Breast Cancer Res Treat (2015) 149:199–209 DOI 10.1007/s10549-014-3220-3
EPIDEMIOLOGY
Geographic access and age-related variation in chemotherapy use in elderly with metastatic breast cancer Shaowei Wan • Steven Jubelirer
Received: 23 August 2014 / Accepted: 21 November 2014 / Published online: 4 December 2014 Ó Springer Science+Business Media New York 2014
Abstract Significant age-related variation in chemotherapy use has been observed among elderly patients with metastatic breast cancer (MBC), which may be partly attributable to geographic access factors such as local area physician practice culture and local health care system capacity. The purpose of the paper was to examine how age may modify the effect of geographic access on chemotherapy use in elderly patients with MBC. This was a retrospective cohort study based on the surveillance, epidemiology, and end results—Medicare-linked database of 1992–2002. Chemotherapy use was defined as at least one chemotherapy-related claim within 6-month postdiagnosis. Geographic access to cancer care was measured by four variables: patient travel time to the nearest oncologist practice, local area per capita number of oncologists, local area per capita number of hospices, and local area chemotherapy rate. Using multivariate logistic regression model, both aggregate models with interaction terms and subgroup analyses were conducted. Among 4,533 elderly with MBC, 30.16 % used chemotherapy. Chemotherapy use decreased with age. Both the aggregate model with interaction terms and the subgroup analysis showed that
Electronic supplementary material The online version of this article (doi:10.1007/s10549-014-3220-3) contains supplementary material, which is available to authorized users. S. Wan (&) Department of Pharmaceutical and Administrative Sciences, The University of Charleston School of Pharmacy, 2300 MacCorkle Ave. SE, Charleston, WV 25304, USA e-mail: [email protected] S. Jubelirer West Virginia University Charleston Division, David Lee Cancer Center, Charleston Area Medical Center, Charleston, WV 25304, USA
local area chemotherapy rate was positively associated with chemotherapy use (P = .0004 in the whole group; in the subgroup analyses, P \ .0001, P = .0006, P = .0006, P = .18, P = .026, respectively). In addition, subgroup analysis showed that, among patients aged 85? years old, local area oncologist supply was negatively associated with chemotherapy use (P = .028). The impact of geographic access to cancer care is the greatest among the oldest group, for whom the clinical evidence is the least certain. Keywords Metastatic breast cancer SEER-Medicare Geographic variation Distance to care Elderly Disparity
Introduction Literature has documented geographic variation in medical services provided to patients receiving post-acute care and end-of-life care across hospital referral regions, health service areas, and among academic medical centers [1–7]. Dartmouth researchers have further shown that variations in regional health care resource use, particularly in preferencesensitive or supply-sensitive care, were closely correlated with geographic access factors such as local area physician practice culture and local health care system capacity [8–11]. Recent research further examined this association by adding more sophisticated measures such as productivity, health status, health functioning, and the severity of comorbidities, as well as in more refined population, and found that these measures accounted for an additional portion of the variation [12–17]. The recent Institute of Medicine report recommended that geographic variation in health care spending needs to be addressed at the decision-making level and that practitioners should consider specific patient characteristics when making these decisions [18].
123
200
An important variable that may play a role in this variation is patient age. Because of increased heterogeneity among patient characteristics and uncertainties regarding treatment effectiveness in elderly patients, it is important to consider how age potentially modifies the effect of geographic access on health care use; in other words, geographic access is likely to affect older patients more than younger patients. Chemotherapy for metastatic breast cancer (MBC) can be considered preference-sensitive care [11]. It is an important treatment option for MBC patients, and it is currently the only therapeutic option for patients with hormone-receptor-negative (HR-) and human epidermal growth factor receptor-2 (HER-2) negative (namely, triple negative) endocrine resistant breast cancer [19, 20]. Over the study period 1992 and 2002 in this study, limited chemotherapy choices were available. A number of new targeted therapy and chemotherapy drugs became available since 2002, with Transtuzumab first approved for HER2 ? MBC in 1998 (see online resources for treatment recommendations for MBC in elderly in 1992, 2002, and 2014). Therefore, comparing with current treatment guidelines, it was similar of the guidelines then in how to choose appropriate treatment options for MBC, which was based on the assessment of HR status, site of metastasis, extent of disease, and age. The major difference in the treatment recommendations since 2002 is the availability of the targeted therapies for HER-2 ? MBC. Although clinical trials have demonstrated the clinical efficacy of chemotherapy in survival and symptom control for MBC, few patients older than 65 years were enrolled and even fewer with comorbidities (e.g., depression) [21, 22]. This lack of evidence resulting from few elderly being involved in clinical trials adds uncertainties in chemotherapy choices because of the potential toxicity of chemotherapy [23]. Furthermore, breast cancer is primarily a disease of elderly, and older women tend to have a higher rate of metastasis at diagnosis [24]. Therefore, the choice of chemotherapy for MBC is particularly complicated for elderly patients because of the trade-offs between the potential survival gains and its effects on their immediate quality of life in the face of limited life expectancy. Great variation exists in chemotherapy use for elderly patients with metastatic breast cancer. Using US Medicare claims, researchers have shown a variation in chemotherapy use for MBC from 38.5 % in patients aged 65–69 years to 9.7 % in patients aged 80? years [25]; other studies have found similar variations in Sweden and France [26, 27]. Several empirical studies based on US integrated health care settings or community cancer centers examined treatment patterns among older women with breast cancer from stage I to stage IV and found that older age was independently associated with decreased use of
123
Breast Cancer Res Treat (2015) 149:199–209
recommended treatments in addition to increased comorbidities, lower functional status, limited remaining life span, and increased risk of adverse events [28–31]. In this study, we examined whether geographic access to cancer care affects chemotherapy use by older patients with MBC more than that of their younger counterparts. The central hypothesis of this study is that age modifies how geographic access factors affect patient treatment choices; specifically, there is a multiplicative interaction effect between age and geographic access variables such that the effect of geographic access on treatment decisions is greater for older patients than for younger patients, and the effect is the greatest among the oldest patients.
Methods Data source This retrospective cohort study used the Surveillance, Epidemiology, and End Results (SEER)—Medicare-linked database. SEER data, also known as Patient Entitlement and Diagnosis Summary File (PEDSF), records patient demographic and clinical variables. Treatment information was captured from the physician, outpatient, and inpatient claims of Medicare data. The Medicare hospice file was used to identify patients who used hospice care. The Medicare provider of service (POS) File was used to obtain the ZIP codes and the number of Medicare-certified hospices. Study population The study subjects were women aged 66? who were initially diagnosed with stage IV metastatic breast cancer between 1992 and 2002. Using SEER data, patients were selected based on the demographic (e.g., patient age) and clinical characteristics (e.g., the cancer site, cancer stage at diagnosis). Patients whose primary cancer diagnoses were not breast cancer or who were initially diagnosed with stage I, II, or III breast cancer were excluded from the study. The extent or severity of cancer (Stage IV) was measured based on the cancer stage defined by the American Joint Committee on Cancer (AJCC) that is recorded in PEDSF file. Patients were restricted to those aged 66? at the time of diagnosis to ensure at least 1 year of Medicare eligibility prior to diagnosis to measure comorbid conditions. Patients were included if they had continuous Medicare Part A and Part B coverage for at least 12 months before diagnosis and the minimum of 12 months or date of death after diagnosis. The analysis excluded patients with Medicare-managed care coverage because their claims were not available in the Medicare datasets. Although
Breast Cancer Res Treat (2015) 149:199–209
201
SEER collects information on first-course treatments, including surgery, chemotherapy, radiation therapy, and hormone therapy, Medicare claims are referred as gold standard to identify treatment-related data [32, 33]. Our study also used Medicare claims to capture chemotherapy and radiation therapy for the study population. Table 1 illustrates the flow diagram for patient selection.
183 days was selected based on the clinical practice that MBC patients who opt for chemotherapy are typically treated within 3-month post-diagnosis. In practice there will be a small subset of patients who have delayed start for a variety of issues such as wound healing or patient/physician choice. Independent variables
Main outcome of interest Chemotherapy was identified from the Medicare physician, outpatient, and inpatient claims using the algorithm validated in previous studies. A list of codes is available in Table A1. These codes were initially validated by the SEER-Medicare program, with sensitivity higher than 88 % [34]. Subsequent epidemiologic studies validated and applied the codes to measure chemotherapy use [25, 35]. A recent update from the SEER-Medicare program suggests that the Medicare claims continue to be the gold standard to capture the treatment claims such as chemotherapy [32]. Only the first chemotherapy claim was kept for each patient. Immediate chemotherapy use was defined in the analysis by any chemotherapy claims made within 6 months (183 days) of diagnosis. The time window of Table 1 The Inclusion and Exclusion Criteria for Selecting Study Subjects Inclusion and exclusion criteria
No. of patients (%)
No. of patients remaining
Total number of breast cancer patients in 2002 PEDSF file Patients who were first diagnosed with MBC
242,121 (100 %)
242,121
9,087 (3.8 %)
9,087
Patients with unknown diagnosis date
36 (0.4 %)
9,051
Patients with an invalid death date
31 (0.3 %)
9,020
All percentages below out of 9087
Diagnosis of MBC after the date of death or an invalid date of death from SEER data but blank date of death from Medicare claims Patients diagnosed between 1992 and 2002 Patients who are 66 years or older at diagnosis
8,521 (94.4 %)
8,521
6,269 (69.0 %)
6,269
Patients with complete Medicare coverage and no HMO coverage from 12 months before diagnosis to the earlier of 1 years after diagnosis or death
4,612 (50.8 %)
4,612
Patients with a zip code that has no latitude and longitude data or not within their SEER areas
79 (0.9 %)
4,533
Age was calculated based on birth date information from the PEDSF file. It was specified as either continuous or categorical consistent with clinical norms. Specifically, the study used two age subgroups, one above and below the median age of 75 years; three age groups of 66–69, 70–79, and 80? years; or five age groups of 66–69, 70–74, 75–79, 80–84, and 85? years. We conceptualized local area treatment style and local health care system capacity as aspects of access to cancer care. The study used four variables to measure patient access to cancer care: (1) patient travel time to the nearest oncologist practice, (2) local area per capita number of medical oncologists, (3) local area per capita number of hospice programs, and (4) local area chemotherapy rate. Patient travel time to the nearest oncologist practice was measured as the minimum travel time from a MBC patient to a medical oncologist practice available at the ZIP code level in the patient’s year of diagnosis [12, 36]. Medical oncologists who were treating breast, prostate, colorectal, and lung cancer patients and who specialized in hematology/oncology, medical oncology, surgical oncology, and gynecology/oncology were identified through Medicare physician services. Using Microsoft’s Geographic Information System MappointÒ 2009, the longitude and latitude of these ZIP codes were used to calculate ZIP code to ZIP code travel times. Local area per capita number of medical oncologists was measured within a defined area around each MBC patient residence in the year of diagnosis. The numerator of the measure was the number of medical oncologists who were treating breast, prostate, colorectal, and lung cancer patients and practicing within a defined area. The same specialties of medical oncologists were identified as in the calculation of patient travel time. The denominator was the number of stage IV lung, colorectal, breast, and prostate cancer patients within the defined area. The same selection criteria for the study population were used to select four types of metastatic cancer patients from 1992 to 2002. Local areas were created based on the driving area for clinical care (DACC) method [12, 37]. Briefly, the DACC method uses driving times (ZIP code to ZIP code) to create local areas around the original patient zip code where other patients with similar conditions reside. The driving time is then expanded incrementally to add patients from the next
123
202
closest zip codes until a threshold number of patients have been reached. The DACC method has been shown to account for additional treatment variation in area practice style than other local area definitions [37, 38]. We defined an area of 50 patients as the measure of a local health care market [39]. The measure was reported as the number of medical oncologists per 1,000 stage IV cancer patients. Similarly, local area per capita number of hospice programs was measured at the patient residence ZIP code level for the year of diagnosis, with the same denominator as defined in local area per capita number of medical oncologists and in the local market size of 50 patients. The numerator of the measure was the number of hospice programs within the area of a MBC patient residence ZIP code. The measure was reported as the number of hospice programs per 1,000 stage IV cancer patients. Local area chemotherapy rate across cancers was measured at each MBC patient residence ZIP code level across years. The denominator was the same as in local area per capita number of medical oncologists and per capita number of hospice programs. The numerator was the number of patients in the denominator who was identified to have received chemotherapy. The measure was reported as the number of patients who received chemotherapy per 100 stage IV cancer patients.
