The Impact of Radiotherapy Delay in Breast Conservation Patients Not Receiving Chemotherapy and the Rationale for Dichotomizing the Radiation Oncology Time-Dependent Standard into Two Quality Measures

Abstract

Introduction:

The Commission on Cancer/National Quality Forum breast radiotherapy quality measure establishes that for women <70 years, adjuvant radiotherapy after breast conserving surgery (BCS) should be started <1 year from diagnosis. This was intended to prevent accidental radiotherapy omission or delay due to a long interval between surgery and chemotherapy completion, when radiation is delivered. However, the impact on patients not receiving chemotherapy, who proceed from surgery directly to radiotherapy, remains unknown.

Patients and Methods:

Patients aged 18–69, diagnosed with stage I-III breast cancer as their first and only cancer diagnosis (2004–2016), having BCS, for whom this measure would be applicable, were reviewed from the National Cancer Database.

Results:

Among 308,521 patients, the median age was 57.0 years, and >99% of all patients were compliant with the measure. The cohort of interest included 186,650 (60.5%) patients not receiving chemotherapy, with a mean age of 57.9 years. 90.5% received external beam radiotherapy (EBRT) and 9.5% brachytherapy. Among them, 24.9% started radiotherapy >8 weeks after surgery. In a multivariable model, delay from surgery to radiotherapy increased the hazard ratios for overall survival to 9.0% (EBRT) per month and 3.0% (brachytherapy) per week.

Conclusion:

While 99.9% of patients undergoing BCS without chemotherapy remain compliant with the current quality measure, 25% have delays >8 weeks to start radiation, which is associated with impaired survival. These data suggest that the current quality measure should be dichotomized into two, with or without chemotherapy, in order to impel prompt radiotherapy initiation and maximize outcomes in all patients.

INTRODUCTION

Time-dependent quality measures in breast cancer have been in existence for several years, but only three have been universally adopted; one for prompt chemotherapy administration, one for endocrine therapy administration, and one for radiotherapy (RT) administration.¹ The last establishes that for women under 70 years of age, adjuvant RT after breast conserving surgery (BCS) should be initiated within one year (365 days) of diagnosis. This quality measure was intended to prevent accidental omission or delay of RT due to the long interval between surgery and completion of chemotherapy, when radiation is typically delivered. It was initially put forth by the American College of Surgeons Commission on Cancer and has been endorsed and adopted by multiple accrediting bodies.^1–7

The rationale for having time-dependent quality measures arises from significant data suggesting that times to treatment can affect outcomes.⁸ RT after BCS provides an improvement in local recurrence and survival,⁹ but increasing times between surgery and RT have been shown to impair that benefit, with the impact of delays varied between studies.^10–14

Compliance with the current measure allows for RT to begin up to one year from diagnosis (rather than measuring time from the last modality), and accounts for the potential time required for chemotherapy administration prior to radiation. If surgery is completed within 90 days of diagnosis,¹⁵ and chemotherapy is begun within 120 days of diagnosis as per the current standard, that leaves 245 days to complete chemotherapy and finish RT simulation and planning. Common chemotherapy regimens vary in length, between 84 days for docetaxel and cyclophosphamide (TC), to 168 days for the cyclophosphamide, methotrexate, 5-fluorouracil (CMF) regimen. If the latter is given, that provides 8 weeks to complete simulation and treatment planning which can occupy 4–6 weeks prior to initiating the delivery of the first fraction, barely leaving 2 weeks to spare. If chemotherapy is not given, however, and surgery is completed within 90 days of diagnosis, there remain 275 days as per the standard to begin radiation administration, allowing 9 months of delay for initiating RT while still be considered compliant for this measure (Figure 1).

Figure 1.

STROBE diagram of inclusions and exclusions.

This study was performed to assess the impact of delaying RT after BCS, specifically in patients who do not receive chemotherapy, to determine its impact on long term outcomes.

METHODS

Patient data from the Participant User Data File (PUF) of the National Cancer Database were analyzed with permission from the American College of Surgeons; the need for informed consent was waived by the Fox Chase Cancer Center institutional review board. All groups of fewer than 10 individuals or groups allowing calculation of such groups were censored as per the privacy policy of the NCDB.

