Screening mammography rate and predictors followingtreatment for colorectal cancer

A. Marshall McBean & Xinhua Yu & Beth A. Virnig

Received: 23 October 2008 /Accepted: 19 January 2009 /Published online: 18 February 2009# Springer Science + Business Media, LLC 2009

AbstractIntroduction Colorectal cancer survivors remain at risk forbreast cancer. Thus, it is important to determine if screeningmammography rates are reduced by the diagnosis andtreatment of incident colorectal cancer.Methods Mammography rates among 7,666 67–79 year-old stage 0-III colorectal cancer survivors were comparedwith rates among 36,433 age-, race/ethnicity-, SEERarea-matched women controlling for pre-diagnosis mam-mography, stage, chemotherapy, income, co-morbidities,treatment in teaching hospital, number of physicianvisits, and gynecologist visits.Results In the first 2 years after diagnosis, the survivors’rate (49.7/100) was 4.2% higher than the controls’ (47.6/100), p<0.001. It was 7.5% higher in the next 2 years, 54.5/100 versus 49.7/100, p<0.001. The higher rates resultedfrom significantly greater rates among survivors withoutprior mammography, 30.9/100, compared with their con-trols (25.3/100) in the first 2 years, for example (O.R. =1.23, 95% C.I. = 1.15–1.32). The strongest predictors ofpost-diagnosis mammography were pre-diagnosis mam-mography (O.R. = 5.76, 95% C.I. = 5.19–6.38), visiting agynecologist (O.R. = 1.83, 95% C.I. = 1.55–2.16),chemotherapy (O.R. = 1.61, 95% C.I. = 1.40–1.86), andmore than nine physician visits. Increasing Charlson scoresand cancer stage were associated with lower mammographyrates.Discussion/Conclusions Overall, the competing demandsof cancer diagnosis and treatment did not reduce mammog-

raphy rates, and these events were associated with increasedrates among previous non-users.Implications for cancer survivors The low mammographyrate among survivors with no history of a prior mammo-gram means that the physicians treating these women mustemphasize the need for such care.

Keywords Mammography . Colorectal neoplasms .

Elderly . Medicare . Health behavior . Health services

Introduction

Approximately 75,000 American women will be newlydiagnosed with colorectal cancer in 2009 [1]. Approxi-mately 63% of them will have local disease, 29% regional,and 6% distant. In women diagnosed with colorectalcancer between 1996 and 2003, the five-year relativesurvival rates were 99% and 84% for those ≥ 50 years ofage with local and regional disease, respectively. Theaverage life expectancy for women 65 years of age is20 years [2]. The lifetime probability of developing breastcancer after this age is 5.7%, and it increases with age [1].Finally, the risk of breast cancer is not different betweenwomen with colorectal cancer and women with no historyof cancer [3, 4]. Given these facts, screening mammogra-phy is an important preventive service for colorectalcancer survivors

The use of mammography in women, including elderlywomen, has been described frequently [5–7]. Among themost recent data, CDC’s Behavioral Risk Factor Surveil-lance System (BRFSS) reported that in 2006, 79.1% ofwomen 65 years of age and older had received amammogram in the past 2 years [5]. Rakowski et al.(2006) used the 2002-2003 Health Information and

J Cancer Surviv (2009) 3:12–20DOI 10.1007/s11764-009-0080-7

A. M. McBean (*) :X. Yu : B. A. VirnigDivision of Health Policy and Management,School of Public Health, University of Minnesota,MMC 97, 420 Delaware St.,S.e., Minneapolis, MN 55455, USAe-mail: mcbea002@umn.edu

National Trends Surveys (HINTS) of the National CancerInstitute to measure recent (81.6%) and repeat (72.2%)mammography in women 42 to 75 years of age [8]. Thus,as these authors calculated, 88.5% (77.2%/81.6%) ofwomen with a recent mammogram had a prior on-schedulemammogram, leading them to conclude that, past mam-mography use is a strong predictor of future behavior.Interestingly, they found no difference in either the repeator recent mammography rates between women who hadever been diagnosed with cancer (excluding breast cancer)and those who had not. This latter finding is inconsistentwith the results we and others have reported for elderlylong-term uterine and female colorectal cancer survivors(those who survived 5 years or more) [9–11]. In thesereports, long-term survivors had significantly higher ratesof mammography than women with no known history ofcancer.

