Breast Cancer Screening and Dialysis: Too Much or Too Little

Breast cancer is a common cause of morbidity and mortality among women1 and current evidence-based clinical practice guidelines recommend periodic mammography for detection and treatment of early disease.2 The applicability of guidelines for mammography screening for women with end-stage renal disease (ESRD) treated by dialysis is unsettled. Cost-effectiveness and life expectancy benefit analyses of breast cancer screening in dialysis patients have concluded that screening may have limited utility.3, 4, 5 Wong and colleagues have extended these analyses in this issue of the American Journal of Kidney Diseases using a deterministic Markov model to estimate the incremental costs and benefits of breast cancer screening in a hypothetical cohort of 1,000 nondiabetic and diabetic women who started dialysis at 50 years of age.6 Over 30 years, breast cancer mammography screening prevented 1 death with a net gain in life expectancy of 1.3 days for each woman in the cohort. The cost per life-years saved (ICER; incremental cost effectiveness ratio) was $109,852 for the overall cohort and close to $1 million when limited to diabetic women on dialysis.

Although the cost-effectiveness of screening dialysis patients for breast cancer varied with the assumptions made by Wong et al, these results were consistent with earlier studies that found screening not to be cost effective when judged on a maximum cost of $50,000 for a 1-year increment in quality-adjusted life as a threshold.7 The authors conclude that “even under the most favorable conditions…breast cancer screening in patients with ESRD on dialysis may not be good value for money,” and suggest that patients' decisions about mammography should reflect an unbiased presentation of risk and benefits, personal breast cancer risk, underlying comorbid conditions, life expectancy, and quality of life.

Clinicians seeking to make sense of these analyses should take note of several issues. First is the epidemiology (risk) of breast cancer among dialysis patients. While dialysis patients are more likely to have cancer,8, 9, 10 the increase in risk is not consistent across all cancers, and there is conflicting evidence about the relative risk of breast cancer in the dialysis population. An international cohort study of more than 372,000 women on dialysis found no increase in breast cancer risk among women on dialysis in European and Australian registries and a 50% increase (standardized incidence ratio [SIR], 1.5; 95% confidence interval, 1.1 to 2.0) among women in the US Renal Data System (USRDS) registry.10 Thus, compared with a hypothetical twin with normal kidney function, a patient on dialysis can be assumed, in the absence of more definitive information from national ESRD and cancer registries, to be at comparable or increased risk for breast cancer.

The second issue is the expected benefit of systematic breast cancer screening. It is instructive to compare Wong et al's results with those from breast cancer screening clinical trials performed in the general population (Table 1). The absolute and relative risk reduction of breast cancer attributed to screening women on dialysis as estimated by Wong et al is comparable to risk reductions reported from trials included in the US Preventive Services Task Force's (USPSTF) summary of evidence supporting breast cancer screening in the general population.11 Thus, the twins might expect, over comparable lifetimes, similar benefits of screening. However, no randomized controlled trial (RCT) has confirmed the efficacy of breast cancer screening in women on dialysis. Thus, robust evidence on the efficacy or effectiveness of breast cancer screening in dialysis patients is lacking.

Table 1. Comparison Between Wong and Colleagues' Findings and Breast Cancer Screening Trials Performed in the General Population

	Authors or Study Name (Year)	Study Design	Study Population	Intervention	Relative Risk Reduction	Absolute Risk Reduction per 1,000 Women
	Wong (2008)	Decision analysis modeling	Women on dialysis, aged 50-69	Mammography (n=1,000) v no screening (n=1,000), annually	0.17	1.0
	Stockholm (1981)	RCT	Women aged 40-64	Mammography alone (n=40,318) v usual care (n=19,943), biennially	0.09	0.288
	Gothenburg (1982)	RCT	Women aged 39-59	Mammography alone (n=28,628) v usual care (n=28,809), every 18 months	0.24	0.878
	Malmo (1978)	RCT	Women aged 45-70	Mammography alone (n=21,088) v usual care (n=21,195), biennially	0.18	1.712
	Swedish 2-County Trial (1977)	RCT	Women aged 40-74	Mammography alone (n=77,080) v usual care (n=55,985), biennially	0.32	1.809
	CNBSS-1 (1980)	RCT	Women aged 40-49	Mammography + CBE (n=25,214) v usual care (n=25,216), annually	0.03	0.12
	CNBSS-2 (1980)	RCT	Women aged 50-59	Mammography + CBE (n=19,711) v CBE (n=19,694), annually	−0.02	−0.097
	HIP (1963)	RCT	Women aged 40-64	Mammography + CBE (n=30,239) v usual care (n=30,256), annually	0.21	1.438
	Edinburgh (1978)	RCT	Women aged 45-64	Mammography + CBE (n=28,628) v usual care (n=26,015), biennially	0.21	1.020

Note: Adapted from Humphrey et al.11 Abbreviations: CBE, clinical breast examination; CNBSS, Canadian National Breast Cancer Screening Study; HIP, Health Insurance Plan of Greater New York; RCT, randomized controlled trial.

