Long-Term Impact of CALGB 9343 on Radiation Utilization

Abstract

Background:

The results of the Cancer and Leukemia Group B (CALGB) 9343 trial showed that radiation therapy (RT) did not improve survival for women older than 70 y with early-stage estrogen receptor + breast cancer treated with breast conserving surgery and adjuvant endocrine therapy. In 2005, guidelines were modified to allow for RT omission; however, minimal change in clinical practice has occurred. The aim of this study was to determine if CALGB long-term follow-up data have affected RT utilization, and to characterize the population still receiving RT after breast conserving surgery.

Methods:

The Surveillance, Epidemiology, and End ResultseMedicare database was used to identify women diagnosed with early-stage breast cancer from 2004 to 2015 who matched the CALGB 9343 inclusion criteria. Multivariate logistic regression was carried out to identify the factors that affect the receipt of radiation therapy. We also plotted the overall use of RT over time juxtaposed with the temporal trends of CALGB 9343 clinical trial data, guideline recommendations, and publishing of long-term survival data.

Results:

The study cohort included 25,723 Medicare beneficiaries, of whom 20,328 (79%) received RT and 5395 (21%) did not receive RT. In a multivariate model, the frequency of RT omission increased over time, with those diagnosed in year 2015 being 2.72 times more likely to omit RT compared with those diagnosed in 2004 (95% confidence interval 2.31–3.19).

Conclusions:

This study investigated the impact of long-term CALGB 9343 data on clinical practice. The results of this study support results from previous studies, extend the dates of analysis, and indicate that after long-term follow-up of CALGB 9343 data, RT was less used, but overall trends did not dramatically decrease.

Introduction

The standard of care for most women diagnosed with early-stage breast cancer is breast conserving therapy, which consists of breast conserving surgery (BCS) and adjuvant radiation therapy (RT).¹ For patients with estrogen receptor (ER) positive tumors, treatment also typically includes 5–10 y of adjuvant endocrine therapy (AET). The risk of developing breast cancer increases with age, and is most common in women older than 70 y of age.² In formulating a treatment plan for older patients diagnosed with early-stage ER + breast cancers, the goal is to avoid both over- and under-treatment of the disease, while taking into consideration patient age and comorbidities. This is further influenced by patients’ preference, and impact of treatment on the patients’ quality of life.

The Cancer and Leukemia Group B (CALGB) 9343 investigators attempted to address the necessity of RT in women older than 70 y treated with breast conserving therapy and AET by randomizing women into RT and no RT arms after BCS. All women in the trial were recommended AET. The results of this randomized controlled trial, first published in 2004, suggest that for women at low risk of tumor recurrence, RT improves the 5-y locoregional recurrence rates (1% for RT versus 4% for no RT), but does not confer a survival advantage. Forgoing RT did not impact mortality or disease-free survival, and did not increase the rates of mastectomy for patients with a cancer recurrence.³ In 2005, the National Comprehensive Cancer Network (NCCN) modified practice guidelines to reflect these results, stating that adjuvant RT could be safely omitted for this patient population.⁴ In 2013, the 10-y survival data of a CALGB 9343 were published, and again demonstrated no significant difference in survival or recurrence when RT was omitted in this population. Breast cancer was the cause of death in only 6% of deceased women in the trial.⁵

Despite the NCCN guideline modification and the long-term results of CALGB 9343, several studies have shown minimal change to RT utilization in clinical practice, many citing continued usage in greater than two-thirds of women in this patient population.^6,7 In 2012, Soulos et al.⁶ performed an analysis using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database to assess the use of RT in women older than 70 y diagnosed with stage I ER + breast cancer between 2001 and 2007. Of the 12,925 patients in their sample, 76.5% received RT; 79% of women received RT before the publication of CALGB 9343, whereas 75% after publication. The authors concluded that CALGB 9343 had minimal impact on RT use, although the overall rates of RT did decrease. Subsequently, the PRIME II trial published in 2015 produced additional evidence that forgoing adjuvant RT among older women who were taking AET was safe with very similar 5-y survivals reported between the RT and no-RT group.⁸ Using the updated SEER-Medicare data, we sought to determine the impact of CALGB 9343 on RT utilization in the modern era, and characterize the population that receives RT after BCS.