It is possible that women would change their priorbehavior regarding mammography after a recent cancerdiagnosis and be less likely to have mammography soonafter a diagnosis of cancer than after surviving 5 or moreyears. Cancer treatment may overwhelm patient who maydelay other health care services, particularly screening andother preventive services. The competing demands model[12] suggests that both patients and physicians may lowerthe priority for screening mammography, or be too busy toobtain it when involved in cancer treatment. Mandelblatt etal. (2006) reported that in the first year following thediagnosis of breast cancer, only 61.9% of women hadfollow-up mammograms [13] which are recommended forall of these women. It is unknown how the use ofmammography is affected when facing competing demandsof care for colorectal or other cancers, and when cancerpatients will resume mammography use after the cancerdiagnosis or treatment. Recently, Snyder et al. have focusedon the importance of the type of physician seen in recentcolorectal cancer survivors [14, 15]. They reported mam-mography rates of 19 to 60% among colorectal cancersurvivors during the first year of survivorship depending onthe mix of physicians seen by the patient. The range for thecumulative use of mammography over 5 years was 25% to66%. However, as they point out, they did not investigatethe role of pre-diagnosis use of mammography, and “pre-diagnosis service use may be one of the strongest predictorsof survivorship service use” [15].

The objective of this study is to examine the associationof newly diagnosed colorectal cancer and other factors,including the use of prior mammography, on the use ofscreening mammography in the first and second, and thirdand fourth years post-diagnosis among 67–79 year-oldMedicare fee-for-service women. We hypothesized thatmammography use among colorectal cancer patients woulddecrease after the cancer diagnosis because of the newly

diagnosed cancer and the added burden of cancer treatment,but that it would be influenced by pre-diagnosis use ofmammography.

Methods

We used the National Cancer Institute’s (NCI) Surveillance,Epidemiology, and End Results (SEER)-Medicare linkeddata from the 2006 merge in this study. Details of the SEERprogram and SEER-Medicare linked data can be found inthe SEER website and prior publications [16]. Briefly,SEER data contains all incident cancer cases and survivaldata for 1973 through 2002 from 14 to 16 population-basedregistries which currently cover approximately 26% of theU.S. population. About 94% of the cancer patients 65 yearsof age and older have been linked to Medicare enrollmentand claims files. We identified 47,809 elderly women withnewly diagnosed colorectal cancer in 1996–2001 using thefirst SEER cancer site recode variable (15–23, 25, 26, colonor rectal cancer, the first primary cancer). Becauseidentifying screening mammography use before the cancerdiagnosis requires 2 years of Medicare claims, we includedonly those beneficiaries 67 years of age and older at thetime of cancer diagnosis, and we excluded women over theage of 79, resulting in 21,390 women. Those with SEERmodified AJCC (American Joint Committee on Cancer)stage IV cancer (n=3,345), and those with unstaged disease(n=882) were excluded, also. This resulted in 1,590patients with AJCC stage 0, 5,401 with stage I, 5,842 withstage III and 4,690 with stage III disease; (total n=17,163).In addition, we excluded those died or were admitted tohospice within a year, n=1,678 and 170, respectively.Similar to other studies using the linked SEER–Medicaredatabase, we further restricted the cohort to women likely tohave a complete history of Medicare reimbursed servicesby sequentially excluding those who did not have bothMedicare Part A and Part B coverage (n=1,222), wereenrolled in managed care (n=4,761), or had end-stage renaldisease (n=41) 2 years before and 4 years after the cancerdiagnosis, resulting in 9,291 women with stage 0-III cancer.

Adjuvant chemotherapy for colorectal cancer was iden-tified using the SEER-Medicare National Claims History(NCH), Inpatient, and Outpatient claims. Based on theestablished algorithm that employs both chemotherapy drugand administration information [17], we further searchedcommon chemotherapy drugs used in the treatment ofcolorectal cancer using HCPCS codes: 5-fluorouracil(J9190) (5-FU), leucovorin (J0640), floxuridine (J9200),oxaliplatin (J9263), irinotecan (J9206), and capecitabine(J8520, J8521). Almost all patients who had chemotherapydrug information were treated with 5-FU (99%). The timeto the initiation of chemotherapy was defined as the period

J Cancer Surviv (20093:12–20 13

from the date of cancer diagnosis (or the date of surgery forthose who had surgery) to the date of the first drug claim.We included only those who initiated chemotherapy within6 months (n=8,778) because the purpose of the chemo-therapy after 6 months (5.5% of total) may have been totreat recurrence [18, 19]. The duration of chemotherapywas estimated as the time between the first and last drugclaim [20]. We included only those whose chemotherapylasted less than 1 year (n=7,674) because the usual periodof chemotherapy is 6 months [21].

A random sample of 5% of women with no knownhistory of cancer, residing in the same SEER areas isavailable from the SEER-Medicare database (n=213,434).Similar enrollment exclusion criteria were applied to them.Each cancer case was individually matched to up to fivenon-cancer controls (1:5 matching) by race/ethnicity(White, Black, and other), age (± 5 years), and SEER area,resulting in 36,433 women for the non-cancer controlgroup, and 7,666 cancer patients.