Breast Cancer Res Treat (2015) 149:199–209
In the aggregate multivariate logistic regression models, age and access variables were specified as categorical or continuous variables with interaction terms between two variables, controlling for other covariates. Both age and the squared term of age were added in the aggregate model to explore the possible nonlinear effects of age. In the subgroup analysis, multivariate logistic regression models were based on two age groups of below/above the median age of 75 years, three age groups with 10-year ranges and five age groups with 5-year ranges, with access variables specified in quartiles, controlling for other covariates. In the aggregate multivariate logistic regression analysis, Chow F-statistics were used to test whether the effect of access on patient chemotherapy choice was stronger among older patients [44]. In the stratified subgroup analysis, the coefficient of each access variable was estimated and statistical significance was assessed for each age category. All data analysis was performed using SAS 9.00 at the significance level .05 (SAS Institute Inc, Cary, NC). This study was approved by the University of Iowa Institutional Review Board. SEER-Medicare officials reviewed and approved the findings for confidentiality assurances.
Results Covariates Control variables included patient demographic characteristics (race and marital status), socioeconomic characteristics (income and education at census level), diagnosis year, residing SEER regions, rural/urban area characteristics, clinical characteristics (comorbidities, hormone receptor status), and radiation therapy. To measure comorbidities, we adopted the Klabunde version of the Charlson comorbidity index [40–42]. The ICD-9-CM diagnosis codes, ICD-9-CM procedure codes, and HCPCS codes to determine major medical conditions for cancer patients are included in Online Resources Table A3-1 and Table A3-2. The source of covariates is available in Online Resources Table A4. A detailed description of covariates in the model is given in Online Resource Table A5.
The average age of the study population was 77.6, ranging from 66.0 to 103.2. There were 30.2 % of 4533 MBC patients who used chemotherapy within a 183-day time window of diagnosis. The chemotherapy rate decreased steadily with age, with 49.3, 40.5, 31.8, 17.6, and 7.7 % in the subgroups aged 66–69, 70–74, 75–79, 80–84, and 85?, respectively. Table 2 shows the clinical and demographic characteristics of the patients by chemotherapy use. Patients who used chemotherapy averaged 74.2-year old while those who did not averaged 79.1-year old. Patients who used chemotherapy were often married and had higher socioeconomic status; those who had no comorbidities or had negative hormone receptor status appeared more likely to use chemotherapy.
Statistical analysis
Bivariate association between access variables and chemotherapy use
To assess the heterogeneity of the association between access and chemotherapy use across age, we began with a statistical test for interaction to determine whether the effect of access to cancer care on chemotherapy use was modified by age in the whole group; we followed with a stratified subgroup analysis [43]. Chemotherapy use was the dependent variable. Age and access variables were key independent variables.
In the whole group analysis, chemotherapy use increased with local area chemotherapy rate (P \ 0.0001). Although it appeared that MBC patients living farther away from the nearest oncologist practice or those living in an area with a higher local area per capita number of oncologists used less chemotherapy, bivariate analysis did not achieve statistical significance (P = 0.855, 0.102 respectively). In addition, there was no obvious trend in chemotherapy use associated
123
Breast Cancer Res Treat (2015) 149:199–209
203
Table 2 Patient characteristic by chemotherapy use
Overall
Percent chemo use
No. of cases
30.2
4,533
P value
Percent chemo use
49.3
No. of cases
SEER area \0.0001*
Age at diagnosis 66–69
Table 2 continued
747
0.9718
Louisiana
34.9
146
Connecticut
31.6
531
31.3
160
70–74
40.5
1,111
Kentucky/Rural Georgiad
75–79
31.8
1,068
Los Angeles
30.9
479
80–84
17.6
856
New Jersey
30.8
454
7.7
751
San Jose
30.4
168
Detroit
30.1
657
1,071
Great California/Hawaiid
30.0
337
30.0
231
85 or older
\0.0001
Area chemo rate 1st quartile
23.7
2nd quartile
29.5
1,063
Atlanta
3rd quartile
31.9
1,075
Seattle
29.4
346
4th quartile
35.8
1,075
Iowa
29.0
507
San Francisco
28.5
253
Utah
27.9
136
New Mexico
25.0
128
1992
26.2
374
1993
29.8
336
1,071
1994
26.5
306
30.7
339
Travel time 1st quartile
30.6
2,141
3rd quartile
30.2
1,070
4th quartile
29.6
1,073
0.8552
2nd quartile
No. of oncologists 1st quartile
Year of diagnosis
0.1016 32.8
0.6677
2nd quartile
30.9
1,070
1995
3rd quartile
29.0
1,071
1996
30.1
336
4th quartile
28.3
1,072
1997
30.5
318
No. of hospices
1998
30.1
289
1st quartile
1999
34.0
312
1,056
2000
31.2
650
31.1
644
30.2
629
2nd quartile
30.7 31.0
1,069
0.7,722
3rd quartile
30.2
1,076
2001
4th quartile
29.1
1,083
2002
Race and ethnicity
0.7810
White
30.1
3,740
Black
29.5
488
Other
31.8
305
Single
25.7
474
Married
39.9
1,391
Divorced/Widowed
25.6
2,497
Unknown
29.8
171
Percentage of low-income residents was above median in 2000 for beneficiary census tract
b
Positive
29.7
1,448
Negative
41.5
1,151
Unknown
23.7
1,934
Percentage of low-income residents above mediana 32.3
1,511
Yes
29.7
1,516
Unknown
28.5
1,506
Comorbiditiesb 0
32.3
1
16.2
734
2
8.4
381
SEER areas were combined based on the proximity due to small size
Big metro
30.1
2,754
Metro
30.7
1,136
Urban
31.6
263
Less urban
28.8
302
Rural
26.3
76
Area chemo rate 100 9 percentage chemotherapy use among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
\0.0001*
Travel time driving time from the patient residency to the nearest oncologist practice. Travel time only has three quartiles because the first quartile and the second quartile are merged into one category due to the first quartile value = 0
0.8796
No. of oncologists 1,000 9 per capita number of oncologists among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
3,418
Rural/urban area characteristicsc
Rural/urban area characteristics were based on area resource file that is available in the PEDSF file
d
0.0658
No
Comorbidities exclude AIDS and all metastatic tumors. The following chronic conditions were measured chronic obstruction pulmonary disease, cerebrovascular disease, congestive heart failure, diabetes, dementia, myocardial infarction, peripheral vascular disease, peptic ulcer disease, paralysis, renal disease, rheumatologic disease
c
\0.0001*
Hormone receptor status
* P \ 0.05 a
\0.0001*
Marital status
P value
No. of hospices 1000 9 per capita number of hospices among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patients
123
204
with the local area per capita number of hospices across metastatic cancers (P = 0.772). Figure 1 shows the bivariate analysis between chemotherapy use and geographic access variables in quartiles by age subgroups. The increasing trend in chemotherapy use associated with local area chemotherapy rate remained in each age category except in the age group of 80–84 years (Fig. 1a). Bivariate association between chemotherapy use and patient travel time to the nearest oncologist in the age subgroup analysis did not achieve statistical significance, although in older age subgroups there was lower chemotherapy use at longer travel times (Fig. 1b). Neither was the association between chemotherapy use and local area per capita number of oncologists significant (Fig. 1c).
