Contraindicated Use of Bevacizumab and Toxicity in Elderly Patients With Cancer

Abstract

Purpose

Drugs are approved on the basis of randomized trials conducted in selected populations. However, once approved, these treatments are usually expanded to patients ineligible for the trial.

Patients and Methods

We used the SEER-Medicare database to identify subjects older than 65 years with metastatic breast, lung, and colon cancer, diagnosed between 2004 and 2007 and undergoing follow-up to 2009, who received bevacizumab. We defined a contraindication as having at least two billing claims before bevacizumab for thrombosis, cardiac disease, stroke, hemorrhage, hemoptysis, or GI perforation. We defined toxicity as first development of one of these conditions after therapy.

Results

Among 16,085 metastatic patients identified, 3,039 (18.9%) received bevacizumab. Receipt of bevacizumab was associated with white race, later year of diagnosis, tumor type, and decreased comorbid conditions. Of patients who received bevacizumab, 1,082 (35.5%) had a contraindication. In multivariate analysis, receipt of bevacizumab with a contraindication was associated with black race (odds ratio [OR] = 2.6; 95% CI, 1.4 to 4.9), increased age, comorbidity, later year of diagnosis, and lower socioeconomic status. Patients with lung (OR = 1.7; 95% CI, 1.1 to 2.4) and colon cancer (OR = 1.4; 95% CI, 1.1 to 1.9) were more likely to have a contraindication. In the group with no contraindication, 30% had a complication after bevacizumab; black patients were more likely to have a complication than were white patients (OR = 1.9; 95% CI, 1.21 to 2.93).

Conclusion

Our study demonstrates widespread use of bevacizumab among patients who had contraindications. Black patients were less likely to receive the drug, but those who did were more likely to have a contraindication. Efforts to understand toxicity and efficacy in populations excluded from clinical trials are needed.

INTRODUCTION

Bevacizumab is a humanized monoclonal antibody against vascular endothelial growth factor (VEGF) that is US Food and Drug Administration (FDA) approved for first-line therapy against a number of common solid tumors.¹ It was first approved for metastatic colorectal cancer in 2004 on the basis of two large randomized trials that showed improvement in overall survival when bevacizumab was added to standard chemotherapy.^2,3 Subsequently, it was also approved for the treatment of nonsquamous non–small-cell lung cancer in 2006 after a study showed improved survival, but increased treatment-related deaths,⁴ and for breast cancer in 2008 after a study showed an increase in progression-free, but not overall, survival.⁵ The FDA reversed their approval of bevacizumab for breast cancer in 2011 when additional clinical trials failed to show a benefit.

Bevacizumab therapy is associated with an increased risk of a number of toxicities, which has resulted in a black box warning for GI perforation, wound healing complications, and hemorrhage.⁶ Because of these and other known serious complications of bevacizumab therapy, patients with a history of bleeding or thrombotic disorders, hemoptysis, cerebral vascular accident, significant cardiac disease (ischemic or congestive heart failure), or GI perforation were excluded from participation in the clinical trials that led to drug approval. In addition, because of the high frequency of increased risk of hypertension, patients were eligible for these trials only if they had controlled hypertension and had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1.^2–5 The drug monograph includes a black box warning as well as a list of contraindications and conditions in which the drug should be administered with caution, and the package label lists these under warnings (Appendix Table A1, online only).⁷

The generalizability of clinical trial results depends on how representative the trial population is of the cancer population. Once a drug is approved by the FDA, its use is typically broadened to patients who would not have met the strict eligibility criteria for the drug in the clinical trial, yet the risks and benefits in those populations are not known.⁸ We performed a population-based analysis to determine the predictors of bevacizumab use in elderly patients with metastatic colon, lung, and breast cancer; the predictors of use in patients with contraindications to the drug; and toxicities associated with use.

PATIENTS AND METHODS

Data Source

We analyzed data from the SEER-Medicare database.⁹ SEER provides information on tumor histology, location, stage of disease, treatment, and survival, along with demographic and selected census tract-level information. The Medicare database includes Medicare A (inpatient) and B (outpatient) eligibility status, billed claims, and diagnoses. These two files are linked and provide the ability to determine treatment, dates of service, and tumor registry information. Exemption was obtained from the Columbia University Institutional Review Board.