The main outcome variable in this study is screeningmammography use. Mammograms that occurred within the2 years before and up to 5 years after the month of cancerdiagnosis were identified using the SEER-Medicare Na-tional Claims History (NCH) and Outpatient files. Thecodes for screening mammogram included the HealthcareCommon Procedure Coding System (HCPCS) codes76092, G0202, and G0203, and the International Classifi-cation of Diseases (Ninth Revision) Clinical Modification(ICD-9-CM) code V76.12. The starting time for identifyingmammograms among the non-cancer controls was the yearand month of cancer diagnosis of the matched cancerpatients.

The following patient socio-demographic variables: age,race/ethnicity, enrollment in at state Medicaid managedprogram (called “state-buy-in”), and Census tract medianhousehold income from the 2000 U.S. Census wereobtained from SEER Patient Entitlement and DiagnosisSummary File. Each person’s Charlson co-morbidity score[22] was calculated from ICD-9 diagnosis codes in theSEER-Medicare NCH, Inpatient and Outpatient claimsduring the 2 years before the month of cancer diagnosis[23, 24]. The Charlson score was grouped as zero, 1, 2, and3 + . The diagnosis of colorectal cancer was excluded incalculating the above scores. Other variables derived fromthe Medicare claims data were whether or not a person wastreated for their cancer in a teaching hospital, as well as theannual number of visits to physicians for new or establishedevaluation & management (E&M) visits, and whether thewoman visited a gynecologists during the first and secondyears, and the third and fourth years after the cancerdiagnosis.

Mammography use was compared between cancerpatients and the matched women with no history of cancer

(controls). The differences of individual characteristicswere tested with t-test for continuous variables and Z-testfor rates. The unadjusted survival curves for the time to thefirst mammogram after the cancer diagnosis were comparedwith the control group during the same period, usingproduct-limit estimates. The time to screening mammogramwas censored at the time of death for those who died duringthe study period. The use of mammography was examinedusing multivariate linear models for rates, and conditionallogistic models for the odds ratios between cancer patientsand non-cancer controls taking account of the matcheddesign. All model results were adjusted for age, race/ethnicity, state buy-in, Charlson comorbidity score, zipcode median income, treatment in a teaching hospital,number of E&M physician office visits and visit to agynecologist [25]. Those who died within the first two, orfirst four, years after the cancer diagnosis (or the sameperiod for controls) were excluded in this part of analysis.Statistical analyses were performed using the StatisticalAnalysis Software, Version 9.1.4 for Windows, SASInstitute Inc., Cary, NC, 2007).

Informed consent from individual patients was notrequired for this study. IRB approval was granted by theUniversity of Minnesota Human Subjects Committee understudy numbers 0312 M55082 and 0312 M54865.

Results

Study populations

The average age of the women newly diagnosed withcolorectal cancer was 73.6 years (Table 1). 7.7% of caseswere Black, 19.1% participated in a state Medicaidadministered program, 32.0% lived in zip codes with amedian income below $38,000, and 31.5% had a Charlsonscore of zero. In addition, 45.9% of cancer patients hadAJCC stage 0 or I, 33.7% had stage II, and 20.4% had stageIII cancer. The matched controls were slightly more likelyto be younger, were less likely to live in areas with thelowest median income, and more likely to have a Charlsonscore of zero.

Pre-diagnosis screening mammography rates

The age adjusted mammography rate in 2 years before thecancer diagnosis was the same, 47.8/100, among cancerpatients and the control group (Table 2). The rate wassignificantly greater among those with stage 0 or I cancer,53.9/100, than among those with stage II or stage III disease,43.8/100 and 42.3/100 respectively (p for trend <0.001).There was no difference among the stage-specific controlgroups.

14 J Cancer Surviv (2009) 3:12–20

Post-diagnosis screening mammography rates

In the first 2 years following the diagnosis of colorectalcancer, the screening mammography rate among cancerpatients was 49.7/100, 4.2% greater than the rate amongthe non-cancer controls (47.6/100), adjusted O.R. = 1.05,95% C.I. = 1.01–1.09, (Table 3). This difference seenbetween the entire survivor cohort and all control was dueentirely to 22.1% higher rate among the cancer patientswho had not had a mammogram prior to the diagnosis ofcancer, 30.9/100, compared with their controls, 25.3/100,adjusted O.R. = 1.23, 95% C.I. = 1.15–1.32. Dependingon the stage of cancer, the rate among the women who hadnot had a recent prior mammogram was 16.6 to 27.3%greater than their controls (adjusted O.R. from 1.18 to1.27).

In the third and fourth year after the diagnosis ofcolorectal cancer, the rate of screening mammographyamong the cancer survivors was 54.5/100, 9.7% greater thanthe rate in the first 2 years. This rate was 7.5% greater thanthe rate among the controls (50.4/100), adjusted O.R. = 1.09,

95% C.I. = 1.05–1.13. The differences between the cancersurvivors and the control women were greatest among thewomen who had not had a mammogram prior to thediagnosis of cancer (18.5% for the total group, with a rangefrom 16.4% to 20.9% depending on the stage at diagnosis).All of these differences were statistically significant in themultivariate analysis.