Breast Cancer Res Treat (2015) 149:199–209
a
b Summary of the results of the interaction term approach and subgroup analysis Because the low to moderate correlation between the four access variables suggested that they measure distinct aspects of geographic access (Online Resource Table A6), we have reported the results based on the model that included the four access variables. In the aggregate models including access variables and age (Online Resource Table A7–A10), results showed that age was consistently significantly associated with chemotherapy choice in all models. We included additional results of each access indicator variable in separate models as online resources (Tables A11–A15). Both the aggregate model and the subgroup analysis showed that the effect of local area chemotherapy rate was consistently positively associated with chemotherapy use. However, the interactions between this variable and age did not achieve statistical significance in the aggregate models. In Table 3, where the results of the five age subgroup analysis are summarized, the parameter estimate of the continuous variable suggested that the odds ratios of chemotherapy use would increase by 1.05 (95 % CI 1.03–1.07), 1.03 (95 % CI 1.01–1.05), 1.03 (95 % CI 1.02–1.05), and 1.05 (95 % CI 1.01–1.09) in the age subgroup of 66–69, 70–74, 75–79, and 85? years, respectively, if one additional patient was treated with chemotherapy per 100 metastatic cancer patients of 65? years old in the area. We originally hypothesized that the per capita oncologist supply was positively associated with chemotherapy use. However, the direction of this access variable has been consistently negative on chemotherapy use and the magnitude was not statistically significant in the aggregate models. Neither were the interaction terms significant. Noticeably, in the five age subgroup analyses, local area oncologist supply was significantly negatively associated
123
c
1st Quartile: 9 -31 2nd Quartile: 32 -38 3rd Quartile: 39 -44 4th Quartile: 45 -70
1st -2nd Quartile: 0 -0.17 hr 3rd Quartile: 0.18 -0.32 hr 4th Quartile: 0.33 -3.72 hr
1st Quartile: 13 -62 2nd Quartile: 63 -99 3rd Quartile: 100-157 4th Quartile: 158-713
Fig. 1 Chemotherapy use by quartiles of local area chemotherapy rate, patient travel time to the nearest oncologist practice, and local area per capita number of oncologists across age subgroups. a. Chemotherapy use by quartiles of local area chemotherapy rate among 100 metastatic breast, colorectal, lung, and prostate cancer patients. b. Chemotherapy use by quartiles of patient travel time to the nearest oncologist practice at the patient ZIP code level. c. Chemotherapy use by quartiles of local area per capita number of oncologists among 1,000 metastatic breast, colorectal, lung, and prostate cancer patients
0.61 (0.28–1.30)
Reference
0.72 (0.43–1.21)
Second quartile
0.5127
First quartile
OR (95 % CI)
P valueà
No. of oncologists
0.1996
OR (95 % CI)
1.06 (0.66–1.69)
Fourth quartile
P value
0.81 (0.53–1.24)
Reference
Third quartile
Second quartile
First quartile
OR (95 % CI)
P valueà
0.5298
1.05 (1.03–1.07)
Travel time
\0.0001*
OR (95 % CI)
1.81 (1.08–3.02) 3.14 (1.80–5.48)
Third quartile Fourth quartile
P value
Reference
1.40 (0.86–2.26)
Second quartile
0.0006*
First quartile
OR (95 % CI)
P valueà
0.84 (0.55–1.27)
Reference
0.3563
0.91 (0.63–1.32)
1.09 (0.78–1.51)
Reference
0.6911
1.65 (1.09–2.50) 1.69 (1.10–2.59)
1.35 (0.90–2.03)
Reference
0.0657
Categorical model
Categorical model
Continuous model
70–74
66–69
Area chemo rate
Variables
0.77 (0.40–1.50)
0.4456
1.03 (1.01–1.05)
0.0006*
Continuous model
0.81 (0.52–1.25)
Reference
0.4702
0.82 (0.53–1.27)
0.98 (0.68–1.40)
Reference
0.6694
1.94 (1.22–3.08) 2.21 (1.36–3.62)
1.31 (0.83–2.08)
Reference
0.0045*
Categorical model
75–79
Table 3 Adjusted estimate of the effect of geographic access on chemotherapy use in age subgroups
0.479 (0.23–1.01)
0.0516
1.03 (1.02–1.05)
0.0006*
Continuous model
1.32 (0.72–2.41)
Reference
0.2974
0.67 (0.37–1.23)
0.76 (0.47–1.22)
Reference
0.3093
1.22 (0.64–2.33) 1.58 (0.84–2.98)
1.80 (0.99–3.29)
Reference
0.2037
Categorical model
80–84
0.39 (0.12–1.27)
0.1181
1.02 (0.99–1.04)
0.1839
Continuous model
0.57 (0.21–1.57)
Reference
0.0463*
0.67 (0.25–1.82)
0.21 (0.08–0.56)
Reference
0.0076*
2.718 (0.90–8.22) 3.87 (1.30–11.53)
1.64 (0.56–4.76)
Reference
0.0857
Categorical model
85?
0.89 (0.14–5.61)
0.9022
1.05 (1.01–1.09)
0.0258*
Continuous model
Breast Cancer Res Treat (2015) 149:199–209 205
123
123
1.00 (0.997–1.002)
1.26 (0.71–2.23)
Fourth quartile P value
OR (95 % CI)
1.00 (0.99–1.01)
0.9256
1.00 (0.996–1.001)
0.3606
0.74 (0.44–1.23)
0.84 (0.53–1.33)
0.90 (0.59–1.38)
Reference
0.7007
0.94 (0.55–1.60)
0.73 (0.46–1.17)
Categorical model
1.00 (0.99–1.00)
0.4666
1.00 (0.997–1.002)
0.6689
Continuous model
P value of a Chow F-statistics testing whether all of the interaction terms are simultaneously equal to zero
1.37 (0.71–2.65)
0.90 (0.49–1.66)
0.77 (0.43–1.38)
Reference
0.3025
0.69 (0.31–1.56)
1.02 (0.54–1.96)
Categorical model
80–84
0.39 (0.12–1.27)
0.1435
1.00 (0.994–1.002)
0.3348
Continuous model
4.68 (1.33–16.55)
3.36 (1.04–10.90)
5.85 (1.98–17.23)
Reference
0.0149*
0.17 (0.04–0.72)
0.24 (0.07–0.81)
Categorical model
85?
1.01 (0.90–1.02)
0.3643
0.992 (0.985–0.999)
0.0277*
Continuous model
No. of hospices 1,000 9 per capita number of hospices among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patients. All four access variables were included in the model. Other variables in the model include patient race, marital status, HR status, radiation therapy use, census tract median household income and education level, residency urban/rural characteristics, residency SEER area, and diagnosis year
No. of oncologists 1,000 9 per capita number of oncologists among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
Travel time driving time from the patient residency to the nearest oncologist practice. Travel time only has three quartiles because the first quartile and the second quartile are merged into one category due to the first quartile value = 0
Area chemo rate 100 9 percentage chemotherapy use among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
* Significance level at 0.05 level
à
1.10 (0.69–1.75)
1.34 (0.88–2.06)
1.10 (0.73–1.65)
Reference
0.5354
0.63 (0.37–1.06)
0.73 (0.47–1.14)
Continuous model
75–79
P value of Wald v2 test testing whether the coefficient is significantly different from zero
0.61 (0.28–1.30)
0.9239 (0.56–1.52)
Third quartile
0.1424
Reference
1.10 (0.67–1.79)
Second quartile
0.6677
First quartile
OR (95 % CI)
P valueà
No. of hospices
0.7454
OR (95 % CI)
1.01 (0.53–1.89)
Fourth quartile
P value
0.83 (0.48–1.43)
Categorical model
Categorical model
Continuous model
70–74
66–69
Third quartile
Variables
Table 3 continued
206 Breast Cancer Res Treat (2015) 149:199–209
Breast Cancer Res Treat (2015) 149:199–209
with chemotherapy use in the subgroup of 85? years old. This effect was not observed among younger age groups.
Discussion It has been widely shown in the literature that older patients with breast cancer were less likely to be treated with chemotherapy [28–31]. Researchers suggested that these differences may be attributed to the age-related bias that older patients would not benefit from chemotherapy to the same degree as their younger counterparts [45]. Our study found that geographic access can explain a portion of the agerelated difference in chemotherapy use in elderly patients with MBC. The insufficient evidence regarding the benefits versus the harm of chemotherapy for elderly patients, coupled with the biological and socioeconomic changes of aging, are major sources of uncertainties for older patients. Older patients suffer from higher rates of comorbid conditions (not shown). The presence of comorbidities appeared strongly associated with not receiving chemotherapy. However, age should not be considered as a surrogate for comorbidities. Controlling for comorbidities, age was independently associated with lower chemotherapy use in our analysis, consistent with previous studies. In addition to comorbidities, considerable age-related heterogeneity in functional ability, tolerance to chemotherapy, and patient preferences exists in elderly women [30, 46–49]. These uncertainties may cause non-clinical factors such as access to matter more to older patients. The direction of the cumulative odds ratios across age groups suggested that local area chemotherapy rate affected chemotherapy choice in a consistent way, with higher chemotherapy rate percentiles predicting higher odds of receiving chemotherapy. These results highlight the importance of local area practice culture and provide insight into how local culture may influence treatment choices of oncologists in the absence of sufficient clinical evidence [8, 50]. The lower chemotherapy rate in older patients may reflect a lack of patient-specific data on efficacy and patient preferences in older patients. The best clinical practice may be preference-based discussion between doctors and patients. It should be noted that although the odds ratio of chemotherapy rate by age are statistically significant, the magnitude is small. Therefore, the interpretation of clinical impact should not be overstated. For patients aged 85? years old, local area per capita number of oncologists also affected chemotherapy choice. This suggested that the effect of access variables may be additive instead of being multiplicative with age suggested by the interaction terms. Contrary to our hypothesis that
207
higher per capita supply of services is associated with higher use [51], oncologist supply had a consistently negative effect on receiving chemotherapy, although this effect did not reach statistical significance in younger age groups. Oncologists may be able to spend more time with each patient in areas with higher per capita number of oncologists and evaluate alternatives, such as hormone therapy and hospice care. Especially for the oldest patients who were 85? years old, treatment benefit and harm of chemotherapy need to be considered in the context of patient life expectancy, the nature and the symptoms of the disease, and patient quality of life [52]. Another possible explanation is that the higher number of oncologists was likely more concentrated in academic practice where they may be more focused on patient preferences. The other access variable, local area per capita number of hospices, was shown to not be associated with chemotherapy use, which may be attributed to a variety of treatment options available to patients with MBC other than chemotherapy and hospice care. Our study has some limitations. Complete data for hormone therapy were not available, although we controlled for hormone receptor status in the regression analysis. Among the HR ? subgroup, we observed a similar pattern of decreasing chemotherapy use with age and local area chemotherapy rate was the only significant factor. There were other unmeasured confounders as discussed earlier, which may not be separated from the effect of age and likely contributed to the insignificant interaction terms in the whole group analysis. This study has several strengths, and opened venues for future studies. Due to heterogeneous patient characteristics, estimating all patients aggregately may mask results. By estimating more homogeneous subgroups, we found that the more uncertain the evidence with age, the more geographic access influenced chemotherapy use. The impact of geographic access was the greatest among the oldest patients for whom the clinical evidence was the least certain. Our results suggested that strategies improving access to cancer care may need to tailor to individual organizations, oncologists and patients in local communities where different health care systems and practices often serve the mix of population rooted in similar environments and characteristics. In addition, our results do not assess the extent to which treatment rates are appropriate among the different age groups. Other factors, such as socioeconomic status, may also exert compounded effects with older age, which need to be examined in future studies. Acknowledgments The authors acknowledge the constructive comments of Dr. John Brooks, Dr. Elizabeth Chrischilles, Dr. Linnea Polgreen, Dr. Alexandra Thomas, and Dr. Bernard Sorofman at the University of Iowa College of Pharmacy. The authors would like to thank Ms. Chris Welch for her support during the manuscript
123
208
Breast Cancer Res Treat (2015) 149:199–209
submission. This study used the linked 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, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database. The collection of the California cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Sect. 103885; the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement #U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the author(s) and endorsement by the State of California, Department of Public Health the National Cancer Institute, and the Centers for Disease Control and Prevention or their Contractors and Subcontractors is not intended nor should be inferred. Disclosures
None.
Ethical standards The study complies with the current laws in the U.S.A where the study was performed. Funding This work was supported by the University of Iowa College of Pharmacy Dissertation Fellowship.
