Impact of evidence-based clinical guidelines on the adoption of postmastectomy radiation in older women


  • This study used the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database.



Although postmastectomy radiation therapy (PMRT) improves survival for patients with high-risk breast cancer, previous literature suggested that it is underused. The impact of recent clinical guidelines on PMRT use is unknown. Accordingly, the authors used the Surveillance, Epidemiology, and End Results (SEER)-Medicare cohort to determine whether the use of PMRT has increased in response to evidence-based guidelines.


In total, 38,322 women aged ≥66 years who underwent mastectomy for invasive breast cancer between 1992 and 2005 were identified. Time trends in the receipt of PMRT for low-risk (T1/T2 N0), intermediate-risk (T1/T2 N1), and high-risk (T3/T4 and/or N2/N3) patients were characterized. Multivariate logistic regression identified risk factors for PMRT omission.


The receipt of PMRT by patients with high-risk breast cancer increased from 36.5% (95% confidence interval, 26%-46.9%) to 57.7% (95% confidence interval, 46.9%-68.4%) between 1996 and 1998 with the publication of landmark clinical trials. However no further increase in PMRT use was observed between 1999 and 2005 despite publication of multiple guidelines endorsing its use; during this period, only 54.8% (2729 of 4978) of high-risk patients received PMRT. Within this high-risk group, patients with smaller tumors or less advanced lymph node disease were at risk for PMRT omission.


After an initial increase in PMRT use in response to clinical trials, the use of PMRT did not increase further in response to guideline publication, and nearly 50% of patients with high-risk breast cancer still do not receive PMRT. Additional research is needed to determine how clinical guidelines can be used to bridge the gap between level I evidence and clinical practice. Cancer 2011;. © 2011 American Cancer Society

It is believed that evidence-based guidelines play a critical role in clinical practice because of their potential to address the complementary needs of providing direction for effective patient care while curtailing the use of futile or potentially harmful therapies. Furthermore, guidelines have the potential to dampen growth in health care costs through endorsement of cost-effective measures and discouragement of wasteful interventions.1, 2 The field of radiation oncology is particularly well positioned to benefit from evidence-based guidelines, both because of its rapidly advancing evidence base and because of the need to improve allocation efficiency when spending limited health care resources on these expensive therapies and procedures.

From 1999 to 2002, a series of evidence-based guidelines3-6 regarding the use of postmastectomy radiation (PMRT) were published in response to 3 landmark randomized, controlled trials that were published first between 1997 and 1999. Those studies demonstrated that PMRT decreased locoregional recurrence and improved survival among patients with high-risk breast cancer (herein defined as stage T3-T4 and/or N2-N3).7-9 Prior research conducted by our group and others demonstrated that publication of these clinical trial results had a positive impact on patient care, and the use of PMRT increased in response to their publication.10, 11 Evidence-based guidelines written between 1999 and 2002 sought to further improve upon this trend by providing health care providers, patients, and the population in general with assessments of current data regarding the receipt of adjuvant therapy for breast cancer.3

It is noteworthy that guidelines regarding use of PMRT relied on passive dissemination of their content from academic publications into clinical practice without any mechanism to promote guideline compliance. To our knowledge, whether PMRT guidelines were adopted successfully in the community through this model has not previously been investigated. Addressing this question is critical, because evidence-based recommendations that fail to be incorporated into clinical practice represent a lost opportunity both for improving outcomes for patients and for responsible stewardship of health care resources.

We hypothesized that the release of evidence-based guidelines regarding indications for PMRT would result in the increased receipt of PMRT by older women with high-risk breast cancer, for whom PMRT generally was recommended, but would not have an impact on the receipt of PMRT by older women with low-risk to intermediate-risk breast cancer, for whom PMRT generally was not recommended. To test this hypothesis, we used the Surveillance, Epidemiology, and End Results (SEER)-Medicare database to identify women aged ≥66 years who underwent mastectomy for invasive breast cancer between 1992 and 2005. We sought to determine the temporal association between guideline publication and changes in the receipt of PMRT by older women with breast cancer stratified according to their risk group. In addition, we sought to identify barriers to guideline adherence by identifying risk factors for the omission of PMRT among women with high-risk breast cancer, for whom PMRT is recommended by current guidelines.