Cohort Selection

We identified all individuals age 65 years or older who had a pathologically confirmed primary diagnosis of stage IV or recurrent breast cancer, non–small-cell lung cancer, or colon cancer¹⁰ between January 1, 2004, and December 31, 2007. If chemotherapy was administered after the first 12 months in patients with stage I-III disease, the patient was categorized as having a treated recurrence. We excluded patients who were enrolled in a non-Medicare health maintenance organization.¹¹ Patients who were enrolled in Medicare because of end-stage renal disease and dialysis, as well as patients with other primary cancers, were also excluded. We classified patients as having received bevacizumab if there was a claim within 12 months of recurrent/metastatic diagnosis. Patients who died within 12 months of diagnosis were excluded, as were patients who had intraocular administration of bevacizumab. Age at diagnosis was categorized into 10-year intervals. We recoded the SEER marital status variable as married, not married, and unknown, and the SEER race variable was categorized as black, Hispanic, white, and other.

Socioeconomic Status Score

We generated an aggregate socioeconomic status (SES) score from education, poverty level, and income data from the 2000 census tract data, as described previously by Du et al.¹² Patients' scores were ranked on a scale of 1 to 5 by use of the formula that incorporated education, poverty, and income weighted equally, with 1 being the lowest value.

Assessment of Comorbid Disease

To assess the prevalence of comorbid disease in our cohort, we used the Klabunde adaptation of the Charlson comorbidity index.^13,14 Medicare inpatient and outpatient claims were searched for 1 year before diagnosis for diagnostic codes of the International Classification of Disease, Ninth Revision, Clinical Modification.¹⁰ Each condition was weighted, and patients were assigned a score that was based on the Klabunde-Charlson index.¹⁴

Treatment Characteristics

We extracted information on chemotherapy from the date of diagnosis from the Medicare files by searching the Level II Healthcare Common Procedure Coding System, Current Procedural Terminology codes, International Classification of Disease, Ninth Revision, Clinical Modification diagnostic codes and procedure codes, diagnostic-related group code, and the center code from physician claims files, the hospital outpatient claims files, or the Medicare provider review files. We searched for Level II Healthcare Common Procedure Coding System codes corresponding to bevacizumab (J9035). All participants had at least one claim for bevacizumab.

Toxicities

Patients were classified as having a contraindication if they had at least two claims within 12 months before first bevacizumab for any of the following diagnoses: severe cardiac disease, thrombosis, hemorrhage, stroke, hemoptysis, or colon perforation. Diagnosis date was used for patients who did not receive bevacizumab. Patients were classified as having a caution if they had hypertension, thrombosis of superficial veins, or prior receipt of doxorubicin (Appendix Table A2, online only). Because we could not differentiate patients who had well-controlled hypertension from those who did not, we considered this a caution. If a patient had both a contraindication and a caution, they were classified as a contraindication. Prior history was defined as two claims within 12 months before receipt of bevacizumab. Subsequent toxicity was defined as two claims for any of the codes for contraindication after the date of first administration of bevacizumab through 3 months after the date of last bevacizumab infusion. Only patients without a prior contraindication were included in the toxicity analysis.¹⁵

Statistical Analysis

Descriptive associations were evaluated using χ² analysis. The associations between outcomes of interest and bevacizumab use were assessed using multivariable logistic regression models that included patient and tumor characteristics. In the subset of patients who received bevacizumab, factors associated with the presence of a contraindication before bevacizumab use were assessed with a multivariable regression. The association of bevacizumab with subsequent toxicity was performed in the subset of patients who received bevacizumab who did not have any contraindication claims before receipt of bevacizumab because the exact timing of any subsequent event could not be determined from the claim. A time-dependent Cox proportional regression analysis was performed evaluating the difference in time to the development of toxicity in patients who did and did not receive bevacizumab. All analyses were conducted with SAS, version 9.13 (SAS Institute, Cary, NC). All statistical tests were two sided.

RESULTS

Among 16,085 patients who met our inclusion criteria, 3,039 (18.9%) received bevacizumab within 12 months of recurrent or metastatic disease. In a univariate analysis, receipt of bevacizumab was associated with age, white race, later year of diagnosis, being married, tumor type, tumor grade, geographic region, and decreased comorbid conditions (Table 1). In a multivariate analysis, black race (odds ratio [OR] = 0.81; 95% CI, 0.67 to 0.98) and Hispanic ethnicity (OR = 0.65; 95% CI, 0.41 to 0.96) were associated with decreased receipt of bevacizumab. Bevacizumab use increased over time and decreased with increased comorbid conditions and contraindications. Patients with colon cancer (OR = 5.34; 95% CI, 4.85 to 5.99) and patients with lung cancer (OR = 1.16; 95% CI, 1.02 to 1.32) were more likely to receive bevacizumab compared with patients with breast cancer. In all three cancers, use increased over time. Patients living in the West (OR = 1.68; 95% CI, 1.49 to 1.88) and Midwest (OR = 1.21; 95% CI, 1.06 to 1.37) were more likely to receive bevacizumab than were those living in the East.