References 1. Wennberg JE, Fisher ES, Stukel TA (2004) Use of hospitals, physician visits, and hospice care during last six months of life among cohorts loyal to highly respected hospitals in the United States. BMJ 328:606 2. Wennberg JE, Fisher ES, Goodman D, Skinner JS (2008) Tracking the care of patients with severe chronic illness: the Dartmouth Atlas of Health Care 2008. The Dartmouth Institute for Health Policy and Clinical Practice, Hanover 3. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ (2004) Variations in the longitudinal efficiency of academic medical centers. Health Aff (Millwood) Suppl Variation:VAR19. doi:10.1377/ hlthaff.var.19 4. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL (2003) The implications of regional variations in medicare spending. Part 1: the content, quality, and accessibility of care. Ann Intern Med 138(4):273 5. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL (2003) The implications of regional variations in medicare spending. Part 2: health outcomes and satisfaction with care. Ann Intern Med 138(4):288 6. Weinstein JN, Bronner KK, Morgan TS, Wennberg JE (2004) Trends and geographic variations in major surgery for degenerative diseases of the hip, knee, and spine. Health Aff (Millwood) Suppl Variation:VAR81. doi:10.1377/hlthaff.var.81 7. MedPac (2011) Regional variation in medicare service use. MedPAC, Washington, DC 8. Fisher ES, Wennberg JE (2003) Health care quality, geographic variations, and the challenge of supply-sensitive care. Perspect Biol Med 46(1):69
123
9. Wennberg JE, Fisher ES, Skinner JS (2002) Geography And The Debate Over Medicare Reform. Health Aff hlthaff.w2.96. doi:10. 1377/hlthaff.w2.96 10. Dartmouth Atlas of Health Care (2013) Supply-sensitive care. http://www.dartmouthatlas.org/keyissues/issue.aspx?con=2937. Accessed 1 Sep 2013 11. Dartmouth Atlas of Health Care (2013) Preference-sensitive care. http://www.dartmouthatlas.org/keyissues/issue.aspx?con=2938. Accessed 1 Sep 2013 12. Brooks JM, Cook EA, Chapman CG, Kulchaitanaroaj P, Chrischilles EA, Welch S, Robinson J (2014) Geographic variation in statin use for complex acute myocardial infarction patients: evidence of effective care? Med Care 52(Suppl 3):S37. doi:10.1097/ MLR.0b013e3182a7fc3d 13. Hadley J, Waidmann T, Zuckerman S, Berenson RA (2011) Medical spending and the health of the elderly. Health Serv Res 46(5):1333. doi:10.1111/j.1475-6773.2011.01276.x;10.1111/j. 1475-6773.2011.01276.x 14. Kaestner R, Silber JH (2010) Evidence on the efficacy of inpatient spending on medicare patients. Milbank Q 88(4):560. doi:10.1111/j.1468-0009.2010.00612.x;10.1111/j.1468-0009. 2010.00612.x 15. Reschovsky JD, Hadley J, Saiontz-Martinez CB, Boukus ER (2011) Following the money: factors associated with the cost of treating high-cost Medicare beneficiaries. Health Serv Res 46(4):997. doi:10.1111/j.1475-6773.2011.01242.x;10.1111/j. 1475-6773.2011.01242.x 16. Silber JH, Kaestner R, Even-Shoshan O, Wang Y, Bressler LJ (2010) Aggressive treatment style and surgical outcomes. Health Serv Res 45(6 Pt 2):1872. doi:10.1111/j.1475-6773.2010.01180. x;10.1111/j.1475-6773.2010.01180.x 17. Zuckerman S, Waidmann T, Berenson R, Hadley J (2010) Clarifying sources of geographic differences in medicare spending. New Engl J Med 363(1):54. doi:10.1056/NEJMsa0909253 18. IOM (Institue of Medicine) (2013) Variation in health care spending: targeting decision making, not geography. The National Academies Press, Washington, DC 19. NCCN Clinical Practice Guidelines in Oncology: Breast Cancer V3 (2014) http://www.nccn.org. Accessed 1 May 2014 20. M.D. Anderson Cancer Center Cancer Treatment Algorithms: Breast Cancer (Invasive) V12 (2013) http://www.mdanderson. org. Accessed 1 Mar 2014 21. Hutchins LF, Unger JM, Crowley JJ, Coltman CA, Albain KS (1999) Underrepresentation of patients 65 years of age or older in cancer-treatment trials. New Engl J Med 341(27):2061 22. Scher KS, Hurria A (2012) Under-representation of older adults in cancer registration trials: known problem, little progress. J Clin Oncol 30(17):2036–2038. doi:10.1200/jco.2012.41.6727 23. Green MC, Hortobagyi GN (2008) Chemotherapy for breast cancer. In: Hunt KK, Robb GL, Strom EA, Ueno NT (eds) Breast Cancer, 2nd edn. Springer, New York. doi:10.1007/978-0-387-34952-7_12 24. Howlander N, Noone NA, Krapcho M (2010) Cancer Statistics Review, 1975–2010 25. Du XL, Goodwin JS (2001) Patterns of use of chemotherapy for breast cancer in older women: findings from medicare claims data. J Clin Oncol 19(5):1455 26. Eaker S, Dickman PW, Bergkvist L, Holmberg L (2006) Differences in management of older women influence breast cancer survival: results from a population-based database in Sweden. PLoS Med 3(3):e25 27. Freyer G, Braud AC, Chaibi P, Spielmann M, Martin JP, Vilela G, Guerin D, Zelek L (2006) Dealing with metastatic breast cancer in elderly women: results from a French study on a large cohort carried out by the ‘Observatory on Elderly Patients’. Ann Oncol 17(2):211
Breast Cancer Res Treat (2015) 149:199–209 28. Enger SM, Thwin SS, Buist DSM, Field T, Frost F, Geiger AM, Lash TL, Prout M, Yood MU, Wei F (2006) Breast cancer treatment of older women in integrated health care settings. J Clin Oncol 24(27):4377 29. Giordano SH, Hortobagyi GN, Kau SC, Theriault RL, Bondy ML (2005) Breast cancer treatment guidelines in older women. J Clin Oncol 23(4):783. doi:10.1200/jco.2005.04.175 30. Hurria A, Togawa K, Mohile SG, Owusu C, Klepin HD, Gross CP, Lichtman SM, Gajra A, Bhatia S, Katheria V, Klapper S, Hansen K, Ramani R, Lachs M, Wong FL, Tew WP (2011) Predicting chemotherapy toxicity in older adults with cancer: a prospective multicenter study. J Clin Oncol 29(25):3457–3465. doi:10.1200/jco.2011.34.7625 31. Muss HB, Woolf S, Berry D, Cirrincione C, Weiss RB, Budman D, Wood WC, Henderson IC, Hudis C, Winer E, Cohen H, Wheeler J, Norton L (2005) Adjuvant chemotherapy in older and younger women with lymph node-positive breast cancer. JAMA 293(9):1073. doi:10.1001/jama.293.9.1073 32. Noone AM, Lund JL, Mariotto A, Cronin K, McNeel T, Deapen D, Warren JL (2014) Comparison of SEER treatment data with medicare claims. Med Care. doi:10.1097/mlr.0000000000000073 33. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF (2002) Overview of the SEER-medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 40(8 Suppl):IV. doi:10.1097/01.mlr.0000020942.47004.03 34. Warren JL, Harlan LC, Fahey A, Virnig BA, Freeman JL, Klabunde CN, Cooper GS, Knopf KB (2002) Utility of the SEER-Medicare data to identify chemotherapy use. Med Care 40(8 Suppl):IV-55-61. doi:10.1097/01.mlr.0000020944.17670.d7 35. Chen-Hardee S, Chrischilles EA, Voelker MD, Brooks JM, Scott S, Link BK, Delgado D (2006) Population-based assessment of hospitalizations for neutropenia from chemotherapy in older adults with non-Hodgkin’s lymphoma (United States). Cancer Causes Control 17(5):647–654. doi:10.1007/s10552-005-0502-4 36. Brooks JM, Chrischilles EA, Scott SD, Chen-Hardee SS (2003) Was breast conserving surgery underutilized for early stage breast cancer? Instrumental variables evidence for Stage II patients from Iowa. Health Serv Res 38(6p1):1385. doi:10.1111/j.1475-6773. 2003.00184.x 37. Fang G, Brooks JM, Chrischilles EA (2010) A new method to isolate local-area practice styles in prescription use as the basis for instrumental variables in comparative effectiveness research. Med Care 48(8):710. doi:10.1097/MLR.0b013e3181e41bb2;10. 1097/MLR.0b013e3181e41bb2 38. Fang G, Brooks JM, Chrischilles EA (2012) Comparison of instrumental variable analysis using a new instrument with risk adjustment methods to reduce confounding by indication. Am J
209
39.
40.
41.
42. 43.
44. 45.
46.
47.
48.
49.
50.
51.
52.
Epidemiol 175(11):1142. doi:10.1093/aje/kwr448;10.1093/aje/ kwr448 Lu-Yao GL, Albertsen PC, Moore DF, Shih W, Lin Y, DiPaola RS, Yao SL (2008) Survival following primary androgen deprivation therapy among men with localized prostate cancer. JAMA 300(2):173. doi:10.1001/jama.300.2.173 Charlson ME, Pompei P, Ales KL, MacKenzie CR (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40(5):373 Klabunde CN, Potosky AL, Legler JM, Warren JL (2000) Development of a comorbidity index using physician claims data. J Clin Epidemiol 53(12):1258 Klabunde CN, Warren JL, Legler JM (2002) Assessing comorbidity using claims data: an overview. Med Care 40(8):26–35 Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM (2007) Statistics in medicine: reporting of subgroup analyses in clinical trials. New Engl J Med 357(21):2189. doi:10.1056/NEJMsr077003 Chow GC (1960) Tests of equality between sets of coefficients in two linear regressions. Econometrica 28(3):591 Crivellari D, Aapro M, Leonard R, von Minckwitz G, Brain E, Goldhirsch A, Veronesi A, Muss H (2007) Breast cancer in the elderly. J Clin Oncol 25(14):1882. doi:10.1200/jco.2006.10.2079 Silliman RA, Balducci L, Goodwin JS, Holmes FF, Leventhal EA (1993) Breast cancer care in old age: what we know, don’t know, and do. J Natl Cancer Inst 85(3):190 Koedoot CG, de Haan RJ, Stiggelbout AM, Stalmeier PFM, de Graeff A, Bakker PJM, de Haes JCJM (2003) Palliative chemotherapy or best supportive care? A prospective study Explaining Patients’ treatment preference and choice. Br J Cancer 89:2219 McQuellon RP, Muss HB, Hoffman SL, Russell G, Craven B, Yellen SB (1995) Patient preferences for treatment of metastatic breast cancer: a study of women with early-stage breast cancer. J Clin Oncol 13(4):858 Bouchardy C, Rapiti E, Blagojevic S, Vlastos A, Vlastos G (2007) Older female cancer patients: importance, causes, and consequences of undertreatment. J Clin Oncol 25(14):1858. doi:10.1200/jco.2006.10.4208 Sirovich B, Gallagher PM, Wennberg DE, Fisher ES (2008) Discretionary decision making by primary care physicians and the cost of U.S. Health care. Health Aff (Millwood) 27(3):813. doi:10.1377/hlthaff.27.3.813;10.1377/hlthaff.27.3.813 Buntin MB, Garten AD, Paddock S, Saliba D, Totten M, Escarce JJ (2005) How much is postacute care use affected by its availability? Health Serv Res 40(2):413. doi:10.1111/j.1475-6773. 2005.00365.x Ganz PA (ed) (2007) Cancer survivorship: today and tomorrow. Springer, New York
123
EPIDEMIOLOGY
Geographic access and age-related variation in chemotherapy use in elderly with metastatic breast cancer Shaowei Wan • Steven Jubelirer
Received: 23 August 2014 / Accepted: 21 November 2014 / Published online: 4 December 2014 Ó Springer Science+Business Media New York 2014
Abstract Significant age-related variation in chemotherapy use has been observed among elderly patients with metastatic breast cancer (MBC), which may be partly attributable to geographic access factors such as local area physician practice culture and local health care system capacity. The purpose of the paper was to examine how age may modify the effect of geographic access on chemotherapy use in elderly patients with MBC. This was a retrospective cohort study based on the surveillance, epidemiology, and end results—Medicare-linked database of 1992–2002. Chemotherapy use was defined as at least one chemotherapy-related claim within 6-month postdiagnosis. Geographic access to cancer care was measured by four variables: patient travel time to the nearest oncologist practice, local area per capita number of oncologists, local area per capita number of hospices, and local area chemotherapy rate. Using multivariate logistic regression model, both aggregate models with interaction terms and subgroup analyses were conducted. Among 4,533 elderly with MBC, 30.16 % used chemotherapy. Chemotherapy use decreased with age. Both the aggregate model with interaction terms and the subgroup analysis showed that
Electronic supplementary material The online version of this article (doi:10.1007/s10549-014-3220-3) contains supplementary material, which is available to authorized users. S. Wan (&) Department of Pharmaceutical and Administrative Sciences, The University of Charleston School of Pharmacy, 2300 MacCorkle Ave. SE, Charleston, WV 25304, USA e-mail: [email protected] S. Jubelirer West Virginia University Charleston Division, David Lee Cancer Center, Charleston Area Medical Center, Charleston, WV 25304, USA
local area chemotherapy rate was positively associated with chemotherapy use (P = .0004 in the whole group; in the subgroup analyses, P \ .0001, P = .0006, P = .0006, P = .18, P = .026, respectively). In addition, subgroup analysis showed that, among patients aged 85? years old, local area oncologist supply was negatively associated with chemotherapy use (P = .028). The impact of geographic access to cancer care is the greatest among the oldest group, for whom the clinical evidence is the least certain. Keywords Metastatic breast cancer SEER-Medicare Geographic variation Distance to care Elderly Disparity
Introduction Literature has documented geographic variation in medical services provided to patients receiving post-acute care and end-of-life care across hospital referral regions, health service areas, and among academic medical centers [1–7]. Dartmouth researchers have further shown that variations in regional health care resource use, particularly in preferencesensitive or supply-sensitive care, were closely correlated with geographic access factors such as local area physician practice culture and local health care system capacity [8–11]. Recent research further examined this association by adding more sophisticated measures such as productivity, health status, health functioning, and the severity of comorbidities, as well as in more refined population, and found that these measures accounted for an additional portion of the variation [12–17]. The recent Institute of Medicine report recommended that geographic variation in health care spending needs to be addressed at the decision-making level and that practitioners should consider specific patient characteristics when making these decisions [18].
