Study Sees Cancer Survival in a More Optimistic Light
By DONALD G. McNEIL Jr.
October 11, 2002
"It's malignant? Then give it to me straight, Doc. How long have I got?"
The correct answer to that question, a new study suggests, may be more optimistic than Doc thinks.
Because statistical methods used in calculating cancer survival are too conservative, the study says, Americans with cancer are actually living longer than many doctors have been led to expect, and patients researching their disease on the Internet may be reading prognoses that are grimmer than the truth.
A study of 24 million Americans in whom cancer was diagnosed from 1978 through 1998, to be published today in The Lancet, a British medical journal, argues that the traditional cohort method used to estimate survival rates does not consider the leaps made in cancer diagnosis and treatment in recent years.
Using a different method called period analysis, the epidemiology study concluded that all cancer patients whose tumors are diagnosed today have a 51 percent chance of living 20 years. The cohort method would give them only a 40 percent chance.
The differences vary widely with the types of cancer. Patients with cancer of the prostate, ovary, rectum or breast, or those with leukemia or Hodgkin's disease, probably have much better survival rates than previous figures suggested, said the study's author, Dr. Hermann Brenner of the German Center for Research on Aging in Heidelberg, who used the National Cancer Institute's database.
For example, Dr. Brenner said, a woman with ovarian cancer may have been told that she has a 35 percent chance of surviving 20 years; in fact, she probably has at least a 50 percent chance of living that long. A man with prostate cancer may be told that he has a 44 percent chance of surviving 20 years; in fact, Dr. Brenner said, he probably has an 81 percent chance.
However, Dr. Brenner conceded that some of the most significant "improvements" in survival come not from miracle cures but from new tests that find tumors -- including tiny, slow-growing ones -- very early. "For prostate cancer, it's clear that early diagnosis or even over-diagnosis may play a role," he said.
But for colorectal cancer, leukemia and other cancers, he said, early detection and better treatment produce genuinely longer lives. The current chance of living 20 years after a diagnosis of rectal cancer is about 49 percent, not 39 percent, his study found, and with leukemia 26 percent rather than 21 percent.
The new analysis included 20-year relative survival rates of 95 percent for thyroid cancer, 88 percent for testicular cancer, better than 80 percent for melanomas and prostate cancer, and 65 percent for breast cancer. The study did not analyze which outcomes stemmed from earlier diagnosis and which stemmed from better treatment.
Other biostatistics experts agreed that Dr. Brenner's analysis probably produces results closer to what a modern prognosis should be.
"In an era when life expectancies are going up, even the period method will understate survival probability," said Dr. David A. Freedman, a professor of statistics at the University of California at Berkeley.
To estimate how many patients will live 20 years, the simpler cohort method looks at patients whose disease was diagnosed in 1978 and counts how many were still alive in 1998. The flaw in using that method to predict the survival of someone whose cancer was diagnosed in 2002 is that it assumes a 1978 treatment plan.
With period analysis, a statistician would use 1997 data to compute the probability of living one year, 1996 data to compute the probability of surviving a second, and so on. These probabilities can be multiplied together to estimate the probability of surviving 20 years. ( Demographers use this method of analysis to compute life expectancy because looking back 80 years to the birth dates of cohorts of aging citizens says little about life expectancy today.)
But Dr. Donald Berry, head of biostatistics at the M. D. Anderson Cancer Center in Houston, cautioned that the study might have little or no effect on what patients hear from their doctors.
"No clinician -- well, almost no clinician," Dr. Berry said, would simply quote to a patient the overall survival numbers for a type of cancer. Any good doctor making a prognosis, he said, takes into account the size of a tumor, how far it has spread, the patient's age, success rates of new treatments and other factors.
Dr. Brenner agreed, but noted that "these survival statistics based on the old numbers are on the Internet everywhere and are seen there by patients."
An advantage of his calculation method, Dr. Brenner said, is that it is faster at making obvious the impact of new therapies. Such observations can be useful to policy makers who parcel out dollars to researchers.
Dr. Brenner used the Surveillance, Epidemiology and End Results database of the National Cancer Institute, which has tracked cancers in about 10 percent of the American population for 30 years.
Dr. Brenda Edwards, the deputy director of the cancer institute's database unit, said Dr. Brenner's conclusions were correct, adding that "we're going to look at some of the features of his method."
Not all the news from Dr. Brenner's recalculation was optimistic.
Some cancers that have always had grim prognoses, including those of the lungs, esophagus, liver and pancreas, as well as multiple myelomas, still have very low 20-year survival rates of 2 percent to 8 percent.