Table 1.

Analysis of Factors Associated With Bevacizumab Use in Patients With Metastatic or Recurrent Colon, Lung, or Breast Cancer

Patient Demographic or Clinical Characteristic	Patients					Adjusted Analysis Associated With Bevacizumab
	No Bevacizumab		Bevacizumab		P*
	No.	%	No.	%		OR	95% CI	P
All patients	13,046	81.1	3,039	18.9
Age at first bevacizumab, years
65-74	7,400	82.1	1,641	17.9	.003	—	—	—
75-84	4,849	80.1	1,229	19.9		1.06	0.97 to 1.16	.16
≥ 85	797	82.6	169	17.4		0.85	0.71 to 1.04	.10
Race/ethnicity
White	11,218	81.0	2,675	19.0	.007	—	—	—
Black	991	84.9	180	15.1		0.81	0.67 to 0.97	.03
Hispanic	157	85.1	29	14.9		0.65	0.41 to 0.96	.04
Missing or other	693	81.5	155	18.5		0.81	0.67 to 0.99	.03
Year of diagnosis
2004	2,752	91.2	267	8.8	< .0001	—	—	—
2005	2,344	82.3	504	17.7		2.54	2.16 to 3.00	< .0001
2006	3,192	79.6	817	20.4		3.14	2.69 to 3.66	< .0001
2007	2,892	76.8	871	23.1		3.91	3.35 to 4.56	< .0001
2008	1,866	76.3	580	23.7		4.03	3.41 to 4.76	< .0001
Area of residence
Metropolitan	11,878	81.3	2,796	18.7	.56	—	—	—
Nonmetropolitan	1,168	82.0	262	18.0		0.94	0.80 to 1.11	.48
Marital status
Married	7,046	80.1	1,795	19.9	< .0001	—	—	—
Unmarried	5,618	83.1	1,154	16.9		0.97	0.88 to 1.06	.24
Unknown	382	81.4	90	18.6		1.03	0.80 to 1.33	.81
Socioeconomic status
Lowest (first) quartile	1,543	82.7	327	17.3	.08	—	—	—
Second quintile	2,387	82.7	514	17.3		0.97	0.82 to 1.14	.75
Third quintile	2,901	81.2	689	18.8		1.04	0.88 to 1.22	.65
Fourth quintile	2,945	81.2	697	18.8		1.06	0.90 to 1.25	.53
Highest (fifth) quartile	3,255	80.3	810	19.7		1.09	0.92 to 1.28	.33
Comorbidity score
0	6,574	80.5	1,596	19.5	.02	—	—	—
1	3,933	82.0	865	18.0		0.90	0.81 to 0.99	.04
> 1	2,801	82.5	596	17.5		0.75	0.66 to 0.84	< .0001
Tumor site
Breast	5,354	88.9	667	11.1	< .0001	—	—	—
Colon	3,085	64.0	1,800	36.0		5.39	4.85 to 5.99	< .0001
Lung	4,607	89.2	572	10.8		1.16	1.03 to 1.34	.03
Region
East	3,257	84.7	591	15.4	< .0001	—	—	—
Midwest	4,785	82.4	1,022	17.6		1.21	1.06 to 1.37	.003
West	5,345	78.7	1,444	21.3		1.68	1.49 to 1.88	< .0001
Tumor grade
High	5,981	77.0	1,836	23.0	< .0001	—	—	—
Low	4,059	84.7	741	15.3		0.87	0.78 to 0.96	.002
Unknown	3,006	86.8	462	13.2		0.94	0.86 to 1.11	.31
Contraindication
No	8,994	82.1	1,957	17.9	< .0001	—	—	—
Yes	4,052	78.9	1,082	21.1		1.23	1.08 to 1.39	.001
Caution
No	7,030	80.2	1,735	19.8	.001	—	—	—
Yes	6,016	82.2	1,304	17.8		1.01	0.90 to 1.13	.87

NOTE. All variables in the multivariate analysis controlled for each other.

Abbreviation: OR, odds ratio.

^*χ² statistic.

