Calculations of cancer mortality depend on accurate determination of the underlying cause of death. The authors argue that cancer mortality should include deaths from treatment for cancer as well as deaths from the disease. They note that clear guidelines for classifying treatment-related cancer deaths do not exist.

"The more we look for cancer and the more we treat people with the diagnosis, the more important it will be to properly assign diagnostic and treatment-related deaths. Otherwise, observed mortality trends may make harmful interventions appear beneficial," said H. Gilbert Welch, M.D., lead author.

As a rule, researchers consider deaths within 30 days of a surgical procedure to be treatment-related when calculating mortality. To determine if this rule is being applied uniformly to cancer patients, the authors used national cancer registry data to find the reported cause of death in patients who, between 1994 and 1998, died within one month of cancer-related surgery to remove a solid tumor. Among the 4,135 deaths within one month of diagnosis and cancer-related surgery, 41 percent were attributed to a cause other than the cancer.

The proportion of cases not attributed to cancer ranged from 13 percent for cervical cancer to 81 percent for laryngeal cancer. There is a trend toward increasing misclassification among cancers such as those of the breast and prostate for which early detection has increased substantially, the authors note.

"Although the estimated effect of this misclassification on overall cancer mortality is modest, it may be indicative of more widespread confusion about how to code treatment-related deaths of patients with cancer," the authors conclude.

They advocate steps to assure that cancer mortality remains a valid indicator of cancer progress and propose development of standard rules such as that all deaths within one month of surgery, radiation therapy, or chemotherapy be attributed to the cancer for which the treatments were initiated.