Women who had mammograms prior to the cancerdiagnosis had rates of mammography in the two follow-upperiods that were approximately twice as great as the ratesamong the women with no history of pre-diagnosismammography. In years 1 and 2, the rates were 69.9/100and 30.9/100, respectively. In years 3 and 4, they were72.4/100 and 37.3/100, respectively.

The patterns of the time to the first mammogram use afterthe cancer diagnosis appear different for those who had apre-diagnosis mammogram, as well as those who did not,compared to their respective control group (Figs. 1 and 2).Because the curves of the time to first mammogram weresimilar for the stage-specific controls of the women who hada pre-diagnosis mammogram, they were combined in Fig. 1.

Table 2 Mammography rates (per 100) in 2 years before cancer diagnosis for female colorectal cancer patients and for non-cancer controls duringthe same period

Cancer patients Controls p value

AJCC stage N Rate p value for trend N Rate p value for trend cancer patients vs. controls

0/I 3,515 53.9 <0.001 16,761 48.3 0.74 <.0001II 2,585 43.8 12,224 48.0 <.0001III 1,566 42.3 7,448 47.7 <.0001Total 7,666 47.8 36,433 47.8 0.8957

Table 1 Characteristics of female colorectal cancer patients and matched non-cancer controls

Cancer patients Controlsn=7,666 n=36,433

Age Mean (SD) 73.6 (3.5) 73.7 (3.7)67–69 16.3% 19.0% p<0.00170–74 37.2% 35.6%75–79 46.5% 45.4%

Race/ethnicity White 85.4% 85.7%Black 7.7% 7.4%Other 6.8% 6.8%

State buy-in status Yes 19.1% 19.2%Zip code median income <$38,000 32.0% 28.4% p<0.001

$38,001–$50,000 37.0% 34.6%$50,001+ 31.0% 39.0%

Charlson score 0 31.5% 43.9% p<0.0011 33.7% 28.1%2 17.9% 14.5%3+ 16.9% 13.5%

AJCC stage 0/I 45.9%II 33.7%III 20.4%

J Cancer Surviv (20093:12–20 15

Likewise, because the curves of the time to first mammo-gram were similar for the stage-specific controls of thewomen who did not have a pre-diagnosis mammogram, theywere combined in Fig. 2. Because these are cumulativecurves, it is the first 24 months after the diagnosis of cancerthat is of primary interest. Among the women who had arecent prior mammogram (Fig. 1), the cumulative percent ofwomen with colorectal cancer who had a post-diagnosismammogram lagged behind that of the control women forthe entire first year. At 6 months, approximately 26% of the

control population had a mammogram compared with 21%of those with stage 0/I, 18% with stage II and 15% withstage III cancer. At 12 months, a similar percentage ofpatients with stage 0/I disease and the control group hadreceived mammograms, and the cumulative percents for thetwo groups are essentially the same for the remainder of the60-month period. Those with stage II and stage III cancerlagged behind the controls during almost the entire 24-monthperiod. However, by 24 months they achieved rates of 67.1/100 and 68.3/100, respectively, which were not statistically

Cancer Stage Non-cancer controls AJCC 0/IAJCC II AJCC III

Mammography Rate

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Months after the Cancer Diagnosis

0 6 12 18 24 30 36 42 48 54 60

Figure 1 Time to the first mam-mogram among women havingmammogram before the cancerdiagnosis.

Table 3 Mammography rates (per 100) and adjusted odds ratios for cancer patients after the cancer diagnosis and for non-cancer controls duringthe same period by history of mammogram

First and second years Third and fourth years

Cancer patientsN=7,666 Rate

ControlsN=36,433 Rate

Adjusted odds ratio(95% C.I.)

Cancer patientsN=7,159 Rate

ControlsN=33,197 Rate

Adjusted odds ratio(95% C.I.)

Total 49.7 47.6 1.05 (1.01–1.09) 54.5 50.4 1.09 (1.05–1.13)AJCC stage0/I 51.6 48.7 1.06 (1.01–1.12) 57.6 51.2 1.08 (1.03–1.14)II 47.5 47.0 1.02 (0.96–1.09) 51.2 49.4 1.09 (1.02–1.16)III 49.0 46.2 1.08 (0.99–1.17) 52.9 50.2 1.12 (1.03–1.22)Prior 69.9 71.7 0.98 (0.94–1.02) 72.4 71.5 1.01 (0.97–1.06)AJCC stage0/I 72.3 72.4 0.99 (0.93–1.06) 73.7 72.0 1.01 (0.95–1.08)II 67.1 71.4 0.95 (0.87–1.03) 70.8 70.7 1.02 (0.94–1.11)III 68.3 70.8 0.99 (0.89–1.11) 71.6 71.7 1.03 (0.92–1.15)No prior 30.9 25.3 1.23 (1.15–1.32) 37.3 30.4 1.25 (1.17–1.33)AJCC stage0/I 31.9 25.5 1.27 (1.15–1.41) 38.8 31.4 1.27 (1.15–1.40)II 29.5 25.3 1.18 (1.05–1.32) 35.3 29.5 1.21 (1.09–1.35)III 31.3 24.6 1.24 (1.07–1.44) 37.8 29.9 1.28 (1.11–1.48)