Of the patients who did not receive bevacizumab, 4,389 (33.9%) had a contraindication before diagnosis. Of the patients who received bevacizumab, 1,082 (35.5%) had a contraindication before its receipt, with 19% having cardiac disease, 15.7% hemorrhage, and 13% thrombosis (Fig 1). An additional 1,315 (43.0%) had a history of a caution. A history of hemorrhage, hemoptysis, or thrombosis was higher in patients with lung cancer than in patients with breast or colon cancer (Fig 1). In a multivariate analysis, receipt of bevacizumab with a prior contraindication was associated with black race (OR = 2.6; 95% CI, 1.4 to 4.9), increased age, increased comorbid conditions, and earlier year of diagnosis (Table 2). Compared with patients with breast cancer, patients with lung cancer (OR = 1.7; 95% CI, 1.1 to 2.5) and patients with colon cancer (OR = 1.4; 95% CI, 1.1 to 2.0) were more likely to have a contraindication. In addition, patients in the highest SES quintile were less likely to receive bevacizumab with a prior contraindication (OR = 0.64; 95% CI, 0.42 to 0.98). We performed a sensitivity analysis requiring that the two claims for a given contraindication to be on separate dates within the year before treatment, and the results were unchanged.

Fig 1.

Percentage of patients with contraindications before receipt of bevacizumab by tumor type.

Table 2.

Analysis of Factors Associated With Bevacizumab Use in Patients With Metastatic or Recurrent Colon, Lung, or Breast Cancer With and Without a Prior Contraindication

Patient Demographics and Clinical Characteristics	Patients							Adjusted Analysis Associated With Prior Contraindication
	No Contraindication		Caution		Contraindication		P*
	No.	%	No.	%	No.	%		OR	95% CI	P
All patients	653	21.5	1,304	43.0	1,082	35.5
Age at first bevacizumab, years
65-74	380	23.5	692	42.8	569	34.8	.02	—	—	—
75-84	245	20.1	538	43.1	410	36.5		1.20	0.95 to 1.51	.12
≥ 85	28	16.5	74	44.1	67	39.4		1.70	1.00 to 2.89	.05
Race/ethnicity
White	604	22.6	1,128	42.3	943	35.1	< .001	—	—	—
Black	15	8.3	87	48.3	78	43.3		2.58	1.37 to 4.84	.003
Hispanic†	< 11		< 11		15	51.7		1.68	0.47 to 5.96	.42
Year of diagnosis
2004	56	20.7	94	35.8	118	43.5	.001	—	—	—
2005	103	20.6	197	39.1	203	40.3		0.94	0.61 to 1.45	.77
2006	190	23.1	370	45.4	257	31.5		0.56	0.37 to 0.85	.006
2007	184	21.3	368	42.3	319	36.4		0.70	0.45 to 1.03	.09
2008	120	20.8	275	47.6	185	31.6		0.64	0.39 to 0.98	.06
Area of residence
Metropolitan	590	21.3	1,196	43.2	991	35.5	.56	—	—	—
Nonmetropolitan	63	24.1	108	41.4	91	34.5		0.69	0.45 to 1.04	.08
Marital status
Married	403	22.4	763	42.7	629	35.0		—	—	—
Unmarried	228	20.0	508	44.0	418	36.1	.49	1.04	0.81 to 1.30	.74
Unknown	22	24.4	33	37.8	35	37.8		1.02	0.52 to 1.81	.95
Socioeconomic status
Lowest quartile	59	18.0	130	40.0	138	42.1	.03	—	—	—
Second quintile	120	23.6	224	43.4	170	32.9		0.57	0.37 to 0.87	.01
Third quintile	140	20.4	280	40.7	269	38.9		0.83	0.54 to 1.25	.36
Fourth quintile	146	21.1	311	44.7	238	34.1		0.78	0.50 to 1.18	.26
Highest (fifth) quartile	188	23.0	358	44.6	264	32.4		0.63	0.42 to 0.98	.03
Comorbidity score
0	478	30.1	723	45.9	381	24.1	< .001	—	—	—
1	143	16.8	397	45.8	323	37.3		2.87	2.21 to 3.64	< .0001
> 1	32	5.5	184	31.4	378	63.1		14.92	9.80 to 21.4	< .0001
Tumor site
Breast	151	22.9	341	50.7	175	53.6	< .001	—	—	—
Colon	389	21.5	735	41.1	676	63.5		1.43	1.05 to 1.95	.03
Lung	113	19.9	228	40.0	231	66.9		1.70	1.15 to 2.51	.01
Region
East	95	16.2	256	44.0	236	39.8	< .001	—	—	—
Midwest	197	19.4	452	44.2	373	36.4		0.74	0.53 to 1.05	.08
West	361	25.2	596	41.8	473	33.0		0.63	0.46 to 0.86	.002
Tumor grade
High	387	21.0	824	44.7	632	34.3	.13	—	—	—
Low	165	22.1	313	41.9	270	36.1		0.94	0.72 to 1.22	.62
Unknown	106	22.7	178	38.2	182	39.1		0.87	0.63 to 1.25	.43

NOTE. All variables controlled for each other in the multivariable analysis.