What, then, drives the lack of cost-effectiveness in our twin sister treated by dialysis? As reported by Wong et al, it reflects the high underlying non–breast cancer mortality experienced by the twin treated by dialysis. The presence of other risks in dialysis patients that “compete” with breast cancer for mortality leads to the fundamental concept of competing risks.12 A crucial point is whether—under competing risks—the survival time (ie, the time span from day 1 of dialysis to death) for women on dialysis is greater than the interval observed before the Kaplan-Meier curves separate in breast cancer screening trials. This analysis should be conducted separately for diabetic and nondiabetic dialysis patients since their survival rates differ.13

Beyond reductions in all-cause mortality risk for the dialysis-treated twin, Wong et al report that the only factor that might render breast cancer screening more cost effective is a higher prevalence of breast cancer in women treated by dialysis. This suggests that one way to increase the cost-effectiveness of breast cancer screening for dialysis patients is to stratify women with respect to expected benefit. Expected longevity might be one such stratification factor. A recent analysis of USRDS data by Walker et al implies that targeted mammography screening based on life expectancy is already a component of current preventive care in US dialysis units.14 A second stratification factor might be risk for breast cancer. For example, factors like a positive breast cancer family history and/or the presence of genetic variant(s) might be considered in order to identify women on dialysis with very high risk. While there is continuing interest in developing such breast cancer risk prediction models for the general population,15, 16, 17 there is no tool that has been shown to be a valid and reliable predictor in the dialysis population.

One important issue that should be addressed in future trials of screening and risk prediction in dialysis patients is the possibility that the efficacy of breast cancer screening in this group is diminished by a high rate of false-positive mammograms due to breast calcification. Several studies have demonstrated that while breast calcification detected during screening mammography of women on dialysis is increased,18, 19 the patterns of calcification were more likely to be benign.19 While there was no increase in the call-back rate, once called back, dialysis patients were more likely to be recommended for biopsy.19 It remains unclear if this increased biopsy rate is justified by an increased detection of breast cancer and additional studies are needed to sort out this issue.

While targeted screening may be a promising method to increase the cost-effectiveness of mammography in the dialysis population, clinicians must consider how cost and criteria balance against the ethical principles of fairness, beneficence (to maximize good), and justice.20 If targeting strategies based on risk stratification of women on dialysis are adopted, they should be based on unbiased, quantitative decision tools and not on variable clinical judgment.

What, then, are clinicians to make of this careful and thought-provoking analysis by Wong and colleagues? An expert panel on cost-effectiveness analysis in health and medicine reminds us that cost-effectiveness analyses are “intended to inform decisions at the level of broad resource allocation and may provide little guidance about optimal bedside management of individuals.”21 As long as evidence of diminished effectiveness from RCTs conducted in that particular group is lacking, we think that a recommendation against screening a subgroup of patients which is based on the lack of cost-effectiveness is indeed questionable. This is particularly true if the conclusion restricts life-saving care. More generally, Wong et al's study leads back to previously unresolved questions22: When should a cost-effectiveness study be performed before the clinical trial? Can a cost-effectiveness study allow clinicians to integrate or change modalities into practice when RCTs are lacking?

When cost-effectiveness studies are performed ahead of RCTs relevant to the particular population—and therefore in the absence of population-specific information on efficacy, as is the case for breast screening for dialysis patients—the results should be viewed either as a result which supports the extension of therapeutic recommendations to the population or as identifying a population where additional evidence is needed to support both decision making by clinicians and policy analyses. These additional trials should explicitly address the issues raised in the cost-effectiveness analysis that call the recommendation into question. For our case, this would include better estimates of breast cancer prevalence and incidence in dialysis populations, better understanding of mammography sensitivity and specificity in the face of increased breast calcifications attendant with dialysis, fuller information about response to breast cancer therapy, and development and utility of risk stratification tools.

While waiting for the next generation of evidence for breast cancer detection and treatment evidence in dialysis patients, we feel that clinicians should follow national practice guideline recommendations2 and recommend mammography to patients like our dialysis-treated twin. Policy makers considering breast cancer screening in the preventive care of dialysis patients should be aware of the limitations of cost-effectiveness models. Disease models (eg, natural history) are at best rough maps of the disease process with many details unknown and analysis dependent on assumptions.23 No model can, nor should, claim to predict the truth.24 After reading this interesting paper, clinicians and policymakers concerned about breast cancer screening for women on dialysis will understand why the authors stressed that the deterministic Markov model was developed “from the perspective of a healthcare payer,” and should insist on appropriate evidence to support or refute these conclusions.