16 J Cancer Surviv (2009) 3:12–20

different from the controls (Table 3). Among the cancerpatients who did not have a mammogram prior to thediagnosis of cancer (Fig. 2), at 6 months, the percent ofwomen who had a mammogram was greatest in the thosewith stage 0/I disease (10%). It was 8% among those withstage II disease and the controls, and 6% among those withstage III disease. By 12 months the percent of women whohad received a mammogram was greater than (stage 0/I), orsimilar to (stage II and stage III), the rate among the controls.As described earlier, by 24 months, the percent of patients ineach of the stage groups who had had a mammogram wassignificantly greater than the percent among the controlwomen (Table 3).

Factors associated with post-diagnosis mammography use

In the multivariate analysis assessing possible factors thatcould be associated with the use of mammography afterthe diagnosis of cancer (Table 4), we found that thecovariate that had the greatest impact was prior screeningmammogram, adjusted O.R. = 5.76 (95% C.I. = 5.19–6.38) for the first and second years, and adjusted O.R. =4.99 (95% C.I. = 4.48–5.55) for the third and fourth years.Other covariates that were positively associated in each ofthe two time periods with the receipt of a mammogramwere visiting a gynecologist, receiving chemotherapy andhaving five or more E&M physician visits. Older women,those in a Medicaid administered program, those with aCharlson score ≥ 2, and those with higher stage cancer (IIand III) were less likely to have a post-diagnosismammogram in each of the two time periods.

Discussion

This study is the first to include recent prior mammographyinformation when examining the short and medium termimpact of newly diagnosed non-breast cancer on subse-quent mammography use. In brief, we found that those withcancer were more likely to have a screening mammogramin the 2 years after the diagnosis of cancer than matchedcontrol women. The difference in favor of these womenwas even greater in the third and fourth year after diagnosis.These findings are consistent with the work of our groupand others that showed higher rates of mammographyscreening among long-term (5 years or more) colorectal anduterine cancer survivors than among non-cancer compari-son groups [9–11]. Of particular interest is the finding thatthe higher rate among the cancer survivors was due to thehigher rate among those women without a history of arecent mammogram compared with their controls. Thepost-diagnosis rates of mammography among patients whohad had a recent prior mammogram were not different fromthe rates in their controls.

We hypothesized that the diagnosis of cancer wouldreduce the use of screening mammography due to thecompeting demands of cancer treatment and follow-up. Wedid not find this. For the entire cancer cohort, the rate ofscreening was 4.2% higher in the first 2 years afterdiagnosis and 7.5% higher in the third and fourth yearsthan among the control group. These patterns suggest thatthe competing demands of cancer care did not have asignificant negative impact on mammography seekingbehavior of the entire cohort of cancer patients. However,

Cancer Stage Non-cancer controls AJCC 0/IAJCC II AJCC III

Mammography Rate

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Months after the Cancer Diagnosis

0 6 12 18 24 30 36 42 48 54 60

Figure 2 Time to the first mam-mogram among women not hav-ing a mammogrambefore the cancer diagnosis.

J Cancer Surviv (20093:12–20 17

we did find signs that there was some impact. For example,the mammography screening rates were significantly loweramong those with stage II and III disease than among thosewith stage 0/I disease (Table 3). The logistic regressionmodels confirmed this pattern for both the initial 2 yearsand the subsequent 2 years (Table 4). Also, the lag in thetime to first mammogram among those with stage II and IIIdisease compared with those with Stage 0/I disease seen inFigs. 1 and 2 is a good representation of the possible impacton each stage-specific cohort of more involved treatment onthose with higher stage disease. For example, if chemo-therapy caused a delay in mammography use, only aminimal effect would be seen on the stage 0/I cohortbecause only 4% received chemotherapy (data not shown).However, 34% of the stage II patients and 76% of the stageIII patients received chemotherapy, and we did see in thefigures a delay in their mammography use.