Abbreviations: OR, odds ratio.

^*Contraindication compared with no contraindication.

^†Details on sample sizes < 11 cannot be reported per SEER-Medicare policy.

In the group of patients who received bevacizumab with no prebevacizumab contraindication, 30.3% developed a complication after receiving bevacizumab. The percentages of patients with complications after receipt of bevacizumab were 22.0%, 34.6%, and 27.7% in patients with breast, colon, and lung cancer, respectively. The most common toxicities were thrombotic (12.1%), followed by cardiac (10.6%), and hemorrhagic (11.9%). Cardiac and hemorrhagic complications were more common in patients with lung cancer, and thrombotic complications were more common in patients with colon cancer (Fig 2). In a multivariate analysis, compared with whites, blacks (OR = 1.8; 95% CI, 1.2 to 2.9) had an increased risk of postbevacizumab toxicity. Increased toxicities were also seen in patients with colon cancer (OR = 1.7; 95% CI, 1.2 to 2.1) compared with patients with breast cancer, and in patients with more comorbidities (Table 3). In a time-dependent Cox proportional hazards model evaluating time to toxicity in the bevacizumab patients without a contraindication before treatment and nonbevacizumab patients without a contraindication before diagnosis, patients who received bevacizumab were more likely to have a toxicity compared with those who did not (hazard ratio [HR] = 1.4; 95% CI, 1.2 to 1.6).

Fig 2.

Percentage of patients with toxicity after receipt of bevacizumab by tumor type.

Table 3.

Analysis of Factors Associated With Bevacizumab Toxicity in Patients With Metastatic or Recurrent Colon, Lung, or Breast Cancer

Patient Demographic or Clinical Characteristic	Patients					Adjusted Analysis Associated With Toxicity
	Toxicity		No Toxicity		P
	No.	%	No.	%		OR	95% CI	P
All patients	575	30.3	1,322	69.7
Age at first bevacizumab, years
65-74	323	30.9	716	69.1	.57	—	—	—
75-84	227	29.7	532	70.2		0.96	0.78 to 1.19	.72
≥ 85	25	26.0	74	74.0		0.90	0.55 to 1.48	.67
Race/Ethnicity
White	490	29.2	1,186	70.8	.012	—	—	—
Black	45	44.1	57	55.9		1.84	1.18 to 2.89	.007
Hispanic*	< 11		< 11			1.14	0.32 to 4.00	.83
Year of diagnosis
2004	41	29.3	103	70.7	< .001	—	—	—
2005	111	37.3	182	62.7		1.76	1.13 to 2.74	.01
2006	180	32.9	366	67.1		1.53	1.00 to 2.32	.05
2007	158	29.3	376	70.7		1.30	0.85 to 1.99	.22
2008	85	22.2	295	77.8		0.93	0.58 to 1.47	.75
Area of residence
Metropolitan	534	30.7	1207	69.3	.12	—	—	—
Nonmetropolitan	41	24.8	124	75.2		0.76	0.50 to 1.16	.21
Marital status
Married	370	32.4	763	67.6	.02	—	—	—
Unmarried	188	26.4	524	73.6		0.76	0.60 to 0.94	.01
Unknown	17	34.0	35	66.0		1.09	0.59 to 2.01	.77
Socioeconomic status
Lowest (first) quartile	54	29.6	131	70.4	.18	—	—	—
Second quintile	84	25.0	249	75.0		0.93	0.60 to 1.43	.74
Third quintile	123	30.1	284	69.9		1.21	0.80 to 1.84	.35
Fourth quintile	146	33.0	294	67.0		1.40	0.92 to 2.13	.12
Highest (fifth) quartile	168	31.3	364	68.7		1.27	0.83 to 1.96	.27
Comorbidity score
0	312	26.7	851	73.3	< .001	—	—	—
1	177	33.4	348	66.6		1.45	1.15 to 1.83	.002
> 1	86	41.0	123	59.0		1.98	1.44 to 2.73	< .001
Tumor site
Breast	104	22.0	375	78.0	< .001	—	—	—
Colon	379	34.6	709	65.4		1.66	1.26 to 2.18	< .001
Lung	92	27.7	238	72.3		1.07	0.77 to 1.59	.60
Region
East	118	34.2	223	65.8	.08	—	—	—
Midwest	196	31.2	437	68.8		0.99	0.73 to 1.35	.96
West	261	27.9	662	72.1		0.87	0.66 to 1.16	.34
Tumor grade
High	361	30.9	805	69.1	.51	—	—	—
Low	140	29.9	320	70.1		1.01	0.79 to 1.29	.91
Unknown	74	27.4	197	72.6		0.87	0.62 to 1.22	< .0001
Hypertension
No	92	26.1	260	73.9	.06	—	—	—
Yes	483	31.3	1,062	68.4		1.29	0.98 to 1.67	.07