The inclusion criteria were governed by the specifications of the radiation quality measure. Specifically, all female patients between 18 and 69 years of age, who had undergone BCS and RT to the breast were included. Patients were limited to those having a diagnosis of non-nonmetastatic, noninflammatory, invasive disease as their first and only breast cancer. Patients who received neoadjuvant therapies, ductal carcinoma in-situ, or unknown entries for time parameters in the database were excluded (Figure 1). Because transfers between institutions have been found to confer significant delays,¹⁶ we included this variable as “surgery elsewhere.”

Demographic, tumor and treatment characteristics were compared by chemotherapy status using Chi-square tests and two sample t-tests. Our primary focus was to examine the interval from surgery to start of radiation in those who did not receive chemotherapy. We analyzed cohorts by type of RT (external beam radiotherapy (EBRT) or brachytherapy. The EBRT category included those not otherwise specified. Brachytherapy included any radioactive sources that delivered therapeutic doses to the tumor bed with a 2–3 cm margin (e.g. “seed implants” or “intracavitary radioisotopes”). Chi-square and two-sample t-tests were utilized to detect significant differences between the patient characteristics. We examined surgery-RT interval as both a continuous variable and a categorical variable. For categories, we used 15 or 30 day intervals starting at the approximate median to describe the right handed tail of the interval distribution. We identified characteristics associated with median surgery-RT interval using quantile regression models. Due to the limitations of the NCDB in documenting recurrence or cause-specific mortality in the PUF dataset, all-cause mortality was used as the primary outcome for this analysis. Overall survival (OS) was measured from date of definitive surgery and OS curves were generated using Kaplan Meier methods. The hazard ratios for the surgery-RT interval (as either a continuous or categorical variable) were calculated using Cox proportional hazards regression with robust standard errors to account for clustering within facilities. Two models were created to account for relevant co-variates, the first included age, race, and Charlson comorbidity index, and the second further adjusted for urban/rural setting, transfer of care, facility volume, ER/PR status, tumor size, lymph node positivity, grade, endocrine therapy, and year of diagnosis (as a restricted cubic spline). In a sensitivity analysis, socioeconomic variables (income, education, type of insurance) were included in the model. Additional sensitivity analyses with time varying covariates or stratification were used to assess the impact of proportional hazards assumption violation in the EBRT Cox models. Analyses were performed using STATA software, release 15 (StataCorp, College Station, TX), and SAS software, version 9.4 (SAS institute, Cary, NC).

RESULTS

Of the 308,521 patients undergoing BCS and radiation in the study period, 186,650 (60.5%) did not receive chemotherapy, thus comprising the cohort of interest. The remainder of the patients (121,871, 39.5%) received chemotherapy and served as the reference cohort. (Figure 1). Characteristics of both cohorts are detailed in Table 1 and Supplemental Table 1. Histologic subtypes are listed in Supplemental Table 2. A higher proportion of Blacks received chemotherapy than Whites (54.0% vs 37.7%), as did Hispanics (45.6% vs 39.0%), while chemotherapy administration was less likely as Charlson comorbidity scores increased. Chemotherapy administration was given in similar proportions regardless of distance to facility, while 76% of those with high-grade tumors received chemotherapy as versus 14.5% with low-grade tumors, and 85.9% with triple negative tumors received chemotherapy as versus 25.1% with receptor-positive, HER2-negative tumors. Receipt of chemotherapy also increased from 23.1% of Stage I patients to 96.2% of Stage III patients.

Table 1.

Characteristics by chemotherapy treatment. P-values are from Chi-square tests.