Stratifying patients based on their recent prior use ofscreening mammography led to perhaps the most intriguingfinding in the study, the increased use of mammographyamong cancer patients with no history of mammography

compared with their matched controls: 22% greater after2 years and 23% greater in years 3 and 4. This was true forall stage-specific groups of women. The diagnosis ofcolorectal cancer appears to have stimulated the use ofscreening mammograms in this group. Put into theframework of the Health Belief Model [26], having arecent diagnosis of colorectal cancer may have increased awoman’s perceived threat of breast cancer. Our results showthat, whatever the prior perceived barriers to mammographymight have been, a significantly larger percentage of thesewomen took action to have a screening mammogram afterthe cancer diagnosis compared with their control group.

Due to the source and nature of our data, we cannotdetermine the reason for the change in behavior amongwomen with no history of a recent mammogram. It maycome from the patient’s perception of increased suscepti-bility to breast cancer, or may be due to a physician’srecommendation. Previous studies have shown that physi-cian recommendation is the key factor in patient’s use ofmammography [27]. In the multivariate analysis (Table 4) agreater number of physician visits (≥5), and a visit to a

Table 4 Determinants of mammography use during the first and second, or third and fourth years after colorectal cancer diagnosis

First and second years Third and fourth yearsAdjusted odds ratio (95% C.I.) Adjusted odds ratio (95% C.I.)

Age in years 67–69 Reference Reference70–74 0.79 (0.68–0.92) 0.76 (0.65–0.89)75–79 0.61 (0.52–0.71) 0.49 (0.42–0.57)

Race/ethnicity White Reference ReferenceBlack 0.80 (0.65–0.98) 1.10 (0.89–1.36)Other 0.88 (0.71–1.08) 0.97 (0.78–1.21)

State buy–in status No Reference ReferenceYes 0.60 (0.52–0.69) 0.53 (0.46–0.61)

Zip code median income <$38,000 Reference Reference$38,001–$55,000 0.97 (0.86–1.1) 0.97 (0.85–1.10)>$55,000 0.92 (0.80–1.05) 0.93 (0.81–1.07)

Charlson score 0 Reference Reference1 0.99 (0.87–1.12) 1.01 (0.89–1.15)2 0.83 (0.71–0.96) 0.83 (0.71–0.97)3+ 0.57 (0.49–0.67) 0.52 (0.44–0.61)

Hospital type Non-teaching Reference ReferenceTeaching 1.06 (0.96–1.18) 1.02 (0.91–1.13)

AJCC cancer stage 0/I Reference ReferenceII 0.68 (0.60–0.77) 0.67 (0.59–0.77)III 0.54 (0.46–0.65) 0.51 (0.43–0.62)

Chemotherapy No Reference ReferenceYes 1.61 (1.40–1.86) 1.93 (1.67–2.24)

Prior mammogram No Reference ReferenceYes 5.76 (5.19–6.38) 4.99 (4.48–5.55)

# of E&M physician visits 0–4 Reference Reference5–9 1.14 (0.99–1.30) 1.25 (1.1–1.43)10–15 1.50 (1.29–1.74) 1.73 (1.47–2.02)15+ 1.33 (1.13–1.57) 1.50 (1.22–1.86)

Gynecologist visit No Reference ReferenceYes 1.83 (1.55–2.16) 1.80 (1.49–2.18)

18 J Cancer Surviv (2009) 3:12–20

gynecologist were both associated with an increase use ofmammography even after adjusting for prior mammogra-phy use. Thus, physicians need to continue to remindwomen of the need for mammography, other screening testsand other preventive services even in the period after acancer diagnosis. Perhaps there is room, also, for studiesthat provide primary care physicians, and possibly gynecol-ogists, with materials or staff to increase the preventionmessage particularly among those with no history of recentmammography or with higher stage disease.

Our results appear to conclusively answer the questionraised by Snyder et al. [15] regarding the predictive powerof pre-diagnosis service use. The adjusted odds ratios forprior mammography use of 5.76 (95% CI = 5.19–6.38 forthe first and second years and 4.99 (95% CI = 4.48–5.55)for the third and fourth years are far greater than any othervariables used in our model.

An interesting, although not surprising, finding in thisstudy was the relationship between mammography useprior to the diagnosis of colorectal cancer and stage atdiagnosis. Prior mammography rates were 23.1% higheramong women with either stage 0 or I cancer comparedwith those with stage II disease, and 27.4% higher thanamong those with stage III disease (Table 2). Likelyexplanations for this are that women who received mammo-grams either were also self-motivated to seek screeningtests for colorectal cancer, or were guided by their healthcare providers to seek screening or receive earlier diagnos-tic testing for gastrointestinal symptoms. Carlos et al.reported that 50 + year –old women who reported thatthey adhered to mammography and Pap smear guidelineswere more likely to adhere to colorectal screening guide-lines [28], and Klabunde et al. [29] have shown that amongelderly Americans, being current on one of three preventiveservices (influenza immunization, pneumococcal polysac-charide vaccination, and colorectal cancer screening) waspredictive of being current with another.