Materials and methods

Data sources

This study was approved by the institutional review board. This study used the SEER-Medicare database, of which approximately 55% of the patients are older than 65 y and 94% have been successfully linked with their Medicare claims.^9–11 The database includes claims for Medicare beneficiaries from 2007 to 2016 and is linked to cancer registry data in 17 US geographic areas.¹² The analytic data points that are in the SEER-Medicare database include patient demographics at the time of diagnosis, information regarding Medicare enrollment, Healthcare Common Procedure Coding System codes, Current Procedural Terminology and Clinical Modification procedure codes and diagnoses, International Classification of Diseases, tumor characteristics, Ninth Revision, and prescription claims data.¹³

Study population

The study cohort was built using the SEER codes C50.0-C50.9 to identify all female patients aged 70 y and older with invasive breast cancer from 2004 to 2016. Only patients who were continuously enrolled in Medicare parts A and B for 12 mo before diagnosis were included. All patients had BCS within 9 mo of the breast cancer diagnosis, and did not have a non-breast cancer diagnosis from initial diagnosis through the first postoperative year. Surgery was first identified from the registry and confirmed from Medicare claims. Tumor specific inclusion criteria refined the cohort to include only patients with tumors that were known to be ER+, 2 cm or less in size, and lymph node pathologically negative or unknown. Patients who had confirmed nodal disease on pathology were excluded from the study. Those patients who did not have a sentinel lymph node biopsy, but had clinically negative nodes, were included. In addition, patients with Her2+ tumors were excluded to match the inclusion criteria of the CALGB 9343 clinical trial. Breast cancer diagnosis was confirmed not to have been discovered from the death report alone.

Variable construction

The following SEER cancer registry database variables were used for this analysis: year of diagnosis, age at the time of diagnosis, race (classified as either white or other), marital status (classified as either single or married), geographic region (Midwest, Northeast, South, or West), and clinical tumor characteristics including stage, tumor size (cm), and lymph node status (negative or unknown). The receipt of adjuvant RT was identified searching claims for Healthcare Common Procedure Coding System. A patient was considered to have received RT if the treatment was received within 9 mo of BCS. The proportion of people living below the poverty line at the census tract level from the 2000 census for cases in 2004–2009, and the American Community Survey 2008–2012 for cases in 2010–2015, available in the SEER-Medicare linked data, were used to define socioeconomic status (SES). This variable was broken into quartiles (7.75% or less of the population living below the poverty line, 7.76%−13.27%, 13.28%−17.15%, and 17.16% or more) with the first quartile representing the highest SES and the fourth quartile representing the lowest SES (greatest level of poverty).

Year of diagnosis and RT status were the variables used for the temporal trends and the multivariate analysis. The multivariate analysis included the following covariates: comorbidity status in the year before breast cancer diagnosis, age (75–79, 80–84, 85+, versus 70–75 y), race, marital status, size of tumor (0–1 cm versus 1–2 cm), lymph node status (unknown versus negative). Comorbidity status was rated using the Charlson Comorbidity Index (CCI) from available SAS macros.^14,15 The CCI was divided into four levels using the Klabunde et al¹⁶ modification. Patient comorbidities were assigned on a scale of 0, 1, 2, 3, and 4 or greater, and were compared. As described elsewhere, all the comorbidities were explored in the 1-y before breast cancer diagnosis in outpatient, physician, and hospital claims.¹⁶ To assess for differences between cohorts over time, the 2004 cohort was compared with the 2015 cohort using chi-square tests.

Results

The study cohort included 25,723 Medicare beneficiaries, of whom 20,328 (79%) received RT and 5395 (21%) did not receive RT (Table 1). There were group differences in those who received versus omitted RT. Specifically, in the group that omitted RT, there was a higher proportion of women aged 80 y and older, single women, women with tumors 0–1 cm, and women with two or more comorbidities (P < 0.001). In addition, the groups differed significantly in geographic region.

Table 1 –

Study sample characteristics (n = 25,723).