			Chemo Status
	ALL		N		Y		Different by Chemo?
	N	Col%	N	Col%	N	Col%	p-value
ALL	308,521		186,650		121,871
Age at Dx							<0.0001
18–29	752	0.2	133	0.1	619	0.5
30–39	9,640	3.1	2,285	1.2	7,355	6.0
40–49	62,300	20.2	29,123	15.6	33,177	27.2
50–59	110,581	35.8	64,782	34.7	45,799	37.6
60–69	125,248	40.6	90,327	48.4	34,921	28.7
Race							<0.0001
White	261,785	84.9	163,017	87.3	98,768	81.0
Black	31,800	10.3	14,641	7.8	17,159	14.1
Asian	9,087	2.9	5,542	3.0	3,545	2.9
Other/Unknown	5,849	1.9	3,450	1.8	2,399	2.0
Hispanic							<0.0001
No	278,289	90.2	169,870	91.0	108,419	89.0
Yes	13,410	4.3	7,290	3.9	6,120	5.0
Unknown	16,822	5.5	9,490	5.1	7,332	6.0
Charlson Score							<0.0001
0	270,305	87.6	162,736	87.2	107,569	88.3
1	32,237	10.4	20,010	10.7	12,227	10.0
2	4,843	1.6	3,133	1.7	1,710	1.4
≥3	1,136	0.4	771	0.4	365	0.3
Setting							0.0021
Large metropolitan	163,361	52.9	99,068	53.1	64,293	52.8
Metropolitan	96,463	31.3	58,304	31.2	38,159	31.3
Urban	16,494	5.3	9,905	5.3	6,589	5.4
Less Urban/Rural	23,962	7.8	14,281	7.7	9,681	7.9
Unknown	8,241	2.7	5,092	2.7	3,149	2.6
Distance to Tx Facility							<0.0001
≤10	170,725	55.3	103,067	55.2	67,658	55.5
11–20	75,260	24.4	45,664	24.5	29,596	24.3
21–40	38,254	12.4	23,028	12.3	15,226	12.5
>40	23,351	7.6	14,388	7.7	8,963	7.4
Unknown	931	0.3	503	0.3	428	0.4
Facility Volume Quartile							<0.0001
1	14,310	4.6	8,442	4.5	5,868	4.8
2	40,084	13.0	23,245	12.5	16,839	13.8
3	75,598	24.5	45,568	24.4	30,030	24.6
4	178,529	57.9	109,395	58.6	69,134	56.7
Surgery Elsewhere							<0.0001
No	247,054	80.1	151,126	81.0	95,928	78.7
Yes	61,467	19.9	35,524	19.0	25,943	21.3
Year of Diagnosis							<0.0001
2004	18,477	6.0	9,017	4.8	9,460	7.8
2005	19,865	6.4	10,085	5.4	9,780	8.0
2006	21,620	7.0	11,018	5.9	10,602	8.7
2007	22,398	7.3	11,860	6.4	10,538	8.6
2008	23,141	7.5	12,850	6.9	10,291	8.4
2009	24,617	8.0	14,076	7.5	10,541	8.6
2010	26,002	8.4	15,299	8.2	10,703	8.8
2011	28,876	9.4	17,471	9.4	11,405	9.4
2012	29,910	9.7	19,062	10.2	10,848	8.9
2013	29,938	9.7	20,354	10.9	9,584	7.9
2014	31,022	10.1	21,884	11.7	9,138	7.5
2015	32,655	10.6	23,674	12.7	8,981	7.4
Histology							<0.0001
Ductal	282,688	91.6	169,592	90.9	113,096	92.8
Lobular	22,383	7.3	16,089	8.6	6,294	5.2
Other	3,450	1.1	969	0.5	2,481	2.0
Grade							<0.0001
1	85,148	27.6	72,822	39.0	12,326	10.1
2	130,245	42.2	85,501	45.8	44,744	36.7
3	78,979	25.6	18,974	10.2	60,005	49.2
Anaplastic/Undifferentiated	762	0.2	219	0.1	543	0.4
Unknown	13,387	4.3	9,134	4.9	4,253	3.5
Tumor Size							<0.0001
≤20mm	240,064	77.8	165,412	88.6	74,652	61.3
21–50mm	63,956	20.7	19,180	10.3	44,776	36.7
>50mm	2,148	0.7	633	0.3	1,515	1.2
Unknown	2,353	0.8	1,425	0.8	928	0.8
Number of Lymph Nodes Examined							<0.0001
1	69,366	22.5	49,711	26.6	19,655	16.1
2	70,977	23.0	49,394	26.5	21,583	17.7
3–4	79,560	25.8	52,893	28.3	26,667	21.9
≥5	88,618	28.7	34,652	18.6	53,966	44.3
Number of Lymph Nodes Positive							<0.0001
0	248,952	80.7	176,441	94.5	72,511	59.5
1–3	48,650	15.8	9,865	5.3	38,785	31.8
4–9	4,783	1.6	173	0.1	4,610	3.8
≥10	6,136	2.0	171	0.1	5,965	4.9
ER/PR status (Borderline incl with Positive)							<0.0001
Both Neg: ER-, PR-	43,193	14.0	7,349	3.9	35,844	29.4
ER+ and/or PR+ (one could be missing)	262,858	85.2	177,650	95.2	85,208	69.9
Both missing	2,470	0.8	1,651	0.9	819	0.7
ER/PR/HER2 status							<0.0001
Unknown - No HER2 2004–2009	130,118		68,906		61,212		(excl from p-value)
Triple Negative: ER-, PR-, HER2-	17,313	9.7	2,433	2.1	14,880	24.5
ER+ and/or PR+, HER2-	143,337	80.3	107,427	91.2	35,910	59.2
Any HER2+	14,025	7.9	5,044	4.3	8,981	14.8
Missing/Undetermined	3,728	2.1	2,840	2.4	888	1.5
NCDB Analytic stage							<0.0001
Stage I	207,904	67.4	159,836	85.6	48,068	39.4
Stage II	89,240	28.9	26,380	14.1	62,860	51.6
Stage III	11,377	3.7	434	0.2	10,943	9.0
Chemo							n/a
Yes	121,871	39.5	.	.	121,871	100.0
No	186,650	60.5	186,650	100.0	.	.
Endocrine							<0.0001
Yes	239,345	77.6	159,450	85.4	79,895	65.6
No	69,176	22.4	27,200	14.6	41,976	34.4