NOTE. All variables controlled for each other in the multivariable analysis. Patients with contraindication before bevacizumab excluded from analysis.

Abbreviations: OR, odds ratio.

^*Details on sample sizes < 11 can not be reported per SEER-Medicare policy.

DISCUSSION

Our findings suggest that bevacizumab use in the community is substantially different from what has shown efficacy in clinical trials. We noted that not only was bevacizumab frequently used in patient groups that were not studied in clinical trials, but the drug was also used commonly in patients with contraindications to the drug. In addition, we found that complications resulting from bevacizumab are significantly higher than have been reported in the randomized trials including only these selected patient populations. Our findings raise concern that, as new treatments are disseminated to populations where the risk-benefit ratio has not been evaluated, higher-than-expected rates of complications will occur, raising issues regarding the generalizability of clinical trials.

Randomized clinical trials are designed to achieve maximal internal validity at the expense of generalizability to increase the likelihood of regulatory approval.¹⁶ Assessing the value of a therapeutic intervention requires reliable evidence of efficacy in patients representative of the general population. However, patients enrolled in clinical trials are often younger, with better performance status and fewer comorbid conditions.^8,17,18 To improve the diversity and generalizability of clinical trial participants, Congress enacted the National Institutes of Health Revitalization Act.¹⁹ Although enrollment of elderly patients has increased, racial and ethnic minorities continue to be under-represented.^20,21 Because expensive cancer medications that have substantial risk of toxicities are frequently used in patients for whom the benefits are uncertain, modifications to clinical trial designs to improve generalizability, such as pragmatic trials, adaptive trial designs, and coverage-with-evidence programs, have been proposed.⁸ More generally, observational studies are used after a drug has been marketed to assess the rates of adverse effects with use of the drug in the general community. The inefficiency of this approach has been criticized, and the FDA has made efforts to try to systematize the method for discovering the rates of long-term toxicities for drugs.

Although a number of studies have examined off-label use of drugs, there has been little data to explore use of medications in patients with contraindications.^22–24 Our findings are notable in that more than one third of patients who received bevacizumab had a major contraindication to the drug. Similarly, in a study investigating the use of the contraindicated agents enoxaparin and eptifibatide among patients undergoing dialysis and percutaneous coronary intervention, investigators found that 22.3% received a contraindicated antithrombotic medication. They also found that receipt of these medications was significantly associated with an increased risk of major bleeding while in hospital.²⁵

Interestingly, although black patients were less likely to receive bevacizumab, when they did receive it, they were more likely to have a prior contraindication. In addition, among patients without a prior contraindication, black patients were more likely to develop a subsequent toxicity. It is possible that this may result from disparities in quality of cancer care. Prior research has shown that the quality of the hospital, as measured by the number of patients who receive guideline care, and the hospital volume explain some of the reported racial disparities in the receipt of definitive breast cancer therapy.^26,27 In addition, increased risk of toxicity may be a function of a higher risk of stroke, cardiovascular disease, and poor health behavior among black patients.²⁸ However, given the known disparities in cancer outcomes, efforts to understand the benefits and the risks of new therapies in real population subgroups are critical.