123
200
An important variable that may play a role in this variation is patient age. Because of increased heterogeneity among patient characteristics and uncertainties regarding treatment effectiveness in elderly patients, it is important to consider how age potentially modifies the effect of geographic access on health care use; in other words, geographic access is likely to affect older patients more than younger patients. Chemotherapy for metastatic breast cancer (MBC) can be considered preference-sensitive care [11]. It is an important treatment option for MBC patients, and it is currently the only therapeutic option for patients with hormone-receptor-negative (HR-) and human epidermal growth factor receptor-2 (HER-2) negative (namely, triple negative) endocrine resistant breast cancer [19, 20]. Over the study period 1992 and 2002 in this study, limited chemotherapy choices were available. A number of new targeted therapy and chemotherapy drugs became available since 2002, with Transtuzumab first approved for HER2 ? MBC in 1998 (see online resources for treatment recommendations for MBC in elderly in 1992, 2002, and 2014). Therefore, comparing with current treatment guidelines, it was similar of the guidelines then in how to choose appropriate treatment options for MBC, which was based on the assessment of HR status, site of metastasis, extent of disease, and age. The major difference in the treatment recommendations since 2002 is the availability of the targeted therapies for HER-2 ? MBC. Although clinical trials have demonstrated the clinical efficacy of chemotherapy in survival and symptom control for MBC, few patients older than 65 years were enrolled and even fewer with comorbidities (e.g., depression) [21, 22]. This lack of evidence resulting from few elderly being involved in clinical trials adds uncertainties in chemotherapy choices because of the potential toxicity of chemotherapy [23]. Furthermore, breast cancer is primarily a disease of elderly, and older women tend to have a higher rate of metastasis at diagnosis [24]. Therefore, the choice of chemotherapy for MBC is particularly complicated for elderly patients because of the trade-offs between the potential survival gains and its effects on their immediate quality of life in the face of limited life expectancy. Great variation exists in chemotherapy use for elderly patients with metastatic breast cancer. Using US Medicare claims, researchers have shown a variation in chemotherapy use for MBC from 38.5 % in patients aged 65–69 years to 9.7 % in patients aged 80? years [25]; other studies have found similar variations in Sweden and France [26, 27]. Several empirical studies based on US integrated health care settings or community cancer centers examined treatment patterns among older women with breast cancer from stage I to stage IV and found that older age was independently associated with decreased use of
123
Breast Cancer Res Treat (2015) 149:199–209
recommended treatments in addition to increased comorbidities, lower functional status, limited remaining life span, and increased risk of adverse events [28–31]. In this study, we examined whether geographic access to cancer care affects chemotherapy use by older patients with MBC more than that of their younger counterparts. The central hypothesis of this study is that age modifies how geographic access factors affect patient treatment choices; specifically, there is a multiplicative interaction effect between age and geographic access variables such that the effect of geographic access on treatment decisions is greater for older patients than for younger patients, and the effect is the greatest among the oldest patients.
Methods Data source This retrospective cohort study used the Surveillance, Epidemiology, and End Results (SEER)—Medicare-linked database. SEER data, also known as Patient Entitlement and Diagnosis Summary File (PEDSF), records patient demographic and clinical variables. Treatment information was captured from the physician, outpatient, and inpatient claims of Medicare data. The Medicare hospice file was used to identify patients who used hospice care. The Medicare provider of service (POS) File was used to obtain the ZIP codes and the number of Medicare-certified hospices. Study population The study subjects were women aged 66? who were initially diagnosed with stage IV metastatic breast cancer between 1992 and 2002. Using SEER data, patients were selected based on the demographic (e.g., patient age) and clinical characteristics (e.g., the cancer site, cancer stage at diagnosis). Patients whose primary cancer diagnoses were not breast cancer or who were initially diagnosed with stage I, II, or III breast cancer were excluded from the study. The extent or severity of cancer (Stage IV) was measured based on the cancer stage defined by the American Joint Committee on Cancer (AJCC) that is recorded in PEDSF file. Patients were restricted to those aged 66? at the time of diagnosis to ensure at least 1 year of Medicare eligibility prior to diagnosis to measure comorbid conditions. Patients were included if they had continuous Medicare Part A and Part B coverage for at least 12 months before diagnosis and the minimum of 12 months or date of death after diagnosis. The analysis excluded patients with Medicare-managed care coverage because their claims were not available in the Medicare datasets. Although
Breast Cancer Res Treat (2015) 149:199–209
201
SEER collects information on first-course treatments, including surgery, chemotherapy, radiation therapy, and hormone therapy, Medicare claims are referred as gold standard to identify treatment-related data [32, 33]. Our study also used Medicare claims to capture chemotherapy and radiation therapy for the study population. Table 1 illustrates the flow diagram for patient selection.
183 days was selected based on the clinical practice that MBC patients who opt for chemotherapy are typically treated within 3-month post-diagnosis. In practice there will be a small subset of patients who have delayed start for a variety of issues such as wound healing or patient/physician choice. Independent variables
Main outcome of interest Chemotherapy was identified from the Medicare physician, outpatient, and inpatient claims using the algorithm validated in previous studies. A list of codes is available in Table A1. These codes were initially validated by the SEER-Medicare program, with sensitivity higher than 88 % [34]. Subsequent epidemiologic studies validated and applied the codes to measure chemotherapy use [25, 35]. A recent update from the SEER-Medicare program suggests that the Medicare claims continue to be the gold standard to capture the treatment claims such as chemotherapy [32]. Only the first chemotherapy claim was kept for each patient. Immediate chemotherapy use was defined in the analysis by any chemotherapy claims made within 6 months (183 days) of diagnosis. The time window of Table 1 The Inclusion and Exclusion Criteria for Selecting Study Subjects Inclusion and exclusion criteria
No. of patients (%)
No. of patients remaining
Total number of breast cancer patients in 2002 PEDSF file Patients who were first diagnosed with MBC
242,121 (100 %)
242,121
9,087 (3.8 %)
9,087
Patients with unknown diagnosis date
36 (0.4 %)
9,051
Patients with an invalid death date
31 (0.3 %)
9,020
All percentages below out of 9087
Diagnosis of MBC after the date of death or an invalid date of death from SEER data but blank date of death from Medicare claims Patients diagnosed between 1992 and 2002 Patients who are 66 years or older at diagnosis
8,521 (94.4 %)
8,521
6,269 (69.0 %)
6,269
Patients with complete Medicare coverage and no HMO coverage from 12 months before diagnosis to the earlier of 1 years after diagnosis or death
4,612 (50.8 %)
4,612
Patients with a zip code that has no latitude and longitude data or not within their SEER areas
79 (0.9 %)
4,533
Age was calculated based on birth date information from the PEDSF file. It was specified as either continuous or categorical consistent with clinical norms. Specifically, the study used two age subgroups, one above and below the median age of 75 years; three age groups of 66–69, 70–79, and 80? years; or five age groups of 66–69, 70–74, 75–79, 80–84, and 85? years. We conceptualized local area treatment style and local health care system capacity as aspects of access to cancer care. The study used four variables to measure patient access to cancer care: (1) patient travel time to the nearest oncologist practice, (2) local area per capita number of medical oncologists, (3) local area per capita number of hospice programs, and (4) local area chemotherapy rate. Patient travel time to the nearest oncologist practice was measured as the minimum travel time from a MBC patient to a medical oncologist practice available at the ZIP code level in the patient’s year of diagnosis [12, 36]. Medical oncologists who were treating breast, prostate, colorectal, and lung cancer patients and who specialized in hematology/oncology, medical oncology, surgical oncology, and gynecology/oncology were identified through Medicare physician services. Using Microsoft’s Geographic Information System MappointÒ 2009, the longitude and latitude of these ZIP codes were used to calculate ZIP code to ZIP code travel times. Local area per capita number of medical oncologists was measured within a defined area around each MBC patient residence in the year of diagnosis. The numerator of the measure was the number of medical oncologists who were treating breast, prostate, colorectal, and lung cancer patients and practicing within a defined area. The same specialties of medical oncologists were identified as in the calculation of patient travel time. The denominator was the number of stage IV lung, colorectal, breast, and prostate cancer patients within the defined area. The same selection criteria for the study population were used to select four types of metastatic cancer patients from 1992 to 2002. Local areas were created based on the driving area for clinical care (DACC) method [12, 37]. Briefly, the DACC method uses driving times (ZIP code to ZIP code) to create local areas around the original patient zip code where other patients with similar conditions reside. The driving time is then expanded incrementally to add patients from the next
123
202
closest zip codes until a threshold number of patients have been reached. The DACC method has been shown to account for additional treatment variation in area practice style than other local area definitions [37, 38]. We defined an area of 50 patients as the measure of a local health care market [39]. The measure was reported as the number of medical oncologists per 1,000 stage IV cancer patients. Similarly, local area per capita number of hospice programs was measured at the patient residence ZIP code level for the year of diagnosis, with the same denominator as defined in local area per capita number of medical oncologists and in the local market size of 50 patients. The numerator of the measure was the number of hospice programs within the area of a MBC patient residence ZIP code. The measure was reported as the number of hospice programs per 1,000 stage IV cancer patients. Local area chemotherapy rate across cancers was measured at each MBC patient residence ZIP code level across years. The denominator was the same as in local area per capita number of medical oncologists and per capita number of hospice programs. The numerator was the number of patients in the denominator who was identified to have received chemotherapy. The measure was reported as the number of patients who received chemotherapy per 100 stage IV cancer patients.