The overall compliance with the radiation quality measure for initiation of RT within one year for patients receiving chemotherapy vs. those not receiving chemotherapy was 99.37% (n=112,107) versus 99.93% (n=186,524), respectively (p<0.0001). The start of RT was >60 days for 22.82% (n=42,585) of patients not receiving chemotherapy (24.9% of EBRT patients vs. 2.8% of brachytherapy patients, p<0.0001).

Figure 2 illustrates histograms of the times between surgery and RT initiation in all patients combined, the subsets of those receiving or not receiving chemotherapy, and among the latter, those receiving EBRT and brachytherapy. Characteristics of those receiving EBRT and brachytherapy are shown in Supplemental Table 3. Intervals between diagnosis and RT are shown in Supplemental Table 4, and the frequency of intervals from definitive surgery to RT (stratified by EBRT vs. brachytherapy) are categorized in Supplemental Table 5.

Figure 2. Histograms of days from definitive surgery to the start of radiotherapy (RT).

Panel A. All patients combined. The x-axis for days from definitive surgery to start of RT is truncated at 365 days arbitrarily. This cut point is not the same as the Dx to RT compliance cutpoint.

Panel B. Patients subdivided by those not receiving (top) and receiving (bottom) chemotherapy

Panel C. Patients not receiving chemotherapy, subdivided by type of RT administered: external beam radiotherapy (EBRT) vs brachytherapy.

In patients receiving EBRT (n=168,933), multivariable analysis accounting for patient and tumor factors, as well as endocrine therapy administration, demonstrated a 9% increase in the hazard rate for death from all causes (HR=1.09, 95% CI 1.07–1.11, p<0.0001) per month of radiotherapy delay. For those receiving brachytherapy (n = 17,717), the increase in hazard rate for death was 3% (HR=1.03, 95%CI 1.00–1.05, p=0.0175) per week (Table 2, Supplemental Table 6). To illustrate using EBRT survival estimates for a difference of 2 months in the surgery-RT interval, Cox model-based estimates of 5-year overall survival are 97.8% for a 1-month surgery-RT interval, and 97.3% for a 3-month surgery-RT interval; 10-year overall survival estimates are 93.3% and 92.1% for 1-month and 3-month surgery-RT intervals, respectively.

Table 2.

External beam radiotherapy (EBRT) patient predictors of overall mortality with surgery to EBRT interval as a continuous variable with full adjustment. N=169,033 patients who received EBRT with 6,972 deaths. Results of a single multivariable cox model, with the outcome as time form start of RT to death or last contact.