Characteristics	Group 1: radiation (n = 20,328)	Group 2: no radiation (n = 5395)	P value*
Age group, y			<0.0001
70–74	7887 (38.80)	1109 (20.56)
75–79	7406 (36.43)	1648 (30.55)
80–84	3908 (19.22)	1564 (28.99)
85+	1127 (5.54)	1074 (19.91)
Race			0.0612
White	18,399 (90.51)	4928 (91.34)
Other	1929 (9.49)	467 (8.66)
Marital status			<0.0001
Single/Other	10,463 (51.47)	3271 (60.63)
Married	9865 (48.53)	2124 (39.37)
Tumor size, cm			<0.0001
0–1	7920 (38.96)	2400 (44.49)
1–2	12,408 (61.04)	2995 (55.51)
Lymph node status			0.3926
Negative	120,316 (99.94)	5390 (99.91)
Unknown	12 (0.06)	5 (0.09)
Geographic region			<0.0001
Midwest	2557 (12.58)	543 (10.06)
Northeast	4824 (23.73)	1077 (19.96)
South	4079 (20.07)	1120 (20.76)
West	8868 (43.62)	2655 (49.21)
Socioeconomic status			0.0207
1st quartile	5690 (27.99)	1517 (28.12)
2nd quartile	4970 (24.45)	1215 (22.52)
3rd quartile	4831 (23.77)	1316 (24.39)
4th quartile	4837 (23.79)	1347 (24.97)
Year of diagnosis			<0.0001
2004	1699 (8.36)	282 (5.23)
2005	1607 (7.91)	361 (6.69)
2006	1674 (8.23)	368 (6.82)
2007	1697 (8.35)	359 (6.65)
2008	1687 (8.30)	408 (7.56)
2009	1764 (8.68)	387 (7.17)
2010	1729 (8.51)	421 (7.80)
2011	1806 (8.88)	484 (8.97)
2012	1651 (8.12)	494 (9.16)
2013	1688 (8.30)	525 (9.73)
2014	1651 (8.12)	627 (11.62)
2015	1675 (8.24)	679 (12.59)
CCI			<0.001
0	12,757 (62.76)	2969 (55.03)
1	4747 (23.35)	1387 (25.71)
2+	2824 (13.89)	1039 (19.26)

^*P value is for a χ² test between each covariate and group.

Figure demonstrates the temporal trends of the use of RT by age group alongside the year that the CALGB 9343 study was published (2004), the year of the announcement and publication of the revised NCCN guidelines (2005), and the year CALGB 9343 long-term survivorship results were published (2013). There was a noted downward trend of women who received RT from the years 2004 to 2015.

Fig –

Temporal trend in the use of radiation therapy by age group. (A) September 2004, Cancer and Leukemia Group B (CALGB) 9343 study was published; (B) March 2005, revised NCCN guidelines were announced; (C) November 2005, revised NCCN guidelines were published; (D) July 2013, CALGB 9343 long-term results published. ***P < 0.001.

In the multivariate model (Table 2), those aged 70–74 y were 85% less likely to have omitted RT compared with those aged 85 y and older (confidence interval [CI] 0.13–0.16). Furthermore, white women (odds ratio [OR] 1.22, CI 1.09–1.36), women who were single (OR 1.20, CI 1.12–1.28), had tumors 0–1 cm (OR 1.39, CI 1.30–1.48), and had two or more comorbidities (OR 1.45, CI 1.33–1.59) were more likely to have omitted RT. Finally, the likelihood that women in this sample omitted RT increased over time, with those diagnosed in year 2015 having been 2.72 times more likely to have omitted RT compared with those diagnosed in 2004 (CI 2.32–3.20).

Table 2 –

Multivariate model–omission of RT.