A major strength of our study is the use of the carefullyvalidated SEER-Medicare linked data. Thus, our finings arelikely to be accurate and can be directly generalized to theU.S. elderly population. Further, we matched each cancercase with up to five non-cancer controls by race, age, andSEER site, and aligned the starting time to mammographyuse in the matched controls with that of the cases. Thisensures accurate estimates and valid comparisons ofmammography use between them. Further, we specificallystratified our study cohorts based on the mammography usebefore the cancer diagnosis. No previous study had adoptedthis strategy to examine mammography among non-breastcancer patient or survivors [15].

This study has limitations. The claims data only provide theevidence of mammography use. We cannot know whethermammography use was due to the patient’s request or the

physician’s recommendation. We also did not have measuresof health beliefs, thus unable to determine actual changes inhealth beliefs. In addition, claims data do not provide detailedclinical information that might influence the use of mammog-raphy, although we had the accurate stage at diagnosisinformation from the SEER registries and used the Charlsonscore to assess patient comorbidities. We also did not examinein detail how the chemotherapy was carried out among cancerpatients, except whether or not the patients received thechemotherapy according to the common practice and clinicalguidelines. However, the narrow range in the time to the onsetof chemotherapy and its duration reported in the Methodssection indicates little variation in its application. Finally,higher rates of mammography have been reported by theCDC’s BRFSS program: 79% of all U.S. elderly women in2006 self-reported the receipt of a mammogram in the past2 years [5]. However, the BRFSS does not distinguishbetween screening and diagnostic examinations, and thescreening mammography rates we found are consistent withthose from recent studies using Medicare data [9–11, 14, 15,30, 31].

In conclusion, following the diagnosis of colorectalcancer, cancer survivors are more likely than controlwomen to have a screening mammogram in the first andsecond, or third and fourth years after the diagnosis (thisstudy) and as long-term (more than 5 years) survivors [9–11]. While a diagnosis of colorectal cancer can bedevastating, it also appears to be an opportunity forproviders to encourage use of other preventive measures.

Acknowledgement This study was supported by grants from theNational Cancer Institute (R01 CA98974) and the National Instituteon Aging (R01 AG25079).

References

1. Ries L, Melbert D, Krapcho M, Stinchcomb DG, Howlader N,Horner MJ, et al. SEER Cancer Statistics Review, 1975–2004,National Cancer Institute. Bethesda, MD. Available from http://seer.cancer.gov/csr/1975_2004/

2. National Center for Health Statistics.Health, United States, 2007.With Chartbook on Trends in the Health of Americans. Hyatts-ville, MD; 2007.

3. Weinberg DS, Newschaffer CJ, TophamA. Risk for colorectal cancerafter gynecologic cancer. Ann Intern Med. 1999;131:189–93.

4. Ahmed F, Goodman MT, Kosary C, Ruiz B, Wu XC, Chen VW,et al. Excess risk of subsequent primary cancers among colorectalcarcinoma survivors, 1975-2001. Cancer. 2006;107(5,Suppl):1162–71. doi:10.1002/cncr.22013.

5. CDC.Behavioral Risk Factor Surveillance System. Ternds andPrevelance Data. Available from http://apps.nccd.cdc.gov/BRFSS/age.asp?cat=WH&yr=2006&qkey=4427&state=UB

6. Rao VM, Levin DC, Parker L, Frangos AJ. Recent trends inmammography utilization in the Medicare population: is there acause for concern? J Am Coll Radiol. 2008;5:652–6. doi:10.1016/j.jacr.2008.01.023.

J Cancer Surviv (20093:12–20 19

7. Kagay CR, Quale C, Smith-Bindman R. Screening mammographyin the American elderly. Am J Prev Med. 2006;31:142–9.doi:10.1016/j.amepre.2006.03.029.

8. Rakowski W, Meissner H, Vernon SW, Breen N, Rimer B,Clark MA. orrelates of repeat and recent mammography forwomen ages 45 to 75 in the 2002 to 2003 Health InformationNational Trends Survey (HINTS 2003. Cancer EpidemiolBiomark Prev. 2006;15:2093–101. doi:10.1158/1055-9965.EPI-06-0301.

9. Earle CC, Neville BA. Under use of necessary care among cancersurvivors. Cancer. 2004;101:1712–9. doi:10.1002/cncr.20560.

10. Yu X, McBean AM, Virnig BA. Physician visits, patientcomorbidities, and mammography use among elderly colorectalcancer survivors. Journal of Cancer Survivorship. 2007;1:275–82.doi:10.1007/s11764-007-0037-7.

11. McBean AM, Yu X, Virnig BA. The use of preventive healthservices among elderly uterine cancer survivors. Am J ObstetGynecol. 2008;198:86 e81–8.

12. Jaen CR, Stange KC, Nutting PA. Competing demands of primarycare: a model for the delivery of clinical preventive services. JFam Pract. 1994;38:166–71.