Data Source

We used the SEER-Medicare database, which tracks incident malignancies in Medicare beneficiaries who reside within a SEER geographic region. Claims codes that were used in this study have been reported previously.11 Citation rates through January 2011 for evidence-based guidelines were extracted from the SCOPUS database (Elsevier, Amsterdam, the Netherlands).12

Study Sample

From 1992 to 2005, 209,579 women aged ≥66 years were diagnosed with breast cancer and reported in the SEER-Medicare cohort. We applied standard exclusions, as outlined in Table 1, to create an analytic cohort of 44,177 patients, of whom 38,332 (87%) could be assigned to a risk group.

Table 1. Number of Observations After Each Exclusion Criterion
StepCriteriaNo. of Remaining Observations
  1. HMO indicates health maintenance organization.

1Women diagnosed with breast cancer between 1992 and 2005209,579
2Include if Medicare Part A/B coverage and no HMO coverage from 12 mo before diagnosis to 12 mo after diagnosis133,655
3Exclude if died within 12 mo of diagnosis132,479
4Exclude if secondary cancer diagnosed within 12 mo of breast cancer diagnosis126,411
5Exclude if another cancer was diagnosed before index breast cancer117,927
6Exclude if histology not consistent with epithelial origin115,250
7Exclude if in situ histology98,159
8Exclude if distant metastasis at time of diagnosis94,794
9Exclude if unknown stage at diagnosis93,369
10Exclude if no pathologic confirmation93,358
11Exclude if patient received radiation before surgery92,724
12Exclude if patient did not undergo mastectomy44,482
13Exclude if patient received neoadjuvant chemotherapy44,177
14Exclude if risk group cannot be determined38,332


The primary outcome for our investigation was receipt of PMRT. Patients were considered to have received PMRT if either SEER or Medicare claims reported treatment with radiation within 1 year of the date of diagnosis.11, 13-18

Treatment-Related Variables

Breast surgery was determined from both SEER and Medicare claims, as reported previously.4-11, 13-19 Definitive surgery was defined as the most extensive surgical procedure reported by SEER or Medicare during the first 9 months after diagnosis. Patients were determined to have undergone breast reconstruction if a reconstructive procedure occurred within the first year of diagnosis as reported by SEER or by Medicare claims. Treatment with chemotherapy within 1 year of diagnosis was determined from Medicare claims, including claims for oral chemotherapeutic agents originating from the Durable Medical Equipment file.20, 21 Adjuvant endocrine therapy is not reported.

Patient-Related and Health Services Variables

Patient characteristics included year of diagnosis, age at diagnosis, race, marital status,22, 23 SEER registry, urban/rural residence, median income of Census tract or zip code,24 the percentage of adults in the Census tract or zip code with some college education,24 and the Charlson comorbidity score spanning an interval from 12 months to 1 month before diagnosis.25-27 To enhance specificity, Part B diagnosis codes were included only if they appeared either more than once over a period exceeding 30 days or in Part A claims as well.28, 29 The density of general surgeons and radiation oncologists practicing within the health service area (HSA) in which the patient resided was determined using data from the Area Resource File.30

Tumor-Related Variables

The tumor characteristics reported by SEER include size, estrogen receptor status, grade, histology,31 T4 tumor classification, and laterality. Margin status and lymph- vascular space invasion are not reported.

Statistical Analysis

To be consistent with current consensus guidelines, we stratified our cohort into low-risk (T1/T2 N0), intermediate-risk (T1/T2 N1) and high-risk (T3/T4 and/or N2/N3) groups.32 We also examined a fourth group of patients for whom PMRT would be most strongly recommended, namely, high-risk patients ages 66 to 79 years with absent or mild comorbidity. Unadjusted associations between covariates and the receipt of PMRT by risk group were tested using the Pearson chi-square test. Trends in PMRT use by calendar year quarter were determined using Joinpoint software (version 3.4.3; National Cancer Institute, Bethesda, Md).33, 34

Because the Joinpoint analysis indicated that PMRT use was relatively stable between 1999 and 2005 for high-risk patients, we created a multivariate logistic regression model to identify risk factors for PMRT omission during this period of stable use. Covariates that were significant at P ≤ .05 in unadjusted logistic regression were included in the final multivariate model. Year of diagnosis was not included in the model given the finding of stable use throughout 1999 to 2005 in the Joinpoint analysis. Goodness of fit was assessed using the Hosmer and Lemeshow test. Finally, because the choice to omit PMRT in patients with severe comorbidity is often due to poor anticipated overall survival, we used the Kaplan-Meier method to determine median survival for patients with high-risk breast cancer who did not receive PMRT.

