Mammography in Asia: The following questions refer to the article "Will Screening Mammography in the East Do More Harm than Good?", by Gabriel M. Leung, Tai-Hing Lam, Thuan Q. Thach, Anthony J. Hedley. American Journal of Public Health 2002 (Nov);92(11):1841-1846. Unless otherwise indicated, please interpret questions in the context of the article. We do not expect you to understand the article completely. Page and column references refer to the Leung et al. article unless otherwise indicated.

  1. A program using screening mammography, as described in the Leung et al. article, is best regarded as: (Choose one best answer)

  A. primordial prevention, since the program attempts to prevent breast cancer risk factors;

  B. primary prevention, since the program attempts to lower breast cancer incidence;

  C. secondary prevention, since the program attempts to prevent or delay death in those who already have breast cancer;

  D. a high-risk prevention approach, since the program targets women at higher risk for breast cancer than other women of the same age.

  2. For a disease to be a suitable candidate for a screening program, it must be a serious public health problem and have a suitable natural history. Which component of natural history (for a non-infectious disease) is essential? (Choose one best answer)

  A. prolonged period during which disease is detectable but not symptomatic;

  B. prolonged period between exposure to the etiologic agent and occurrence of recognizable symptoms;

  C. characteristic symptom or symptom profile;

  D. case fatality rate should be close to 100%.

  3. A hypothetical observational breast cancer study enrolled 10,000 women age 50 years and older during 1990 and followed each woman for 10 years. All women received a free medical exam, including a mammogram, at baseline and each year thereafter. All deaths were fully investigated, and there was no loss to follow-up. At the baseline examination, 80 women were diagnosed as having early breast cancer. 20 of these women died during the 10 year follow-up. By the end of follow-up, 240 other women had developed breast cancer, including 30 who died from their disease. 200 other women (who did not develop breast cancer) died from other causes. Answer the following questions based on this information (suggestion: drawing a flow chart may help). Show calculations, give answers to 3 significant figures, and show units.

3a. What was the prevalence of (diagnosed) early breast cancer in the study population at baseline?

3b. What was the 10-year case-fatality “rate” in women with early breast cancer at baseline? Why did we place “rate” in quotation marks?

3c. What was the cumulative incidence (incidence proportion) of breast cancer during the 10 year follow-up? (Remember: show calculation and answer to 3 significant figures.)

3d. What was the incidence density (incidence rate) of breast cancer during the 10 years, under reasonable assumptions about when deaths and new cases occurred?

3e. What was the breast cancer mortality rate in the study population during the 10 years?

3f. What was the all-cause mortality rate in the study population during the 10 years?

3g. What was the prevalence of (diagnosed) breast cancer in 2000?

  4. In order to make projections of potential benefit of mammographic screening programs for women in Asia, Leung et al. carried out a meta-analysis of randomized controlled trials of the effectiveness of mammographic screening in reducing mortality from breast cancer. The authors excluded from their meta-analysis studies that enrolled primarily symptomatic patients with breast lumps, pain, nipple discharge, or enlarged lymph nodes (p1841, middle column). Briefly explain the rationale for this exclusion.

  5. Depending on their study design, evaluations of the efficacy of breast cancer screening programs are vulnerable to three phenomena related to the nature and natural history of the disease: a) lead time bias, b) length bias, and c) overdiagnosis bias. For each of these types of bias give a succinct statement of what this bias is and explain whether it is a concern (and if it is, how the bias could occur) in studies that analyze breast cancer mortality in women randomly assigned to a screening group or a non-screening group.

  5a. lead time bias

  5b. length bias

  5c. overdiagnosis bias

  6. The American Cancer Society (ACS) publication Cancer Facts & Figures, 2003 (http://ww3.cancer.org/downloads/STT/CAFF2003PWSecured.pdf) gives the annual incidence of invasive female breast cancer in the U.S. as 136.7 per 100,000 women during 1995-1999 (adjusted to the 2000 U.S. standard population), whereas in Figure 2 (in Leung et al.) the annual incidence of invasive breast cancer is above 300 per 100,000. What is the most likely explanation for this difference?

  7. “In Hong Kong, there were 868 new cases of breast cancer for women aged 50 years or more in 1996, an incidence of 123.3 per 100,000.” (p1843, left col. - note that “50 years or more” includes women over 69 years of age). If the incidence rate of 123.3 per 100,000 women-years is a crude rate (i.e., not standardized), what was the approximate number of Hong Kong women in this age group in 1996? (3 significant figures)

  8. The authors give an estimate of 146.1 per 100,000 women for Hong Kong’s projected breast cancer “prevalence ratio” at screening (p 1843, left col., bottom).

  8a. Derive the authors’ estimate.

  8b. Use the prevalence estimate (146.1 per 100,000 women age 50+ years) to estimate the number of cases of breast cancer that would be available to detect with the screening program if there were 700,000 women age 50+ years old in Hong Kong in 1996 and all participated?

  9. At the top of the middle column of page 1843, the authors project a range for the PPV for screening in Hong Kong women (1.8% to 13.4%). This question asks you to derive that range, step by step, and to interpret it.

  9a. Use your answer to question 8 to derive low and high estimates of the number of cases that would be detected based on the “optimal ranges” stated at the bottom of page 1843, column 1.

  9b. Use your answer to question 8 and the “optimal ranges” to derive low and high estimates for the number of women without breast cancer correctly classified as such (i.e., “true negatives”) if there were 700,000 women age 50 years or more in Hong Kong in 1996 and all were screened.

  9c. Using your answers above, derive the range for PPV.

  9d. Show the source of the percentages in the authors’ statement “So we would expect at least 86% and as many as 98%, of positive screens to be false positives.” (page 1843, top of column 2).

10. On page 1842 (right column), the authors state that “the case detection rate (i.e., screen-detected cancers plus interval cancers) in the screened group is the best proxy available for the prevalence of the condition at screening.” Which of the following conditions do the authors include in their usage of “the condition”? Indicate all that apply.

a. Neoplastic growths that will not result in clinical disease (i.e., cause symptoms or tissue damage)

b. Invasive cancers that are asymptomatic but detectable through screening

c. Invasive cancers that are symptomatic but not yet diagnosed

d. Invasive cancers that have been diagnosed

11. The cumulative false-positive rate cited to Elmore et al. (p1844, middle col.) corresponds to a cumulative specificity of only 50.9%. Why is that specificity so much lower than the range of 93.5%-99.1% cited on page 1843, bottom of left column?

12. “Miller et al. argued that the excellent survival (87.5% alive at 10 years) of women with DCIS and early impalpable invasive cancer in CNBSS II was almost certainly due to a combination of lead time and length bias.” (middle of left col. on p1844) In 1-3 sentences, express this statement in a manner that an educated person who had never encountered epidemiology would understand (refer to “DCIS and early impalpable invasive cancer” as “very early breast cancer”).