Breast Cancer Res Treat (2015) 149:199–209
In the aggregate multivariate logistic regression models, age and access variables were specified as categorical or continuous variables with interaction terms between two variables, controlling for other covariates. Both age and the squared term of age were added in the aggregate model to explore the possible nonlinear effects of age. In the subgroup analysis, multivariate logistic regression models were based on two age groups of below/above the median age of 75 years, three age groups with 10-year ranges and five age groups with 5-year ranges, with access variables specified in quartiles, controlling for other covariates. In the aggregate multivariate logistic regression analysis, Chow F-statistics were used to test whether the effect of access on patient chemotherapy choice was stronger among older patients [44]. In the stratified subgroup analysis, the coefficient of each access variable was estimated and statistical significance was assessed for each age category. All data analysis was performed using SAS 9.00 at the significance level .05 (SAS Institute Inc, Cary, NC). This study was approved by the University of Iowa Institutional Review Board. SEER-Medicare officials reviewed and approved the findings for confidentiality assurances.
Results Covariates Control variables included patient demographic characteristics (race and marital status), socioeconomic characteristics (income and education at census level), diagnosis year, residing SEER regions, rural/urban area characteristics, clinical characteristics (comorbidities, hormone receptor status), and radiation therapy. To measure comorbidities, we adopted the Klabunde version of the Charlson comorbidity index [40–42]. The ICD-9-CM diagnosis codes, ICD-9-CM procedure codes, and HCPCS codes to determine major medical conditions for cancer patients are included in Online Resources Table A3-1 and Table A3-2. The source of covariates is available in Online Resources Table A4. A detailed description of covariates in the model is given in Online Resource Table A5.
The average age of the study population was 77.6, ranging from 66.0 to 103.2. There were 30.2 % of 4533 MBC patients who used chemotherapy within a 183-day time window of diagnosis. The chemotherapy rate decreased steadily with age, with 49.3, 40.5, 31.8, 17.6, and 7.7 % in the subgroups aged 66–69, 70–74, 75–79, 80–84, and 85?, respectively. Table 2 shows the clinical and demographic characteristics of the patients by chemotherapy use. Patients who used chemotherapy averaged 74.2-year old while those who did not averaged 79.1-year old. Patients who used chemotherapy were often married and had higher socioeconomic status; those who had no comorbidities or had negative hormone receptor status appeared more likely to use chemotherapy.
Statistical analysis
Bivariate association between access variables and chemotherapy use
To assess the heterogeneity of the association between access and chemotherapy use across age, we began with a statistical test for interaction to determine whether the effect of access to cancer care on chemotherapy use was modified by age in the whole group; we followed with a stratified subgroup analysis [43]. Chemotherapy use was the dependent variable. Age and access variables were key independent variables.
In the whole group analysis, chemotherapy use increased with local area chemotherapy rate (P \ 0.0001). Although it appeared that MBC patients living farther away from the nearest oncologist practice or those living in an area with a higher local area per capita number of oncologists used less chemotherapy, bivariate analysis did not achieve statistical significance (P = 0.855, 0.102 respectively). In addition, there was no obvious trend in chemotherapy use associated
123
Breast Cancer Res Treat (2015) 149:199–209
203
Table 2 Patient characteristic by chemotherapy use
Overall
Percent chemo use
No. of cases
30.2
4,533
P value
Percent chemo use
49.3
No. of cases
SEER area \0.0001*
Age at diagnosis 66–69
Table 2 continued
747
0.9718
Louisiana
34.9
146
Connecticut
31.6
531
31.3
160
70–74
40.5
1,111
Kentucky/Rural Georgiad
75–79
31.8
1,068
Los Angeles
30.9
479
80–84
17.6
856
New Jersey
30.8
454
7.7
751
San Jose
30.4
168
Detroit
30.1
657
1,071
Great California/Hawaiid
30.0
337
30.0
231
85 or older
\0.0001
Area chemo rate 1st quartile
23.7
2nd quartile
29.5
1,063
Atlanta
3rd quartile
31.9
1,075
Seattle
29.4
346
4th quartile
35.8
1,075
Iowa
29.0
507
San Francisco
28.5
253
Utah
27.9
136
New Mexico
25.0
128
1992
26.2
374
1993
29.8
336
1,071
1994
26.5
306
30.7
339
Travel time 1st quartile
30.6
2,141
3rd quartile
30.2
1,070
4th quartile
29.6
1,073
0.8552
2nd quartile
No. of oncologists 1st quartile
Year of diagnosis
0.1016 32.8
0.6677
2nd quartile
30.9
1,070
1995
3rd quartile
29.0
1,071
1996
30.1
336
4th quartile
28.3
1,072
1997
30.5
318
No. of hospices
1998
30.1
289
1st quartile
1999
34.0
312
1,056
2000
31.2
650
31.1
644
30.2
629
2nd quartile
30.7 31.0
1,069
0.7,722
3rd quartile
30.2
1,076
2001
4th quartile
29.1
1,083
2002
Race and ethnicity
0.7810
White
30.1
3,740
Black
29.5
488
Other
31.8
305
Single
25.7
474
Married
39.9
1,391
Divorced/Widowed
25.6
2,497
Unknown
29.8
171
Percentage of low-income residents was above median in 2000 for beneficiary census tract
b
Positive
29.7
1,448
Negative
41.5
1,151
Unknown
23.7
1,934
Percentage of low-income residents above mediana 32.3
1,511
Yes
29.7
1,516
Unknown
28.5
1,506
Comorbiditiesb 0
32.3
1
16.2
734
2
8.4
381
SEER areas were combined based on the proximity due to small size
Big metro
30.1
2,754
Metro
30.7
1,136
Urban
31.6
263
Less urban
28.8
302
Rural
26.3
76
Area chemo rate 100 9 percentage chemotherapy use among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
\0.0001*
Travel time driving time from the patient residency to the nearest oncologist practice. Travel time only has three quartiles because the first quartile and the second quartile are merged into one category due to the first quartile value = 0
0.8796
No. of oncologists 1,000 9 per capita number of oncologists among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
3,418
Rural/urban area characteristicsc
Rural/urban area characteristics were based on area resource file that is available in the PEDSF file
d
0.0658
No
Comorbidities exclude AIDS and all metastatic tumors. The following chronic conditions were measured chronic obstruction pulmonary disease, cerebrovascular disease, congestive heart failure, diabetes, dementia, myocardial infarction, peripheral vascular disease, peptic ulcer disease, paralysis, renal disease, rheumatologic disease
c
\0.0001*
Hormone receptor status
* P \ 0.05 a
\0.0001*
Marital status
P value
No. of hospices 1000 9 per capita number of hospices among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patients
123
204
with the local area per capita number of hospices across metastatic cancers (P = 0.772). Figure 1 shows the bivariate analysis between chemotherapy use and geographic access variables in quartiles by age subgroups. The increasing trend in chemotherapy use associated with local area chemotherapy rate remained in each age category except in the age group of 80–84 years (Fig. 1a). Bivariate association between chemotherapy use and patient travel time to the nearest oncologist in the age subgroup analysis did not achieve statistical significance, although in older age subgroups there was lower chemotherapy use at longer travel times (Fig. 1b). Neither was the association between chemotherapy use and local area per capita number of oncologists significant (Fig. 1c).
Breast Cancer Res Treat (2015) 149:199–209
a
b Summary of the results of the interaction term approach and subgroup analysis Because the low to moderate correlation between the four access variables suggested that they measure distinct aspects of geographic access (Online Resource Table A6), we have reported the results based on the model that included the four access variables. In the aggregate models including access variables and age (Online Resource Table A7–A10), results showed that age was consistently significantly associated with chemotherapy choice in all models. We included additional results of each access indicator variable in separate models as online resources (Tables A11–A15). Both the aggregate model and the subgroup analysis showed that the effect of local area chemotherapy rate was consistently positively associated with chemotherapy use. However, the interactions between this variable and age did not achieve statistical significance in the aggregate models. In Table 3, where the results of the five age subgroup analysis are summarized, the parameter estimate of the continuous variable suggested that the odds ratios of chemotherapy use would increase by 1.05 (95 % CI 1.03–1.07), 1.03 (95 % CI 1.01–1.05), 1.03 (95 % CI 1.02–1.05), and 1.05 (95 % CI 1.01–1.09) in the age subgroup of 66–69, 70–74, 75–79, and 85? years, respectively, if one additional patient was treated with chemotherapy per 100 metastatic cancer patients of 65? years old in the area. We originally hypothesized that the per capita oncologist supply was positively associated with chemotherapy use. However, the direction of this access variable has been consistently negative on chemotherapy use and the magnitude was not statistically significant in the aggregate models. Neither were the interaction terms significant. Noticeably, in the five age subgroup analyses, local area oncologist supply was significantly negatively associated
123
c
1st Quartile: 9 -31 2nd Quartile: 32 -38 3rd Quartile: 39 -44 4th Quartile: 45 -70
1st -2nd Quartile: 0 -0.17 hr 3rd Quartile: 0.18 -0.32 hr 4th Quartile: 0.33 -3.72 hr
1st Quartile: 13 -62 2nd Quartile: 63 -99 3rd Quartile: 100-157 4th Quartile: 158-713
Fig. 1 Chemotherapy use by quartiles of local area chemotherapy rate, patient travel time to the nearest oncologist practice, and local area per capita number of oncologists across age subgroups. a. Chemotherapy use by quartiles of local area chemotherapy rate among 100 metastatic breast, colorectal, lung, and prostate cancer patients. b. Chemotherapy use by quartiles of patient travel time to the nearest oncologist practice at the patient ZIP code level. c. Chemotherapy use by quartiles of local area per capita number of oncologists among 1,000 metastatic breast, colorectal, lung, and prostate cancer patients
0.61 (0.28–1.30)
Reference
0.72 (0.43–1.21)
Second quartile
0.5127
First quartile
OR (95 % CI)
P valueà
No. of oncologists
0.1996
OR (95 % CI)
1.06 (0.66–1.69)
Fourth quartile
P value
0.81 (0.53–1.24)
Reference
Third quartile
Second quartile
First quartile
OR (95 % CI)
P valueà
0.5298
1.05 (1.03–1.07)
Travel time
\0.0001*
OR (95 % CI)
1.81 (1.08–3.02) 3.14 (1.80–5.48)
Third quartile Fourth quartile
P value
Reference
1.40 (0.86–2.26)
Second quartile
0.0006*
First quartile
OR (95 % CI)
P valueà
0.84 (0.55–1.27)
Reference
0.3563
0.91 (0.63–1.32)
1.09 (0.78–1.51)
Reference
0.6911
1.65 (1.09–2.50) 1.69 (1.10–2.59)
1.35 (0.90–2.03)
Reference
0.0657
Categorical model
Categorical model
Continuous model
70–74
66–69
Area chemo rate
Variables
0.77 (0.40–1.50)
0.4456
1.03 (1.01–1.05)
0.0006*
Continuous model
0.81 (0.52–1.25)
Reference
0.4702
0.82 (0.53–1.27)
0.98 (0.68–1.40)
Reference
0.6694
1.94 (1.22–3.08) 2.21 (1.36–3.62)
1.31 (0.83–2.08)
Reference
0.0045*
Categorical model
75–79
Table 3 Adjusted estimate of the effect of geographic access on chemotherapy use in age subgroups
0.479 (0.23–1.01)
0.0516
1.03 (1.02–1.05)
0.0006*
Continuous model
1.32 (0.72–2.41)
Reference
0.2974
0.67 (0.37–1.23)
0.76 (0.47–1.22)
Reference
0.3093
1.22 (0.64–2.33) 1.58 (0.84–2.98)
1.80 (0.99–3.29)
Reference
0.2037
Categorical model
80–84
0.39 (0.12–1.27)
0.1181
1.02 (0.99–1.04)
0.1839
Continuous model
0.57 (0.21–1.57)
Reference
0.0463*
0.67 (0.25–1.82)
0.21 (0.08–0.56)
Reference
0.0076*
2.718 (0.90–8.22) 3.87 (1.30–11.53)
1.64 (0.56–4.76)
Reference
0.0857
Categorical model
85?