All statistical analyses were 2-sided with P ≤ .05 and were conducted using SAS statistical software (version 9.3; SAS Institute, Inc., Cary, NC). Our institutional review board granted this study exempted status.


Baseline Characteristics

Of 38,332 women, 23,126 (60.3%) were low risk, 7211 (18.8%) were intermediate risk, and 7995 (20.9%) were high risk (Table 2). The receipt of PMRT was strongly associated with risk group, and 6.6% of low-risk patients, 16% of intermediate-risk patients, and 48.5% of high-risk patients received PMRT (P < .0001). Receipt of PMRT also was correlated with age at diagnosis, marital status, comorbidity, SEER registry, median income in Census tract or zip code, estrogen receptor status, number of involved lymph nodes, and receipt of chemotherapy. Health service characteristics that were investigated, including the density of surgeons and radiation oncologists in a given health service area, were not consistently associated with PMRT use across all risk groups (Table 2).

Table 2. Associations Between the Receipt of Postmenopausal Radiation Therapy and Demographic, Health Services, Tumor, and Treatment Characteristicsa
 Low-Risk Group (T1-T2/N0)Intermediate-Risk Group (T1-T2/N1)High-Risk Group (T3-T4 and/or N2-N3)
CharacteristicNo.Received PMRT, %PNo.Received PMRT, %PNo.Received PMRT, %P
  • PMRT indicates postmastectomy radiation therapy; HSA, health service area.

  • a

    Thirteen percent of patients (5845 of 44,177) were excluded because they could not be assigned to a risk group.

  • b


  • c

    For these variables, the unknown groups were excluded when calculating the chi-square P value. In addition, the data for the bilateral/unknown laterality groups were suppressed to protect confidentiality for those with cell sizes <11 cm.

  • d

    Quartiles were created based on the population of HSAs and not on the population of patients in this study, so the quartiles are not of equal size for the patients in this study.

Overall cohort (N=38,332)b23,1266.6 721116 799548.5 
 Year of diagnosis         
  200516666.9 55321 83054.3 
 Age, y         
  70-7464497.9 192318.4 191955.7 
  75-7959896.3 183516.2 194252.2 
  ≥8062164.1 19749.4 260632.5 
  Hispanic2907.2 10018 11841.5 
  Black12119.7 51117 74847.6 
  Asian5846.3 13720.4 16160.9 
  Others/unknown4509.3 16920.7 12756.7 
 Charlson comorbidity index         
  156846.4 176015.3 188547.6 
  ≥235565.5 129013.9 133542.9 
 Unknown9317.8 38416.9 76345.1 
 Adults in Census tract or zip code with at least some college education         
   Lowest quartile55936.1.0003177216.1.21212145.5.0008c
   Second quartile58275.8 177214.7 190549.3 
   Third quartile57996.8 176816.2 193050.1 
   Highest quartile57097.7 183117.3 195050.2 
   Unknownc198<5.6 68<16 8934.8 
Health service characteristics (1999-2005)         
 No. of radiation oncologists in HSA/HSA populationd         
   Lowest quartile7493.7<.000123814.3.1925447.2.05
   Second quartile12035.9 36918.7 37755.2 
   Third quartile46467.1 146419.1 172854.1 
   Highest quartile65098.6 231420.1 261956 
 No. of general surgeons in HSA/HSA populationd         
   Lowest quartile11335.5<.000137015.4.00833954.78
   Second quartile19986.4 60218.6 61153.2 
   Third quartile29765.7 95816.9 105055.7 
   Highest quartile70009 245521 297854.9 
Tumor characteristics         
 Tumor size, cm         
  2.0 to ≤5.086227.5 444818.7 375249.7 
  >5.0  296847.9 
  Unknown/not applicable  32650.6 
 Tumor histology         
  Lobular25768.2 74218.2 142557.6 
  Other/unknown43986.5 111617.8 144346.5 
 Estrogen receptor status         
  Negative/borderline31867.9 103422.3 149351.3 
  Unknown39905.5 108611.9 115439.5 
 No. of positive lymph nodes         
  0 (All negative)23,1266.6111234.4<.0001
  1-3 721116 111443.7 
  4-9  343851 
  ≥10  173462.1 
 None examined/not specified/unknown  59730.3 
 T4 tumor         
  No23,1266.6 721116 607749.4.0087
  Yes  191845.9 
Treatment characteristics         
 Receipt of chemotherapy         
  Yes254913.1 253724.7 367166.4 
 Breast reconstruction         
  Yes10185.9 28917.7 28259.9 