Predictor	n Patients	N deaths	Hazard Ratio (vs base)	Robust SE	LCL	UCL	p-value for Predictor
Surgery to EBRT, months (Hazard Ratio is for 1 month difference)	168,933	6,972	1.09	0.01	1.07	1.11	<0.0001
Age at diagnosis
18–39	2,388	74	1.00	(base)			<0.0001
40–49	27,753	548	0.76	0.10	0.59	0.98
50–59	58,617	1,799	1.17	0.14	0.92	1.48
60–69	80,175	4,551	2.16	0.26	1.71	2.75
Race
White	146,979	5,931	1.00	(base)			<0.0001
Black	13,466	837	1.41	0.05	1.31	1.52
Asian	5,293	108	0.64	0.07	0.53	0.79
Other/Unknown	3,195	96	0.83	0.09	0.68	1.01
Charlson Comorbidity Index
0	147,471	5,178	1.00	(base)			<0.0001
1	17,937	1,263	2.03	0.07	1.91	2.17
2	2,822	363	3.91	0.22	3.50	4.37
3	703	168	7.99	0.68	6.77	9.43
Setting
Large Metro	88,916	3,364	1.00	(base)			<0.0001
Metropolitan	53,471	2,376	1.12	0.04	1.05	1.19
Urban	9,131	413	1.07	0.06	0.96	1.19
Less Urban/Rural	12,832	651	1.20	0.06	1.09	1.32
Unknown	4,583	168	1.02	0.10	0.85	1.23
Transfer of Care
No	136,322	5,570	1.00	(base)			0.2144
Yes	32,611	1,402	1.04	0.03	0.98	1.11
Facility Volume Quartile
1	7,864	432	1.00	(base)			<0.0001
2	21,417	1,062	0.91	0.06	0.81	1.04
3	41,332	1,756	0.76	0.05	0.68	0.86
4	98,320	3,722	0.74	0.04	0.66	0.82
Tumor Size
≤20mm	148,810	5,567	1.00	(base)			<0.0001
21–50mm	18,223	1,303	1.88	0.06	1.75	2.01
>50mm	597	50	1.80	0.27	1.34	2.42
Unknown	1,303	52	0.75	0.10	0.58	0.98
Number of LN positive
0	158,861	6,122	1.00	(base)			<0.0001
1–3	9,728	757	1.75	0.07	1.61	1.89
4–9	173	39	3.34	0.59	2.35	4.73
≥10	171	54	4.70	0.76	3.43	6.46
ER-PR status
ER- and PR-	6,641	734	1.00	(base)			<0.0001
ER+ and/or PR+	160,871	6,151	0.79	0.04	0.71	0.88
Missing	1,421	87	0.57	0.07	0.44	0.73
Grade
1	65,201	2,056	1.00	(base)			<0.0001
2	77,850	3,156	1.19	0.04	1.12	1.26
3	17,320	1,408	1.65	0.07	1.52	1.80
Anaplastic/Undifferentiated	196	24	1.78	0.39	1.16	2.74
Unknown	8,366	328	1.11	0.07	0.99	1.25
Hormone Therapy
Yes	144,854	5,101	1.00	(base)			<0.0001
No	24,079	1,871	1.43	0.05	1.33	1.53

^*year of diagnosis included as Restricted cubic spline

SE = standard errors, LCL = lower confidence limits, UCL = upper confidence limits

The hazard ratios for overall survival in six Cox models, including three sensitivity analyses, with increasing degrees of adjustment, range between 1.19 (95%CI 1.17–1.21, p<0.0001) and 1.08 (95%CI 1.067–1.11, p<0.0001), and are listed in Supplemental Table 7. Plots of overall survival with increasing levels of adjustment, including one sensitivity analysis are illustrated in Figure 3 and Supplemental Figure 1. The hazard ratios, further stratified by surgery to RT delay length, for the three primary Cox models and the sensitivity analysis containing the largest adjustment are elaborated in Table 3. Predictors of times from surgery to EBRT and brachytherapy initiation are shown in Supplemental Tables 8 and 9.

Figure 3. Overall survival plots by increasing levels of adjustment.

Panel A. Overall survival (OS) by surgery-radiotherapy (RT) interval for patients who received EBRT (left) or brachytherapy (right), estimated using Kaplan-Meier methods.