0.89 (0.14–5.61)
0.9022
1.05 (1.01–1.09)
0.0258*
Continuous model
Breast Cancer Res Treat (2015) 149:199–209 205
123
123
1.00 (0.997–1.002)
1.26 (0.71–2.23)
Fourth quartile P value
OR (95 % CI)
1.00 (0.99–1.01)
0.9256
1.00 (0.996–1.001)
0.3606
0.74 (0.44–1.23)
0.84 (0.53–1.33)
0.90 (0.59–1.38)
Reference
0.7007
0.94 (0.55–1.60)
0.73 (0.46–1.17)
Categorical model
1.00 (0.99–1.00)
0.4666
1.00 (0.997–1.002)
0.6689
Continuous model
P value of a Chow F-statistics testing whether all of the interaction terms are simultaneously equal to zero
1.37 (0.71–2.65)
0.90 (0.49–1.66)
0.77 (0.43–1.38)
Reference
0.3025
0.69 (0.31–1.56)
1.02 (0.54–1.96)
Categorical model
80–84
0.39 (0.12–1.27)
0.1435
1.00 (0.994–1.002)
0.3348
Continuous model
4.68 (1.33–16.55)
3.36 (1.04–10.90)
5.85 (1.98–17.23)
Reference
0.0149*
0.17 (0.04–0.72)
0.24 (0.07–0.81)
Categorical model
85?
1.01 (0.90–1.02)
0.3643
0.992 (0.985–0.999)
0.0277*
Continuous model
No. of hospices 1,000 9 per capita number of hospices among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patients. All four access variables were included in the model. Other variables in the model include patient race, marital status, HR status, radiation therapy use, census tract median household income and education level, residency urban/rural characteristics, residency SEER area, and diagnosis year
No. of oncologists 1,000 9 per capita number of oncologists among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
Travel time driving time from the patient residency to the nearest oncologist practice. Travel time only has three quartiles because the first quartile and the second quartile are merged into one category due to the first quartile value = 0
Area chemo rate 100 9 percentage chemotherapy use among metastatic breast, prostate, lung, and colorectal cancer patients in the local area of 50 patients (four types of metastatic cancer patients) surrounding the residency of a metastatic breast cancer patient
* Significance level at 0.05 level
à
1.10 (0.69–1.75)
1.34 (0.88–2.06)
1.10 (0.73–1.65)
Reference
0.5354
0.63 (0.37–1.06)
0.73 (0.47–1.14)
Continuous model
75–79
P value of Wald v2 test testing whether the coefficient is significantly different from zero
0.61 (0.28–1.30)
0.9239 (0.56–1.52)
Third quartile
0.1424
Reference
1.10 (0.67–1.79)
Second quartile
0.6677
First quartile
OR (95 % CI)
P valueà
No. of hospices
0.7454
OR (95 % CI)
1.01 (0.53–1.89)
Fourth quartile
P value
0.83 (0.48–1.43)
Categorical model
Categorical model
Continuous model
70–74
66–69
Third quartile
Variables
Table 3 continued
206 Breast Cancer Res Treat (2015) 149:199–209
Breast Cancer Res Treat (2015) 149:199–209
with chemotherapy use in the subgroup of 85? years old. This effect was not observed among younger age groups.
Discussion It has been widely shown in the literature that older patients with breast cancer were less likely to be treated with chemotherapy [28–31]. Researchers suggested that these differences may be attributed to the age-related bias that older patients would not benefit from chemotherapy to the same degree as their younger counterparts [45]. Our study found that geographic access can explain a portion of the agerelated difference in chemotherapy use in elderly patients with MBC. The insufficient evidence regarding the benefits versus the harm of chemotherapy for elderly patients, coupled with the biological and socioeconomic changes of aging, are major sources of uncertainties for older patients. Older patients suffer from higher rates of comorbid conditions (not shown). The presence of comorbidities appeared strongly associated with not receiving chemotherapy. However, age should not be considered as a surrogate for comorbidities. Controlling for comorbidities, age was independently associated with lower chemotherapy use in our analysis, consistent with previous studies. In addition to comorbidities, considerable age-related heterogeneity in functional ability, tolerance to chemotherapy, and patient preferences exists in elderly women [30, 46–49]. These uncertainties may cause non-clinical factors such as access to matter more to older patients. The direction of the cumulative odds ratios across age groups suggested that local area chemotherapy rate affected chemotherapy choice in a consistent way, with higher chemotherapy rate percentiles predicting higher odds of receiving chemotherapy. These results highlight the importance of local area practice culture and provide insight into how local culture may influence treatment choices of oncologists in the absence of sufficient clinical evidence [8, 50]. The lower chemotherapy rate in older patients may reflect a lack of patient-specific data on efficacy and patient preferences in older patients. The best clinical practice may be preference-based discussion between doctors and patients. It should be noted that although the odds ratio of chemotherapy rate by age are statistically significant, the magnitude is small. Therefore, the interpretation of clinical impact should not be overstated. For patients aged 85? years old, local area per capita number of oncologists also affected chemotherapy choice. This suggested that the effect of access variables may be additive instead of being multiplicative with age suggested by the interaction terms. Contrary to our hypothesis that
207
higher per capita supply of services is associated with higher use [51], oncologist supply had a consistently negative effect on receiving chemotherapy, although this effect did not reach statistical significance in younger age groups. Oncologists may be able to spend more time with each patient in areas with higher per capita number of oncologists and evaluate alternatives, such as hormone therapy and hospice care. Especially for the oldest patients who were 85? years old, treatment benefit and harm of chemotherapy need to be considered in the context of patient life expectancy, the nature and the symptoms of the disease, and patient quality of life [52]. Another possible explanation is that the higher number of oncologists was likely more concentrated in academic practice where they may be more focused on patient preferences. The other access variable, local area per capita number of hospices, was shown to not be associated with chemotherapy use, which may be attributed to a variety of treatment options available to patients with MBC other than chemotherapy and hospice care. Our study has some limitations. Complete data for hormone therapy were not available, although we controlled for hormone receptor status in the regression analysis. Among the HR ? subgroup, we observed a similar pattern of decreasing chemotherapy use with age and local area chemotherapy rate was the only significant factor. There were other unmeasured confounders as discussed earlier, which may not be separated from the effect of age and likely contributed to the insignificant interaction terms in the whole group analysis. This study has several strengths, and opened venues for future studies. Due to heterogeneous patient characteristics, estimating all patients aggregately may mask results. By estimating more homogeneous subgroups, we found that the more uncertain the evidence with age, the more geographic access influenced chemotherapy use. The impact of geographic access was the greatest among the oldest patients for whom the clinical evidence was the least certain. Our results suggested that strategies improving access to cancer care may need to tailor to individual organizations, oncologists and patients in local communities where different health care systems and practices often serve the mix of population rooted in similar environments and characteristics. In addition, our results do not assess the extent to which treatment rates are appropriate among the different age groups. Other factors, such as socioeconomic status, may also exert compounded effects with older age, which need to be examined in future studies. Acknowledgments The authors acknowledge the constructive comments of Dr. John Brooks, Dr. Elizabeth Chrischilles, Dr. Linnea Polgreen, Dr. Alexandra Thomas, and Dr. Bernard Sorofman at the University of Iowa College of Pharmacy. The authors would like to thank Ms. Chris Welch for her support during the manuscript
123
208
Breast Cancer Res Treat (2015) 149:199–209
submission. This study used the linked 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, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database. The collection of the California cancer incidence data used in this study was supported by the California Department of Public Health as part of the statewide cancer reporting program mandated by California Health and Safety Code Sect. 103885; the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program under contract N01-PC-35136 awarded to the Northern California Cancer Center, contract N01-PC-35139 awarded to the University of Southern California, and contract N02-PC-15105 awarded to the Public Health Institute; and the Centers for Disease Control and Prevention’s National Program of Cancer Registries, under agreement #U55/CCR921930-02 awarded to the Public Health Institute. The ideas and opinions expressed herein are those of the author(s) and endorsement by the State of California, Department of Public Health the National Cancer Institute, and the Centers for Disease Control and Prevention or their Contractors and Subcontractors is not intended nor should be inferred. Disclosures
None.
Ethical standards The study complies with the current laws in the U.S.A where the study was performed. Funding This work was supported by the University of Iowa College of Pharmacy Dissertation Fellowship.