Trends in PMRT Use

For low-risk patients, PMRT use was essentially stable at 6% in 1992 compared with 6.9% in 2005 (Fig. 1A). For intermediate-risk patients, PMRT use increased linearly from 10.2% in 1992 to 21% in 2005 (Fig. 1B). The rate of change in PMRT use was 0.25% per quarter (95% confidence interval [CI], 0.21%-0.30% per quarter).

Figure 1.

The use of postmastectomy radiotherapy is illustrated between 1992 and 2005 for patients with (A) low-risk, (B) intermediate-risk, (C) high-risk, and (D) favorable high-risk breast cancer in the Surveillance, Epidemiology, and End Results-Medicare cohort. Squares indicate point estimates for each calendar year quarter. Error bars indicate the 95% confidence interval for the quarterly estimate. Solid lines indicate trends in PMRT use estimated by Joinpoint analysis. Vertical lines indicate publication dates of major evidence-based guidelines for PMRT: Line 1, American Society for Therapeutic Radiology and Oncology (Harris et al, 19994); Line 2, American Society of Clinical Oncology (Recht et al, 20015); Line 3, National Institutes of Health (Eifel et al, 20013); Line 4, National Comprehensive Cancer Network (Carlson et al, 20016).

For high-risk patients, receipt of PMRT was characterized by 3 distinct intervals (Fig. 1C). During the first interval, which extended from the first quarter of 1992 through the second quarter of 1996, PMRT use did not significantly change (rate of change, 0.25% per quarter; 95% CI, -0.22% -0.72% per quarter). During the second interval, which extended from the second quarter of 1996 through the second quarter of 1998, PMRT use rose significantly from 36.5% (95% CI, 26%-46.9%) to 57.7% (95% CI, 46.9%-68.4%) at the rate of 2.34% per quarter (95% CI, 0.40%-4.28% per quarter). During the third interval, which extended from the second quarter of 1998 to the end of the study period in the fourth quarter of 2005, PMRT use did not change significantly (rate of change, 0.01% per quarter; 95% CI, -0.16% -0.19% per quarter) despite the publication of multiple evidence-based guidelines early in this period. The annual percentage change in PMRT use differed significantly for the second interval compared with the first interval (P < .05) and for the third interval compared with the second interval (P < .05). Similar results were noted for high-risk patients who had a Charlson comorbidity index of 0 or 1 and were ages 66 to 79 years (the group for whom PMRT is most likely to be beneficial) (Fig. 1D). Interestingly, although there was no change in PMRT use for high-risk patients after PMRT guidelines were published, we observed that these guidelines were cited a total of 820 times in the academic literature.3-6

Risk Factors for PMRT Omission

Of 4978 women with high-risk breast cancer who were diagnosed between 1999 through 2005, 2729 women (54.8%) received PMRT. In multivariate analysis, patient factors that were associated with the omission of PMRT included advanced age and moderate to severe comorbidity (Table 3). The tumor factors that were associated with the omission of PMRT included smaller size, absence of lymph node involvement (eg, patients with T3N0 or T4N0 tumors), ductal histology (compared with lobular histology), and no T4 component. The only treatment factor that was associated with the omission of PMRT was the omission of chemotherapy as well. Undergoing breast reconstruction was not associated with the omission of PMRT. Model fit was acceptable (P = 0.10; area under the receiver-operator curve = 0.75). Among women with high-risk breast cancer who did not receive PMRT, the median survival was 4.5 years.