Panel B. Overall survival curves for EBRT (left) and brachytherapy(right) patients, estimated from Cox models without covariates. Hazard ratio estimate for longest vs shortest interval group is 2.36 (95%CI=2.10–2.65, p<0.001) for EBRT and 1.49 (95%CI=1.03–2.13, p=0.035) for brachytherapy. From separate Cox models with the surgery-RT interval as a continuous variable, unadjusted HR for a 1 month difference = 1.19 (95%CI 1.17–1.21, p<0.0001) for EBRT and HR for a 1 week difference = 1.02 (95%CI 1.00–1.04, p=0.0229) for brachytherapy.

Panel C. Overall survival curves for EBRT (left) and brachytherapy(right) patients, estimated from Cox models with adjustment for age, race, and Charlson comorbidity index (CCI). Adjusted hazard ratio estimate for longest vs shortest interval group is 2.43 (95%CI 2.17–2.73, p<0.001) for EBRT and 1.59 (95%CI 1.10–2.31, p=0.014) for brachytherapy. From separate age, race and CCI-adjusted Cox models with the surgery-RT interval as a continuous variable, adjusted HR for a 1 month difference = 1.19 (95%CI 1.17–1.21, p<0.0001, same as univariate) for EBRT and HR for a 1 week difference = 1.03 (95%CI 1.01–1.05, p=0.0084) for brachytherapy.

Panel D. Overall survival curves for EBRT (left) and brachytherapy(right) patients, estimated from Cox models with further adjustment for age, race, Charlson comorbidity index, urban/rural setting, transfer of care, facility volume, ER/PR status, tumor size, lymph node positivity, grade, endocrine therapy, and year of diagnosis (with year of diagnosis as a restricted cubic spline while others are categorical). Adjusted hazard ratio estimate for longest vs shortest interval group is 1.40 (95%CI 1.23–1.59, p<0.001) for EBRT and 1.46 (95% CI 0.99–2.17, p=0.057) for brachytherapy. From separate adjusted Cox models with the surgery-RT interval as a continuous variable, adjusted HR for a 1 month difference = 1.09 (95%CI 1.07–1.11, p<0.0001) for EBRT, and adjusted HR for a 1 week difference = 1.03 (95%CI 1.00–1.05, p=0.0175) for brachytherapy.

Table 3.

Results of four separate Cox models, with increasing degrees of adjustment. For each model, the outcome was time to death from the start of EBRT, with censoring at last contact. Also for each model, on the surgery to EBRT variable is shown, which is a categorical variable with 6 levels. The lowest interval is the reference group of 1–45 days. The hazard ratios are for each interval relative to the 1–45 day interval. The yellow highlights the longest interval (vs the shortest). With adjustment for covariates, the hazard ratio decreases. Robust standard errors were based on within PUF facility cluster adjustment.

	HR estimate	Robust SE	Z for pairwise	Pairwise with base, p-value	HR LCL	HR UCL	overall p-value for interval variable
Unadjusted for other covariates
1–45 days	1.00	(Base)					<0.0001
46–60 days	1.08	0.04	2.23	0.026	1.01	1.15
61–75 days	1.29	0.05	6.54	0.000	1.20	1.40
76–90 days	1.61	0.08	9.45	0.000	1.46	1.78
91–120 days	1.88	0.10	11.43	0.000	1.69	2.10
121+ days	2.36	0.14	14.63	0.000	2.10	2.65
Adjusted for age, race, and CCI
1–45 days	1.00	(Base)					<0.0001
46–60 days	1.08	0.03	2.35	0.019	1.01	1.15
61–75 days	1.30	0.05	6.67	0.000	1.20	1.46
76–90 days	1.62	0.08	9.42	0.000	1.46	1.79
91–120 days	1.77	0.10	10.67	0.000	1.60	1.97
121+ days	2.43	0.14	15.29	0.000	2.17	2.73
Adjusted for age, race, CCI, setting, transfers of care, facility volume, ER/PR status, tumor size, lymph node positivity, grade, endocrine therapy, and year of diagnosis (year of diagnosis as restricted cubic spline, others are categorical)
1–45 days	1.00	(Base)					<0.0001
46–60 days	1.05	0.03	1.38	0.167	0.98	1.12
61–75 days	1.21	0.05	5.02	0.000	1.13	1.31
76–90 days	1.42	0.07	6.95	0.000	1.29	1.57
91–120 days	1.45	0.08	6.60	0.000	1.30	1.63
121+ days	1.40	0.09	5.12	0.000	1.23	1.59
Sensitivity analysis: Addition of income, education, and type of insurance to adjusted model above
1–45 days	1.00	(Base)					<0.0001
46–60 days	1.05	0.03	1.51	0.131	0.99	1.12
61–75 days	1.21	0.05	4.92	0.000	1.12	1.31
76–90 days	1.40	0.07	6.64	0.000	1.27	1.55
91–120 days	1.40	0.08	5.96	0.000	1.26	1.57
121+ days	1.37	0.09	4.79	0.000	1.21	1.56