Understanding populations that benefit from new targeted cancer therapies is of particular interest from a public policy perspective because of the costs associated with their use. In 2007, the sales of bevacizumab were approximately $2.3 billion dollars in the United States.²⁹ Much of this may have been in settings where the benefits were small. Reassuringly, bevacizumab use was higher in patients with colon cancer, for whom the benefits were more clearly substantiated by clinical trials showing a 35% reduction in the HR for overall mortality with the use of bevacizumab. Along the same lines, use was higher in patients with contraindications who had a diagnosis of colon cancer. However, in women with breast cancer, for whom the use of bevacizumab has been controversial,³⁰ and ultimately resulted in the reversal of FDA approval in 2011, the percentage of women who received bevacizumab increased from 3% in 2005 to 20% in 2007, with 54% receiving it with a contraindication and 25% of women having at least one complication after its use. Presumably, the use of bevacizumab in patients with breast cancer continued to rise between 2007 and 2011, given that the FDA did not grant approval until 2008.

It should be noted that, because of selection factors, we did not look at toxicity rates among patients who did not receive bevacizumab. For comparison purposes, in the clinical trials for colon cancer, thrombotic complications ranged from 13.8% to 19.4%, bleeding complications were approximately 3%, and GI perforation was 1.5%.^2,31 In the lung cancer trials, 4.4% had bleeding events and thrombotic and cardiovascular events were not reported.⁴ In the breast cancer trial, 1.6% had a thrombotic event, and less than 1% had either a stroke, left ventricular dysfunction, hemorrhage, or GI perforation.⁵ In patients without a known contraindication, we found that overall, approximately 30% of patients had at least one complication, and the types of complications differed by tumor type. Given the substantially greater toxicity we noted, our data raise concern for the comparative effectiveness of bevacizumab in elderly patients. Often when drugs are approved, short-term toxicities are evaluated but the long-term risks are not. Ideally, postmarketing studies of new drugs to address the long-term issues associated with new medications would provide this information.³² One way to do this is by active monitoring of claims data to identify safety concerns associated with new therapies as they become disseminated.

We acknowledge several important limitations of our study and of the SEER-Medicare database in general.³³ It is possible that not all patients with Medicare claims had the complication we assigned; however, we required participants to have two claims within the prior 12 months to reduce misclassification bias, which has been done routinely when looking for diagnoses.^34–36 In addition, we performed a sensitivity analysis requiring the two codes to be on different days. Although the numbers of toxicities decreased slightly, the multivariate analysis was unchanged. It is also possible that our results are not generalizable to younger patients, for whom the complication rates are likely to be much lower. SEER-Medicare lacks data on the severity of the toxicity; therefore, it is possible that these complications were mild and not life threatening. Alternatively, because mild complications are sometimes not captured with billing data, our findings may have underestimated some complications in the elderly. As with any analysis of administrative data, it is impossible to determine individual patient and physician preferences that may have influenced patterns of use. Importantly, the goal of our analysis was to explore patterns of use and toxicity and, as such, survival was not examined. Further studies of the effect of bevacizumab on survival in elderly patients with cancer with and without contraindications to treatment are clearly needed. Given that clinical trials for bevacizumab received widespread publicity, and that the patient population included stage IV incurable cancer, the risk-benefit ratio may have favored treatment despite prior contraindications.

Our study demonstrated that, despite known toxicities, bevacizumab was used frequently in settings where the risks and benefits have not been fully evaluated. The toxicities associated with bevacizumab use were substantially higher than documented in clinical trials. Of great concern is that, although black patients are less likely to receive bevacizumab, they are both more likely to receive the drug with a prior contraindication and are more likely to develop toxicities from therapy compared with white patients. This study emphasizes the importance of widening clinical trial inclusion criteria so that patients and providers can make informed decisions, and stresses the need for active monitoring of claims data and other sources to identify safety concerns associated with new therapies and drug efficacy in real world populations.

Appendix

Table A1.

Bevacizumab Contraindications and Cautions

Bevacizumab contraindications

Hypersensitivity to drug

Surgery, major within 28 days

Unhealed surgical wounds

Severe hemorrhage

Recent hemoptysis

GI perforation

Uncontrolled hypertension

Severe/arterial thromboembolism

Bevacizumab cautions

Congestive heart failure

Concurrent or past anthracycline use

Proteinuria

Thromboembolism history

Age > 65 years

NOTE. Data adapted.¹

Table A2.