Characteristics	Point estimate	95% Wald confidence limits
Age, y
70–74 versus 85+	0.144	0.129–0.161
75–79 versus 85+	0.234	0.211–0.259
80–84 versus 85+	0.429	0.386–0.476
Race
White versus others	1.219	1.089–1.363
Marital status
Single/Other versus married	1.197	1.121–1.277
Tumor
0–1 cm versus 1–2 cm	1.386	1.300–1.478
Geographic region
Midwest versus West	0.661	0.59–0.737
Northeast versus West	0.741	0.682–0.806
South versus West	0.941	0.865–1.025
Socioeconomic status
2nd versus 1st quartile	0.869	0.794–0.951
3rd versus 1st quartile	1.008	0.921–1.102
4th versus 1st quartile	1.013	0.926–1.108
Year
2005 versus 2004	1.322	1.108–1.577
2006 versus 2004	1.291	1.083–1.539
2007 versus 2004	1.271	1.066–1.517
2008 versus 2004	1.415	1.191–1.681
2009 versus 2004	1.280	1.075–1.523
2010 versus 2004	1.432	1.206–1.700
2011 versus 2004	1.617	1.368–1.913
2012 versus 2004	1.794	1.517–2.122
2013 versus 2004	1.970	1.668–2.327
2014 versus 2004	2.588	2.199–3.046
2015 versus 2004	2.723	2.317–3.199
CCI
1 versus no comorbidities	1.208	1.120–1.303
2+ versus no comorbidities	1.453	1.332–1.585

There were significant differences between the earliest (2004) and latest (2015) cohorts (Table 3). The women in the later cohort were of younger age, lower SES, and had less comorbidities overall.

Table 3 –

Study sample characteristics - Year 2004 versus Year 2015

Characteristics	Year 2004 (n = 1981)	Year 2015 (n = 2354)	P value
Age group, y			<0.0001
70–74	642 (32.41)	973 (41.33)
75–79	743 (37.51)	763 (32.41)
80–84	453 (22.87)	429 (18.22)
85+	143 (7.22)	189 (8.03)
Race			0.0052
White	1826 (92.18)	2112 (89.72)
Other	155 (7.82)	242 (10.28)
Marital status			0.0111
Single/Other	1109 (55.98)	1227 (52.12)
Married	872 (44.02)	1127 (47.88)
Tumor size, cm			0.2046
0–1	747 (37.71)	932 (39.59)
1–2	1234 (62.29)	1422 (60.41)
Geographic region			0.0022
Midwest	248 (12.52)	262 (11.13)
Northeast	506 (25.54)	515 (21.88)
South	378 (19.08)	530 (22.51)
West	849 (42.86)	1047 (44.48)
Socioeconomic status			<0.0001
1st quartile	837 (42.25)	587 (24.94)
2nd quartile	340 (17.16)	627 (26.64)
3rd quartile	438 (22.11)	595 (25.28)
4th quartile (low)	366 (18.48)	545 (23.15)
CCI			<0.0001
0	1260 (63.60)	1409 (59.86)
1	486 (24.53)	527 (22.39)
2+	235 (11.86)	418 (17.76)

Discussion

The results of our study, which analyzed modern SEER-Medicare data, suggest that in the years after the publication of the CALGB 9343 clinical trial and the change in NCCN guidelines, RT utilization in women older than 70 y with early-stage ER + breast cancer is trending downward. However, there persists a significant proportion of women in this population who continue to be treated with RT, revealing minimal change in clinical practice. Our study looked at RT use for each year compared to 2004—when the CALGB 9343 study was first publishe—dto look at its impact on clinical practice. Additional factors found to be associated with the omission of RT were age, marital status, tumor size, and comorbidity index. We also found significant regional differences between those treated and not treated with RT.

The relatively minor change in RT utilization over time is likely multifactorial in nature, and related to sociodemographic status, patient and tumor characteristics, and physician recommendations. The rate of RT after treatment with BCS varies in the literature from 65% to 95%.¹⁷ A prior analysis of the SEER database found that approximately 20% of the women with stage I cancer diagnosed from 1988 to 2007 and treated with BCS did not receive RT.¹⁸ This proportion did not change over time. In this previous study, increasing age, small tumor size, African American race, single/divorced marital status, and ER negativity were all significant predictors of not receiving RT.