HR = hazard ratio, SE = standard errors, CCI = Charlson comorbidity index, LCL = Lower confidence level, UCL = upper confidence level

There were only weak correlations (r=0.057 for EBRT, r=−0.055 for brachytherapy) between the time from diagnosis to surgery and time from surgery to radiotherapy and little association between treatment intervals and facility volume for either RT type.

DISCUSSION

Although breast conservation therapy has been standard of care for over thirty-five years,¹⁷ the paradigm has shifted such that we now recognize that the RT contributes not only to reducing the relative risk of local-regional recurrences by approximately 50–60%, but also can result in a modest, but significant, benefit in overall survival for some patients.⁹ Attempts have been made to discern a subgroup of patients for whom RT may be omitted as standard of care,¹⁸ but it remains essential for all but the lowest risk patients (i.e. hormone receptor-positive, node-negative, T1 primaries, having endocrine therapy, over age 70).¹⁹

To that end, a quality measure was developed by the Commission on Cancer⁵ and that has been endorsed by the National Quality Forum (NQF#0219),^3,4 The National Comprehensive Cancer Network (NCCN), American Society of Clinical Oncology (ASCO),² The National Accreditation Program for Breast Centers (NAPBC),¹ and the National Consortium of Breast Centers’ National Quality Measures for Breast Cancer Centers (NQMBC) program.⁶ It has also been endorsed by the American Society of Therapeutic Radiation Oncology (ASTRO).⁷ This measure specifies that female patients, aged 18–69, who have their first diagnosis of AJCC stage I, II, or III breast cancer who undergo BCS, should be administered breast RT starting within 1 year (365 days) of the date of diagnosis. This measure mandates actual administration with initiation of RT within the time frame indicated, in contrast to the current time-dependent quality measures for chemotherapy (NQF #0559) and endocrine therapy (NQF #0220) where either recommendation or administration of those therapies are defined as sufficient within their specified period of time.^20,21 This difference stems from the fact that BCS and radiation are both considered integral to breast conserving therapy, and creates accountability to preoperatively educate patients about the need for radiotherapy before any decision to embark on BCS. For this measure, the specified time frame of one full year allows for completion of surgery, and the completion of any currently utilized chemotherapy regimen, while still allowing a small buffer for radiation oncology consultation, simulation, and prompt initiation of RT.⁸ However this same quality measure also applies to those breast conservation patients not receiving chemotherapy, with the unintended implication that a delay of nine months to initiate postoperative radiation treatment in these patients is acceptable.⁸

Although components of evaluation, in and of themselves, have an inherent time cost that cannot be eliminated, delaying treatment by a small amount,^22–24 timely treatment should still be pursued. This is because timely breast cancer surgery,^15,25 chemotherapy,^26,27 and radiation^10,12,14 all maximize survival. It remains clear that delays in RT affect recurrence, but there is no universally-accepted threshold of delay,^11–14 and delays as short as 6–8 weeks after breast conservation surgery have been shown to potentially to compromise long-term outcomes.^10,11 Thus, the existing radiation measure does not measure potential delays in radiation for patients who do not receive systemic chemotherapy, and who should proceed promptly from surgery to radiation and allows for a significant time lapse between surgery and radiation without identifying these patients at risk. Thus, the goal of this study was to assess the timeline, current patterns of care, and outcomes in the United States for patients undergoing breast conservation therapy who do not receive chemotherapy.

There was a marked difference in the timing of RT for chemotherapy versus no chemotherapy patients undergoing breast conservation. As expected, the overlap in the timeline was minimal (Figure 2); nevertheless, nearly 25% of patients in the no-chemotherapy group still received their RT more than 3 months (90 days, 8.6 weeks) after their surgery, suggesting that more specific guidance on optimal timeliness of radiation for these patients is warranted.

