By Aaron Krol

November 19, 2013 | While clinical screening for cervical and colorectal cancers in middle-aged Americans is standard practice that is generally agreed to improve health outcomes, screening for these cancers in the elderly has been more controversial. “The potential benefits of early detection with cancer screening decline with age,” said Dr. Robin Yabroff, an epidemiologist at the National Cancer Institute, in an email to Clinical Informatics News, “yet the potential harms, including complications from follow-up testing, do not.” Because the benefits from early detection of cancer may not be felt for years after discovery, elderly patients, whose prior life expectancy is lower, cannot confidently expect improved rates of survival even if a screening does detect cancer in its early stages. In the worst case, an elderly patient may be exposed to aggressive cancer treatments that reduce quality of life without improving life expectancy. For these reasons, the U.S. Preventive Services Task Force recommends that screening be discontinued at age 75 for colorectal cancer, and at age 65 for cervical cancer.

Age, however, is a crude tool for making cost-benefit analyses of cancer screening. Dr. Yabroff is part of a team at the NCI who set out to create a more flexible life expectancy table for Americans in the age range that represents a gray area for cancer screens, which could guide decisions to continue or discontinue these procedures. The team wanted to take into account comorbidity – non-cancer conditions that influence life expectancy – as well as sex and race to make the best recommendations. To access a large set of patient data that could confidently exclude individuals with previous cancer diagnoses, the study examined a random selection of 5% of Medicare beneficiaries from the SEER (Surveillance, Epidemiology, and End Results) cancer registry areas, a representative geographic cross-section of the country for which exhaustive data on cancer diagnoses is available. Patients with a history of cancer were eliminated from the study, and Medicare claims from 1991-2005 were collected to gather information on comorbidity. A total of over 400,000 relevant patient histories were available from which to construct tables.

Each study subject was given a comorbidity score based on weighted conditions such as liver disease, dementia, ulcers, AIDS, and myocardial infarctions, and the population sorted into no comorbidity, low/medium comorbidity, and high comorbidity groups. In addition, subjects were further stratified by sex, race, and by the three most common comorbidity factors: congestive heart failure, chronic obstructive pulmonary disease, and diabetes. Each of these variables was then examined for its effect on life expectancy in the subject population, broken down year by year from ages 66 to 90. Existing life expectancy tables were consulted to extend the curves when life expectancies exceeded the range of the study (i.e., when patients survived beyond 2005).

As expected, the study revealed that certain patient groups can expect to live considerably longer than their ages alone would indicate. For example, a 70-year-old woman with no comorbidity factors has a similar life expectancy to an average 65-year-old – suggesting that if a cervical cancer screening would be recommended for a 65-year-old, it should also benefit her. On average, 75-year-olds with no comorbidity live three years longer than typical 75-year-olds, and six years longer than those with high comorbidity. The full study appeared yesterday in Annals of Internal Medicine.

This greater precision allows more informed choices to be made about preventive healthcare. “This research is important because the life expectancy estimates we developed can be used to help physicians optimize cancer screening strategies with their elderly patients, including decisions about continuing and stopping screening,” said Dr. Yabroff.

“Because these decisions are complex,” she added, “patient preferences play a critical role.” The study authors caution that their work has certain necessary limitations because of potential biases from considering only Medicare patients from SEER areas, a group chosen because of the depth of information available on its members; and that, while this information is a valuable tool for weighing risks and benefits, patients should always be consulted before making decisions about care.