Does Patient Life Expectancy (LE) Affect Receipt of Routine Cancer Screening in the United States? A Population-Based Study

Reporter: Abigail Berman, MD
The Abramson Cancer Center of the University of Pennsylvania
Last Modified: October 31, 2012

Presenting Author: Trevor Royce, MS4
Affiliation: University of North Carolina at Chapel Hill, Chapel Hill, NC


  • Routine cancer screening methods exist for a variety of malignancies including prostate cancer, breast cancer, colorectal cancer, and cervical cancer. Screening, however, has unproven benefit for patients with limited life expectancy.
  • Many organizations recommend cessation of screening for certain cancers at a specified life expectancy or at an arbitrary age cut-off as a surrogate for life expectancy.
  • The United States Preventative Services Task Force (USPSTF) has existing guidelines for cancer screening as follows:
    • Colorectal cancer (2008, grade A recommendation) as a fecal occult blood test, sigmoidoscopy, or colonoscopy from ages 50-75
    • Breast cancer (2009; grade B) as a mammogram every 2 years from ages 50-74
    • Cervical cancer (2012; grade A) as a pap smear every 3 years from ages 21-65
    • Prostate cancer (2012; grade D) as no screening. However, the AUA recommends PSA testing with cessation when life expectancy (LE) is less than10 years.
  • The goal of this study is to examine patterns of cancer screening in patients with different LE using data collected by the population-based National Health Interview Survey (NHIS).

Materials and Methods

  • NHIS is a study sponsored by the Center for Disease Control, and designed to collect data from samples representative of the US population.
    • Subjects are surveyed once, collecting information regarding demographics, comorbidities, and receipt of health services.
    • Approximately 90,000 patients are sampled every year.
  • A total of 25,286 subjects age 65 years and older in NHIS cohorts from 2000, 2003, 2005, 2008, and 2010 were included. Patients were excluded if life expectancy could be calculated, had a history positive for screened cancer in question, or a prior hysterectomy for the cervical cancer screening analysis.
  • Using a validated index specifically for NHIS cohorts, the 9-year risk of mortality was calculated.
    • The mortality index included sex, age, smoking status, body mass index, comorbid conditions, number of hospitalizations, perceived health, and functional measures.
  • Subjects were grouped into low (<25%), intermediate (25-50%), or high (>50%) risk of mortality to examine the proportion of each subgroup receiving cancer screening.
  • A logistic regression model was performed for each screening test to assess potential factors associated with receipt of screening in subjects with a high risk of mortality. Analyses accounted for population sampling weights.


  • PSA screening within the last 2 years occurred in 69.8% in the low risk group, 65.3% in the intermediate risk, and 56% in the high risk.
  • Breast cancer screening within the last 2 years occurred in 73.9% in the low risk group, 63.1% in the intermediate risk, and 4.8% in the high risk.
  • Cervical cancer screening within the last 3 years occurred in 70.6% in the low risk group, 55.0% in the intermediate risk, and 41.0% in the high risk.
  • Colorectal cancer screening within the last 5 years occurred in 50.7% in the low risk group, 47.6% in the intermediate risk, and 41.2% in the high risk.
  • On multivariable analyses in subjects with a high mortality risk, there was a significantly lower risk (OR 0.63) risk of prostate cancer screening for age >84, lower risk of breast screening for age 70-74 (OR 0.49), 80-84 (OR 0.52), age >84 (OR 0.63), lower risk of cervical screening for age 75-79 (OR 0.02), 80-84 (OR 0.34), age >84 (OR 0.18), and higher risk of colorectal screening for age 80-84 (OR 1.30), age >84 (OR 2.08).
  • Controlling for age, race, sociodemographic variables and geographic region, there was more Pap smear (2010 vs. 2000, OR 1.7, p<.001) and colorectal cancer screening (2010 vs. 2000, OR 1.8, p<.001) in recent years.

Author's Conclusions

  • A significant proportion of the US population with high risk of 9-year mortality received cancer screening that is unlikely to provide benefit; for cervical and colorectal cancer screening, this use increased from 2000 to 2010.
  • These results raise concerns about potential over-screening in those with limited life expectancy. There is opportunity for eliminating health care expenditures while improving patient outcomes through prevention of unnecessary harms from procedures and treatments.

Clinical Implications

  • This is an interesting study regarding the overuse of screening in the elderly population. Despite clear screening guidelines, this study shows that patients with a high risk of a low life expectancy are often screened for cancers, including PSA in 56%, mammograms in 47%, pap smears in 42%, and colorectal in 41%.
    • In addition, they show that overscreening continues to occur in more recent years.
  • Strengths of this study include a very large population (>25,000 patients) and a rigorous multivariate analysis.
  • There have been several randomized studies that do not show a benefit in prostate cancer screening in men with comorbidities. The PLCO study (Crawford ED, J Clin Oncol. 2011 Feb 1;29(4):355-361) found that in men with no or minimal comorbidity, the decrease in PCA-specific mortality associated with screening was significant (22 v 38 deaths). However, in men with at least 1 significant comorbidity, there was no decrease in PCSM with screening. This was one of many studies that formed the basis of the USPSTF group.
  • There are several limitations to this study:
    • It uses the NHIS comorbidity index which is only one of many validated comorbidity indices. The authors could consider validating this data with the Social Security Death Index or Charlson comorbidity index.
    • The authors arbitrarily grouped subjects into low (<25%), intermediate (25-50%), or high (>50%) risk of mortality. Therefore, there is potential bias that these proportions do not represent true risk, and these groupings require further validation.
    • This study does not address cancer treatment. For example, if prostate cancer is diagnosed in an elderly man, his practitioner may recommend active surveillance. Therefore, the risks of overdiagnosis may not be as great if patients are not also undergoing unnecessary treatment.
  • Future studies could include the formation of a risk score calculator to determine whether a patient should be screened for a given cancer.
    • For example, the risk of diagnosis of prostate cancer and death from prostate cancer should take into account not only age and comorbidities but also race, as there is ample evidence to suggest that african americans and hispanics are diagnosed with more biologically-aggressive prostate cancer.
  • This study remains of interest and may assist physicians with narrowing screening populations rather than omitting screening entirely for large, heterogenous, populations.