Furthermore, Chu et al.⁷ investigated the sociodemographic variation of adjuvant RT use in the post-CALGB 9434 era. Using the National Cancer Database, they analyzed a cohort of 120,308 women older than 70 y with early-stage breast cancer diagnosed between 2004 and 2012, and found significant variation in RT usage across sociodemographic strata.⁷ The variables they found to be associated with increased odds for omission of RT included advanced age, African American race, Medicaid recipients, increased great circle distance, therapy under academic research program, residents of East-South-Central region, living in a rural population <2500 not adjacent to a metropolitan area, and low-income level. Contrary to their results, our study suggests that white women are less likely to receive RT. This discrepancy may be explained by the utilization of different databases and more recent data in our study. The SEER program includes registries that cover about 26% of the US population, in comparison to the National Cancer Database that represents approximately 70% of newly diagnosed cancer cases.^19,20 Furthermore, the SEER areas are limited to only a fraction of defined geographic areas, thereby making conclusions about sociodemographic variables difficult to generalize and compare. Our results are consistent with Soulos et al.,⁶ who also analyzed the SEER database in 2012, and reported an increase in the use of RT in African American women from 69.3% before the study to 74.7% after the study. Almost a full decade later, there has been very little change in radiation use in this cohort examined via the same database.

Our results suggest that women with smaller tumor size and greater number of comorbidities were less likely to receive RT, and this finding is consistent across multiple studies.^7,21,22 Some studies of smaller samples do not support the impact of tumor size on the utilization of RT. Christian et al.²³ analyzed a single-center tumor registry for all patients older than 70 y with ER + stage I breast cancer from 2009 to 2017 and found no association with tumor grade and size regarding the use of RT. In their study, only age was found to be predictive of being treated with RT after publication of the CALGB 9343 data. The lack of significance of tumor characteristics in their study is likely explained by their small sample size.

There are additional physician-related reasons as to why the omission of RT for women older than 70 y with early-stage breast cancer has not been integrated into clinical practice. The medical community may be more likely to incorporate new guidelines that suggest adding a treatment, and less likely to withhold one.⁸ This is further complicated by financial incentives that favor treatment over no treatment, especially when both are suitable options.⁸ In addition, many surgeons and radiation oncologists may be uninformed of the current guidelines and the lack of benefit of RT in this patient population. It is frequently stated that it takes an average of 17 y for research evidence to reach clinical practice.²⁴ Shumway et al.²⁵ conducted a nationwide survey of radiation oncologists and surgeons in 2017 to evaluate physician attitudes, knowledge, communication, and recommendations regarding RT omission in older women with early-stage breast cancer. Of the 825 responses they received, 40% surgeons and 20% radiation oncologists felt that omission of RT in patients older than 70 y with stage I ER + breast cancer, treated with BCS and AET was unreasonable. Those with these views were found to have lower breast cancer patient volume, greater number of years in practice, and less involvement in multidisciplinary tumor boards. Many mistakenly associated RT in this population with increased survival, and overestimated the risk of locoregional recurrence in older women with omission of RT. The strongest correlates of acceptance of RT omission were knowledge regarding lack of survival benefit and correct estimation of life expectancy.²⁵ Quite often, it is the surgeon who is the first physician to discuss treatment options, and a surgeon’s recommendation has been found to be the most influential factor affecting older women’s treatment decisions.²⁶ In a frequently changing surgical specialty, it is essential for surgeons to be knowledgeable of current guidelines for the treatment of breast cancer to provide the best evidence-based patient care.

There are limitations in the present analysis. Although using the SEER database provides strength in the number of patients, the data are retrospective and subject to the limitations of the SEER database. Our comorbidity assessment is based on Medicare claims, which are collected for administrative, rather than clinical purposes. Furthermore, patient preferences are likely to play a significant role in the treatment planning, and are unable to be captured in the SEER database.

This is the most recent study to investigate the temporal trends of the translation of the CALGB 9343 into clinical practice decisions. The results of this study highlight the need for a personalized approach to treatment planning in women older than 70 y with early-stage breast cancer. When applying the results of the CALGB 9343 study, it is important to note that their conclusions were drawn in a clinical trial setting where adherence to AET can be assumed to be higher than outside of this setting. Studies have suggested that increasing age showed significant association to not receive AET.^27–29 It is prudent to incorporate AET adherence to help with the RT decision-making, as nonadherence to AET could result in undertreatment of breast cancer, and thus reconsideration of RT should be warranted in the initial postsurgical period.