Table 3. Multivariate Model of High-Risk Patients Treated Between 1999 and 2005 (N=4975)a
VariableAdjusted OR for Omission of PMRT95% CIP
  • OR indicates odds ratio; PMRT, postmastectomy radiation; HSA, Health Service Area.

  • a

    The multivariate analysis also was adjusted for marital status and Surveillance, Epidemiology, and End Results registry (data not shown).

  • b

    The ratio of the number of radiation oncologists in the HSA to the HSA population.

 Age, y   
 Charlson comorbidity index   
 Median income in census tract or zip code   
  Lowest quartile1.00  
  Second quartile0.850.70-1.03.103
  Third quartile0.900.72-1.13.358
  Highest quartile0.890.68-1.17.412
 Adults with some college education in census tract or zip code   
   Lowest quartile1.00  
   Second quartile0.790.65-0.97.022
   Third quartile0.820.66-1.03.084
   Highest quartile0.850.66-1.10.222
Health services characteristics(1999-2005)   
 Radiation oncologist HSA ratiob   
  Lowest quartile1.00  
  Second quartile0.830.58-1.19.316
  Third quartile0.820.59-1.13.228
  Highest quartile0.790.57-1.09.154
Tumor characteristics   
 Tumor size, cm   
  2.0 to ≤5.00.860.69-1.06.154
  Unknown/not applicable0.770.52-1.14.190
 Tumor histology   
 Estrogen receptor status   
 No. of positive lymph nodes   
  0 (All negative)1.00  
  None examined/not specified/unknown1.040.77-1.41.789
 T4 tumor classification   
Treatment characteristics   
 Receipt of chemotherapy   
 Breast reconstruction   


In this study, we observed that 45.2% of older women who were diagnosed with high-risk breast cancer between 1999 and 2005 did not receive potentially life-saving PMRT despite the publication of 4 major guidelines between 1999 through 2002 recommending its use.3-6 Paradoxically, PMRT use did increase in intermediate-risk patients through 2005, albeit at a modest pace, even though the published guidelines did not strongly recommend PMRT for such patients. Our observations demonstrate the failure of evidence-based guidelines to satisfy their intended goal of summarizing and disseminating clinical evidence to everyday practice.

In a previous publication,11 we reported that the use of PMRT in postmenopausal women increased sharply after publication of the premenopausal Danish Breast Cancer Group (DBCG) 82b and British Colombia Cancer Agency clinical trials.7, 8 However, after peaking at 66% in 1997, the use of PMRT in postmenopausal women subsequently stabilized at 53% and failed to increase in response to the postmenopausal DBCG 82c trial, the results of which were published in 1999.9 The failure of PMRT use in postmenopausal women to increase in response to the DBCG 82c trial underlined a disconnection between trial publication and everyday practice, in effect illustrating the gap that evidence-based guidelines are meant to bridge. Unfortunately, our current results suggest that radiation oncology guidelines regarding indications for PMRT have not successfully bridged this gap between scientific evidence and everyday clinical practice. In contrast, Punglia and colleagues reported that PMRT was received by 83.6% of high-risk patients who were treated at National Comprehensive Cancer Network (NCCN) institutions, underlining the large differential in adoption between specialized cancer centers and broader clinical practice as reflected in our SEER-Medicare-derived cohort.1

Several possibilities may explain the apparent inability of evidence-based guidelines to yield their intended result in this specific clinical situation. First, the omission of PMRT in high-risk patients is likely to be multifactorial, and PMRT may have been appropriately contraindicated in certain patients. However, 3 observations argue against the hypothesis that rational clinical decision-making alone can explain the observed poor compliance with PMRT guidelines. First, we observed that use of PMRT in the relatively younger and healthier group of high-risk patients also did not increase after the publication of evidence-based guidelines. Second, patients in the high-risk group with smaller tumors yet extensive lymph node involvement (T1/T2 N2/N3), larger tumors yet limited lymph node involvement (T3/T4 N1), or ductal (as opposed to lobular) histology were less likely to receive PMRT, although there is little evidence-based justification for the omission of PMRT in patients with these clinical factors.35, 36 Finally, the median survival for the high-risk patients in our study who survived the first year of their cancer diagnosis and did not receive PMRT was 4.5 years, which suggests that there is a subset of patients who are not currently receiving PMRT but whose life expectancy is long enough to derive at least a locoregional control, if not overall survival, benefit from PMRT. Our findings suggest that the exclusion of PMRT in almost half of older patients with high-risk breast cancer in the United States cannot be explained by rational clinical decision-making alone and that, indeed, many of these women are receiving suboptimal care.