13. Mandelblatt JS, Lawrence WF, Cullen J, Stanton AL, KrupnickJL, Kwan L, et al. Patterns of care in early-stage breast cancersurvivors in the first year after cessation of active treatment. J ClinOncol. 2006;24:77–84. doi:10.1200/JCO.2005.02.2681.

14. Snyder CF, Earle CC, Herbert RJ, Neville BA, Blackford AL,Frick KD. Trends in follow-up and preventive care for colorectalcancer survivors. J Gen Intern Med. 2008;23:254–9. doi:10.1007/s11606-007-0497-5. Epub 2008 Jan 16.

15. Snyder CF, Earle CC, Herbert RJ, Neville BA, Blackford AL,Frick KD. Preventive care for colorectal cancer survivors: a 5-yearlongitudinal study. J Clin Oncol. 2008;26:1073–9. doi:10.1200/JCO.2007.11.9859.

16. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF.Overview of the SEER-Medicare data: content, research applica-tions, and generalizability to the United States elderly population.Med Care. 2002;40(8, Suppl):IV-3–IV-18.

17. Warren JL, Harlan LC, Fahey A, Virnig BA, Freeman JL,Klabunde CN, et al. Utility of the SEER-Medicare data to identifychemotherapy use. Med Care. 2002;40(8, Suppl):IV-55–IV-61.

18. Schrag D, Rifas-Shiman S, Saltz L, Bach PB, Begg CB. Adjuvantchemotherapy use for Medicare beneficiaries with stage II coloncancer. J Clin Oncol. 2002;20:3999–4005. doi:10.1200/JCO.2002.11.084.

19. Hershman D, Hall MJ, Wang X, Jacobson JS, Mcbride R, GrannVR, et al. Timing of adjuvant chemotherapy initiation after

surgery for stage III colon cancer. Cancer. 2006;107:2581–8.doi:10.1002/cncr.22316.

20. Neugut AI, Matasar M, Wang X, McBride R, Jacobson JS, TsaiWY, et al. Duration of adjuvant chemotherapy for colon cancerand survival among the elderly. J Clin Oncol. 2006;24(15):2368–75. doi:10.1200/JCO.2005.04.5005.

21. Benson AB 3rd, Schrag D, Somerfield MR, Cohen AM,Figueredo AT, Flynn PJ, et al. American Society of ClinicalOncology recommendations on adjuvant chemotherapy for stageII colon cancer. J Clin Oncol. 2004;22:3408–19. doi:10.1200/JCO.2004.05.063.

22. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new methodof classifying prognostic comorbidity in longitudinal studies:development and validation. J Chronic Dis. 1987;40:373–83.doi:10.1016/0021-9681(87) 90171-8.

23. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidityindex for use with ICD-9-CM administrative databases. J ClinEpidemiol. 1992;45:613–9. doi:10.1016/0895-4356(92) 90133-8.

24. Klabunde CN, Harlan LC, Warren JL. Data sources for measuringcomorbidity: a comparison of hospital records and medicareclaims for cancer patients. Med Care. 2006;44:921–8.doi:10.1097/01.mlr.0000223480.52713.b9.

25. Ghali WA, Quan H, Brant R, Melle GV, Norris CM, Faris PD, et al.Comparison of 2 methods for calculating adjusted survival curvesfrom proportional hazards models. Journal of the American MedicalAssociation. 2001;286:1494–7. doi:10.1001/jama.286.12.1494.

26. Glanz K, Rimer BK, Lewis FM. Health Behavior and HealthEducation: Theory, Research, and Practice (3 rd ed). Jossey-Bass;2002.

27. Walter LC, Lewis CL, Barton MB. Screening for colorectal, breast,and cervical cancer in the elderly: a review of the evidence. Am JMed. 2005;118:1078–86. doi:10.1016/j.amjmed.2005.01.063.

28. Carlos RC, Fendrick AM, Patterson SK, Bernstein SJ. Associa-tions in breast and colon cancer screening behavior in women.Acad Radiol. 2005;12:451–8. doi:10.1016/j.acra.2004.12.024.

29. Klabunde CN, Meissner HI, Wooten KG, Breen N, Singleton JA.Comparing colorectal cancer screening and immunization status inolder americans. Am J Prev Med. 2007;33:1–8. doi:10.1016/j.amepre.2007.02.043.

30. Earle CC, Burstein HJ, Winer EP, Weeks JC. Quality of non-breast cancer health maintenance among elderly breast cancersurvivors. J Clin Oncol. 2003;21:1447–51. doi:10.1200/JCO.2003.03.060.

31. McBean AM, Yu X. The underuse of screening services amongelderly women with diabetes. Diabetes Care. 2007;30:1466–72.doi:10.2337/dc06-2233.

20 J Cancer Surviv (2009) 3:12–20