the tumor otherwise is not life-threatening. Including such cases, whose numbers are higher in the U.S. than in Europe, makes survival data bogus.
The Health Affairs study showed survival gains in the U.S. versus Europe were greatest for prostate cancer; breast cancer claimed the second-best U.S. survival data–the two cancers where lead-time bias figures most prominently in overdiagnosis.
It’s interesting that Europe had the survival edge in data for melanoma and colorectal and uterine cancer survival gains.
According to Reuters, U.S. cancer mortality places the U.S. in the middle of countries reporting to the Organization for Economic Co-operation and Development.
If that’s not enough to prompt questions about the “Spend More! Live Longer!” theory of cancer survival, consider this: Even the study’s researchers concede that it’s impossible to state that improved survival is a direct result of spending money on cancer care. It might result from improved screenings that detect the “pseudo-disease,” or nonaggressive, nonthreatening tumors that artificially enhance survival data.
And Philpson said, “In the last decade, spending in the U.S. has increased more than in Europe. I would be extremely surprised if the survival gains haven’t continued. But it is a much more open question whether that additional spending has been accompanied by an increase in longevity.”
According to Reuters, in 2004 (the last year for which figures were available), the U.S. spent $72 billion on cancer care. It also noted that Philipson’s research was supported in part by Bristol-Myers Squibb Co. That company makes a melanoma drug, Yervoy, that costs $120,000 for a full course of treatment.
Certainly the cost of cancer drugs have increased. Dendreon Corp. makes Provenge for prostate cancer at $93,000 per treatment. Bristol and Eli Lilly and Co. make Erbitux at $100,000 per year. The researchers said their analysis “does not imply that all treatments are cost-effective.”
Remember, Philipson is an economist. His scholarship concerns how much an additional year of life is worth. His researchers assumed the value to be $150,000 to $360,000.
No wonder economics is referred to as “the dismal science.”
Footnote: Here is a good explanation from “the Incidental Economist” blog of the “lead-time bias” problem in medical statistics and why the correct number to focus on is death rates, not survival rates. However, important caveat from the same blog: Survival rates are very important to any individual patient, because they can tell you how long YOU might live with a particular cancer at a particular stage. The importance of death rates is on the macro level: Does early screening bend the death curve down, or not?