An alternative explanation is that, in many instances, access to the radiation resources required to perform PMRT is deficient. There is evidence for this mechanism, because investigators have demonstrated that variables like the distance to a radiation oncology facility influence the receipt of PMRT by elderly patients.37 Other published literature underscores this point; for example, there is higher likelihood of PMRT guideline compliance at NCCN institutions in which multidisciplinary care, including radiotherapy, is readily available.1 Moreover, Jagsi et al reported in an analysis of SEER patients in Detroit and Los Angeles that active surgeon referral to radiation specialists was an important predictor of appropriate PMRT receipt.38 However, it is unlikely that access is the sole determinant for PMRT omission. Although we did not have accurate data for measuring distance between patients and radiation facilities, we observed that other surrogate factors usually implicated in radiation access, including neighborhood education level,39 race,40 income, and density of radiation oncologists, did not significantly influence rates of PMRT omission in our multivariate analysis.38, 41

Additional research is required to explain the apparent failure of evidence-based guidelines to impact the adoption of PMRT for older women with high-risk breast cancer. One important mechanism may be reliance on passive dissemination for raising awareness of guidelines in treating and referring physicians. In a thorough review of both successful and unsuccessful instances of guideline implementation, Smith and Hillner observed that successful guidelines promoted both active dissemination of their content and accountability for guideline adherence.2 For example, since the mid-1970s, the province of British Columbia has systematically released guidelines for the treatment of lymph node-negative breast cancer and monitored compliance through internal audits. Consequently, in British Columbia, the rate of compliance with adjuvant radiotherapy guidelines was excellent at 97%.42 Likewise, a program of active dissemination and accountability improved breast cancer guideline compliance in France from 19% to 54%.43

In contrast, the United States Community Hospital Oncology Program, which was enacted in 1987, released guidelines through local mailings without an additional plan for implementation or accountability. Those guidelines failed to influence practice in breast cancer; for example, only 33% compliance was reported with the recommendation to record cancer stage.44 A parallel effort in Italy, also in 1987, likewise lacked explicit feedback on physician performance and, subsequently, had no impact on care patterns.45

Given these historic examples, we speculate that measures to improve accountability with PMRT guidelines may improve PMRT use in appropriate patients. To date, certain strategies already are taking root. For example, the American College of Surgeons Commission on Cancer (CoC) explicitly ties their accreditation to adherence to evidence-based guidelines.46 The CoC already tracks institutional use of radiotherapy after breast-conserving surgery for women aged ≤70 years who are diagnosed with invasive breast cancer, and we believe that compliance with PMRT should be added as an additional quality measure. Similarly, the National Quality Forum already tracks 4 quality measures in breast cancer and should consider adding compliance with PMRT recommendations as a fifth measure. Finally, payers with widespread networks have an opportunity to influence evidence-based practice with financial incentives.

The findings of this study are limited to patients aged >65 years who reside within a SEER registry and, thus, may not apply to other populations. In addition, the data cannot discriminate between physician decision-making and patient preferences. This is an important point, because many patients choose mastectomy with the specific bias of wishing to avoid radiotherapy; therefore, it is possible that the omission of PMRT in some candidates may be attributable to this phenomenon rather than to the deficient adoption of clinical guidelines by practicing physicians.38 Another limitation is that our study did not explicitly measure passive dissemination and begs the question whether physicians were exposed adequately to the evidence-based guidelines referenced in this study. Although there is no definitive way to measure dissemination, we note that, since their publication, the 4 guidelines referenced in the literature have been cited cumulatively 820 times in the academic literature, which suggests that lack of dissemination alone is unlikely to fully account for our observations.12

In conclusion, despite the publication of 4 major cancer PMRT guidelines, nearly 50% of older women with high-risk breast cancer still do not receive PMRT. Research is needed to develop novel methods that promote adherence to evidence-based guidelines.


A portion of this work was funded by a research grant from Varian Medical Systems.


The authors made no disclosures.