References 1. Wennberg JE, Fisher ES, Stukel TA (2004) Use of hospitals, physician visits, and hospice care during last six months of life among cohorts loyal to highly respected hospitals in the United States. BMJ 328:606 2. Wennberg JE, Fisher ES, Goodman D, Skinner JS (2008) Tracking the care of patients with severe chronic illness: the Dartmouth Atlas of Health Care 2008. The Dartmouth Institute for Health Policy and Clinical Practice, Hanover 3. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ (2004) Variations in the longitudinal efficiency of academic medical centers. Health Aff (Millwood) Suppl Variation:VAR19. doi:10.1377/ hlthaff.var.19 4. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL (2003) The implications of regional variations in medicare spending. Part 1: the content, quality, and accessibility of care. Ann Intern Med 138(4):273 5. Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL (2003) The implications of regional variations in medicare spending. Part 2: health outcomes and satisfaction with care. Ann Intern Med 138(4):288 6. Weinstein JN, Bronner KK, Morgan TS, Wennberg JE (2004) Trends and geographic variations in major surgery for degenerative diseases of the hip, knee, and spine. Health Aff (Millwood) Suppl Variation:VAR81. doi:10.1377/hlthaff.var.81 7. MedPac (2011) Regional variation in medicare service use. MedPAC, Washington, DC 8. Fisher ES, Wennberg JE (2003) Health care quality, geographic variations, and the challenge of supply-sensitive care. Perspect Biol Med 46(1):69
123
9. Wennberg JE, Fisher ES, Skinner JS (2002) Geography And The Debate Over Medicare Reform. Health Aff hlthaff.w2.96. doi:10. 1377/hlthaff.w2.96 10. Dartmouth Atlas of Health Care (2013) Supply-sensitive care. http://www.dartmouthatlas.org/keyissues/issue.aspx?con=2937. Accessed 1 Sep 2013 11. Dartmouth Atlas of Health Care (2013) Preference-sensitive care. http://www.dartmouthatlas.org/keyissues/issue.aspx?con=2938. Accessed 1 Sep 2013 12. Brooks JM, Cook EA, Chapman CG, Kulchaitanaroaj P, Chrischilles EA, Welch S, Robinson J (2014) Geographic variation in statin use for complex acute myocardial infarction patients: evidence of effective care? Med Care 52(Suppl 3):S37. doi:10.1097/ MLR.0b013e3182a7fc3d 13. Hadley J, Waidmann T, Zuckerman S, Berenson RA (2011) Medical spending and the health of the elderly. Health Serv Res 46(5):1333. doi:10.1111/j.1475-6773.2011.01276.x;10.1111/j. 1475-6773.2011.01276.x 14. Kaestner R, Silber JH (2010) Evidence on the efficacy of inpatient spending on medicare patients. Milbank Q 88(4):560. doi:10.1111/j.1468-0009.2010.00612.x;10.1111/j.1468-0009. 2010.00612.x 15. Reschovsky JD, Hadley J, Saiontz-Martinez CB, Boukus ER (2011) Following the money: factors associated with the cost of treating high-cost Medicare beneficiaries. Health Serv Res 46(4):997. doi:10.1111/j.1475-6773.2011.01242.x;10.1111/j. 1475-6773.2011.01242.x 16. Silber JH, Kaestner R, Even-Shoshan O, Wang Y, Bressler LJ (2010) Aggressive treatment style and surgical outcomes. Health Serv Res 45(6 Pt 2):1872. doi:10.1111/j.1475-6773.2010.01180. x;10.1111/j.1475-6773.2010.01180.x 17. Zuckerman S, Waidmann T, Berenson R, Hadley J (2010) Clarifying sources of geographic differences in medicare spending. New Engl J Med 363(1):54. doi:10.1056/NEJMsa0909253 18. IOM (Institue of Medicine) (2013) Variation in health care spending: targeting decision making, not geography. The National Academies Press, Washington, DC 19. NCCN Clinical Practice Guidelines in Oncology: Breast Cancer V3 (2014) http://www.nccn.org. Accessed 1 May 2014 20. M.D. Anderson Cancer Center Cancer Treatment Algorithms: Breast Cancer (Invasive) V12 (2013) http://www.mdanderson. org. Accessed 1 Mar 2014 21. Hutchins LF, Unger JM, Crowley JJ, Coltman CA, Albain KS (1999) Underrepresentation of patients 65 years of age or older in cancer-treatment trials. New Engl J Med 341(27):2061 22. Scher KS, Hurria A (2012) Under-representation of older adults in cancer registration trials: known problem, little progress. J Clin Oncol 30(17):2036–2038. doi:10.1200/jco.2012.41.6727 23. Green MC, Hortobagyi GN (2008) Chemotherapy for breast cancer. In: Hunt KK, Robb GL, Strom EA, Ueno NT (eds) Breast Cancer, 2nd edn. Springer, New York. doi:10.1007/978-0-387-34952-7_12 24. Howlander N, Noone NA, Krapcho M (2010) Cancer Statistics Review, 1975–2010 25. Du XL, Goodwin JS (2001) Patterns of use of chemotherapy for breast cancer in older women: findings from medicare claims data. J Clin Oncol 19(5):1455 26. Eaker S, Dickman PW, Bergkvist L, Holmberg L (2006) Differences in management of older women influence breast cancer survival: results from a population-based database in Sweden. PLoS Med 3(3):e25 27. Freyer G, Braud AC, Chaibi P, Spielmann M, Martin JP, Vilela G, Guerin D, Zelek L (2006) Dealing with metastatic breast cancer in elderly women: results from a French study on a large cohort carried out by the ‘Observatory on Elderly Patients’. Ann Oncol 17(2):211
Breast Cancer Res Treat (2015) 149:199–209 28. Enger SM, Thwin SS, Buist DSM, Field T, Frost F, Geiger AM, Lash TL, Prout M, Yood MU, Wei F (2006) Breast cancer treatment of older women in integrated health care settings. J Clin Oncol 24(27):4377 29. Giordano SH, Hortobagyi GN, Kau SC, Theriault RL, Bondy ML (2005) Breast cancer treatment guidelines in older women. J Clin Oncol 23(4):783. doi:10.1200/jco.2005.04.175 30. Hurria A, Togawa K, Mohile SG, Owusu C, Klepin HD, Gross CP, Lichtman SM, Gajra A, Bhatia S, Katheria V, Klapper S, Hansen K, Ramani R, Lachs M, Wong FL, Tew WP (2011) Predicting chemotherapy toxicity in older adults with cancer: a prospective multicenter study. J Clin Oncol 29(25):3457–3465. doi:10.1200/jco.2011.34.7625 31. Muss HB, Woolf S, Berry D, Cirrincione C, Weiss RB, Budman D, Wood WC, Henderson IC, Hudis C, Winer E, Cohen H, Wheeler J, Norton L (2005) Adjuvant chemotherapy in older and younger women with lymph node-positive breast cancer. JAMA 293(9):1073. doi:10.1001/jama.293.9.1073 32. Noone AM, Lund JL, Mariotto A, Cronin K, McNeel T, Deapen D, Warren JL (2014) Comparison of SEER treatment data with medicare claims. Med Care. doi:10.1097/mlr.0000000000000073 33. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF (2002) Overview of the SEER-medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 40(8 Suppl):IV. doi:10.1097/01.mlr.0000020942.47004.03 34. Warren JL, Harlan LC, Fahey A, Virnig BA, Freeman JL, Klabunde CN, Cooper GS, Knopf KB (2002) Utility of the SEER-Medicare data to identify chemotherapy use. Med Care 40(8 Suppl):IV-55-61. doi:10.1097/01.mlr.0000020944.17670.d7 35. Chen-Hardee S, Chrischilles EA, Voelker MD, Brooks JM, Scott S, Link BK, Delgado D (2006) Population-based assessment of hospitalizations for neutropenia from chemotherapy in older adults with non-Hodgkin’s lymphoma (United States). Cancer Causes Control 17(5):647–654. doi:10.1007/s10552-005-0502-4 36. Brooks JM, Chrischilles EA, Scott SD, Chen-Hardee SS (2003) Was breast conserving surgery underutilized for early stage breast cancer? Instrumental variables evidence for Stage II patients from Iowa. Health Serv Res 38(6p1):1385. doi:10.1111/j.1475-6773. 2003.00184.x 37. Fang G, Brooks JM, Chrischilles EA (2010) A new method to isolate local-area practice styles in prescription use as the basis for instrumental variables in comparative effectiveness research. Med Care 48(8):710. doi:10.1097/MLR.0b013e3181e41bb2;10. 1097/MLR.0b013e3181e41bb2 38. Fang G, Brooks JM, Chrischilles EA (2012) Comparison of instrumental variable analysis using a new instrument with risk adjustment methods to reduce confounding by indication. Am J
209
39.
40.
41.
42. 43.
44. 45.
46.
47.
48.
49.
50.
51.
52.
Epidemiol 175(11):1142. doi:10.1093/aje/kwr448;10.1093/aje/ kwr448 Lu-Yao GL, Albertsen PC, Moore DF, Shih W, Lin Y, DiPaola RS, Yao SL (2008) Survival following primary androgen deprivation therapy among men with localized prostate cancer. JAMA 300(2):173. doi:10.1001/jama.300.2.173 Charlson ME, Pompei P, Ales KL, MacKenzie CR (1987) A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 40(5):373 Klabunde CN, Potosky AL, Legler JM, Warren JL (2000) Development of a comorbidity index using physician claims data. J Clin Epidemiol 53(12):1258 Klabunde CN, Warren JL, Legler JM (2002) Assessing comorbidity using claims data: an overview. Med Care 40(8):26–35 Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM (2007) Statistics in medicine: reporting of subgroup analyses in clinical trials. New Engl J Med 357(21):2189. doi:10.1056/NEJMsr077003 Chow GC (1960) Tests of equality between sets of coefficients in two linear regressions. Econometrica 28(3):591 Crivellari D, Aapro M, Leonard R, von Minckwitz G, Brain E, Goldhirsch A, Veronesi A, Muss H (2007) Breast cancer in the elderly. J Clin Oncol 25(14):1882. doi:10.1200/jco.2006.10.2079 Silliman RA, Balducci L, Goodwin JS, Holmes FF, Leventhal EA (1993) Breast cancer care in old age: what we know, don’t know, and do. J Natl Cancer Inst 85(3):190 Koedoot CG, de Haan RJ, Stiggelbout AM, Stalmeier PFM, de Graeff A, Bakker PJM, de Haes JCJM (2003) Palliative chemotherapy or best supportive care? A prospective study Explaining Patients’ treatment preference and choice. Br J Cancer 89:2219 McQuellon RP, Muss HB, Hoffman SL, Russell G, Craven B, Yellen SB (1995) Patient preferences for treatment of metastatic breast cancer: a study of women with early-stage breast cancer. J Clin Oncol 13(4):858 Bouchardy C, Rapiti E, Blagojevic S, Vlastos A, Vlastos G (2007) Older female cancer patients: importance, causes, and consequences of undertreatment. J Clin Oncol 25(14):1858. doi:10.1200/jco.2006.10.4208 Sirovich B, Gallagher PM, Wennberg DE, Fisher ES (2008) Discretionary decision making by primary care physicians and the cost of U.S. Health care. Health Aff (Millwood) 27(3):813. doi:10.1377/hlthaff.27.3.813;10.1377/hlthaff.27.3.813 Buntin MB, Garten AD, Paddock S, Saliba D, Totten M, Escarce JJ (2005) How much is postacute care use affected by its availability? Health Serv Res 40(2):413. doi:10.1111/j.1475-6773. 2005.00365.x Ganz PA (ed) (2007) Cancer survivorship: today and tomorrow. Springer, New York
123