Additional rationale for dichotomizing the radiation quality measure, so that patients without chemotherapy are assessed using a different timeline, is our finding that when proceeding directly from surgery to RT, there was a 9% decline per month in the HR for survival for those receiving external beam radiotherapy, and a 3% decline per week in those patients receiving brachytherapy. Even though the overall hazard for death from breast cancer has diminished significantly in the last decade due to a multitude of factors, it was notable that after maximal adjustment, there was still a 37% increase in the hazard for death in patients with surgery and radiation intervals greater than 120 days (Table 3). Even if surgery were performed 3 full months after diagnosis, in those not needing postoperative chemotherapy, as it is written currently, the quality measure allows more than twice that amount (275 days or ~9 months) from diagnosis until the start of RT, and we therefore believe it should be split into two.

While no time-dependent quality measure defining the maximal time between diagnosis and surgery has yet been adopted, our group has previously proposed a threshold of 90 days^8,15 based upon our analysis of patterns of care and outcomes benefits. Furthermore, the weak correlation between the surgery to radiation interval, and the diagnosis to surgery interval suggest that these two intervals likely require independent measurement and optimization. A major limitation of the current radiation quality measure is that it specifies the time interval as measured from date of diagnosis. This was consistent with the goal of this early breast cancer quality measure was to identify accidental omission or delay of RT. However, due to the long interval between surgery and completion of chemotherapy, the measure inherently also captures surgical and chemotherapy delays in that interval and fails to adequately address significant delays in those not receiving chemotherapy. We believe that the quality measure should be primarily reflective of the care it is designed to optimize, and given these limitations to the current measure as discussed above, our findings support a change to be consistent with the data where the interval delay has been measured from time of surgery to RT, and not the date of diagnosis to radiation administration. This would provide uniform definitions for these delays and avoid confounding from inclusion of delays related to treatments delivered by the other disciplines.

The process of quality measure creation should always be based around a need, where the existing pattern of care diverges enough from the optimal pattern of care to warrant imposing a measure, or rule. In this study, when we looked at compliance for the current standard, over 99% of patients received radiation within a year of diagnosis, compliant with the standard. For those not receiving chemotherapy, 25% of patients waited 3 months or more for RT, suggesting that the measure is failing to identify a detrimental delay for those patients not receiving chemotherapy, while still allowing for an extraordinarily high compliance rate with this measure.

These data are consistent with patient attitudes about delays and RT. A survey study from a tertiary care center in Toronto demonstrated that, in contrast to the average delay of 13 weeks to start radiation at their center, patients responded that they would be willing to wait on average 7 weeks before wanting to leave their city in order to obtain the treatment more promptly.²⁸

It must be acknowledged that this study is limited by the fact that the NCDB is not a random or population sample. While it is subject to quality assurance, includes the majority of all cases in the United States, and is felt to be representative,²⁹ there may also be data that is germane to patient outcomes that are not adequately captured by this or any such large registry dataset, such as significant comorbidities that affect timeliness and RT decisions. Therefore associations cannot confirm causation. The NCDB does contain the data, however, from all Commission on Cancer hospitals, who are bound by the metrics described here, and for that reason, is felt to be generalizable to the population and the most appropriate sample for this analysis.

In conclusion, our findings suggest that while the current radiation quality measure provides an adequate time interval to account for systemic chemotherapy prior to RT, it fails to detect clinically significant delays in those patients not receiving chemotherapy. Our findings demonstrate that up to 25% of patients proceeding directly from surgery to radiation have delays greater than 60 days, and importantly, that these delays result in significant increases in the hazard of mortality in these patients. Thus, we propose that the radiation quality measure be dichotomized into two groups, with a defined shorter time interval for those not receiving chemotherapy. Furthermore, to harmonize the existing data demonstrating delays in radiation with adverse outcomes with the radiation quality measure, we propose that the definition of the new quality measure (for the non-chemotherapy subgroup), be measured from the date of surgery and not diagnosis. These adjustments will allow for more clinically relevant guidance for clinicians, better overall assessment of radiation oncology performance, appropriate identification of centers with unacceptable radiation delays for both chemotherapy and non-chemotherapy patients, and ultimately, they should maximize outcomes for all patients undergoing BCS.