Codes for Contraindications, Toxicity, and Cautions

Type of Event	ICD-9	HCPCS/CPT
Contraindication/toxicity
Thrombosis	Pulmonary embolism: 415.1, 415.11, 415.19	Venous thrombectomy: 34,401, 34,421, 34,451, 34,471, 34,490
Deep venous system	Portal vein thrombosis:452
	Other venous embolism and thrombosis: 453-453.9
Thrombosis arterial	Arterial embolism and thrombosis: 444, 444.0, 444.1, 444.21, 444.22, 444.8, 444.89, 444.9	Arterial embolectomy: 34,001, 34,051, 34,101, 34,111, 34,151, 34,201, 34,203
	Atheroembolism: 445, 445.0, 445.01, 445.02, 445.8, 445.81, 445.89
Hemoptysis	786.3
Cardiac	Acute myocardial infarction: 410-410.99	Coronary artery bypass: 33,510-33548
	Other acute and subacute forms of ischemic heart disease: 411-411.99	Prolonged extracorporeal circulation for cardiopulmonary insufficiency: 33,960
	Cardiac arrest: 427.5-427.59	Ventricular assist device: 33,983
	Heart failure: 428-428.99	Carotid endartarectomy: 33,572
	Sequelae of myocardial infarction: 429.7-429.79	Catheter placement in coronary arteries and left heart catheterization: 93,508-93524
Cardiac procedures	Defibrillator placement: 37.94-37.98
AICD, cardiac catheterization	Extracorporeal circulation, auzillary to open heart surgery: 39.61-39.63
	Percutaneous cardiopulmonary bypass: 39.65-39.66
	Percutaneous procedures (angioplasty/stent placement) 36.0-36.09
	Implantation of heart and circulatory assist systems: 37.61-37.69
Hemorrhage	Diseases of the pericardium: 423.0	Reopening of recent laparotomy: 49,002
	Intracerebral hemorrhage: 431, 432.9	Endoscopy with control of bleeding: 46,614
	Goodpasture's Syndrome: 446.21	Transurethral fulgration for postoperative bleeding occurring after the usual follow-up time: 52,606
	Hemorrhage, unspecified: 459.0	Thoracoscopy with control of traumatic hemorrhage: 32,654
	Esophageal hemorrhage: 530.82 Other specified disorders of the stomach and duodenum: 537.89	Colonoscopy through stoma with control of bleeding: 44,391
	Hemoperitoneum (nontraumatic): 568.81	Colonoscopy with control of bleeding: 45,382, 45,317,
	Hemorrhage of the GI tract: 578.9	Sigmoidoscopy with control of bleeding: 45,334
	Spontaneous ecchymoses: 782.7	Esophagoscopy with control of bleeding: 43,227
	Epistaxis: 784.7	Upper GI endoscopy with control of bleeding: 43,255
	Hemopytsis: 786.3	Small intestine endoscopy with control of bleeding: 44,366, 44,378
		Management of liver hemorrhage: 47,350
		Thoracotomy with control of traumatic hemorrhage: 32,110
		Sinus endoscopy with control of nasal hemorrhage: 31,238
		Curretage, postpartum: 59,160, Introduction of hemostatic agent or pack for spontaneous or traumatic vaginal hemorrhage: 57,180
		Control of nasal hemorrhage: 30,901-30906
		Control of nasopharyngeal hemorrhage: 42,970-42972
		Control of oropharyngeal hemorrhage: 42,960-42962
Stroke	Occlusion of cerebral arteries: 434.91, 434.01, 434.11
	Occlusion and stenosis of precerebral arteries: 433-433.9
GI perforation	Perforation of instestines: 569.8	Gastrotomy with suture repair of bleeding ulcer: 43,501
	Perforation of esophagus: 530.4-530.49
	Gastroesophageal laceration-hemorrhage syndrome: 530.7-530.79
	Gastric ulcer, chronic or acute with perforation: 531.1-531.29, 531.5-531.69
	Duodenal ulcer, chronic or acute with perforation: 532.1-532.29, 532.5-532.69
	Peptic ulcer, site unspecified, chronic or acute with perforation: 533.1-533.29, 533.5-533.69
	Gastrojejunal ulcer, chronic or acute with perforation: 534.1-534.29, 534.5-534.69
Caution
Thrombosis	Phlebitis of saphenous vein:451.0
Superficial venous system	Phlebitis of superficial veins of upper extremities: 451.82
	Phlebitis and thrombophlebitis: 451.1, 451.11, 451.19, 451.2, 451.8, 451.81, 451.83, 451.84, 451.89, 451.9
Hypertension	Essential and secondary hypertension: 401-405
Anthracycline		Doxorubicin: J9000,J9001, Epirubicin: J9178,J9180, Idarubicin: J9211

NOTE. Data adapted.⁷

Abbreviations: AICD, automatic implantable cardioverter defibrillator; ICD-9, International Classification of Disease (9th revision); HCPCS/CPT, Healthcare Common Procedure Coding System/Current Procedural Terminology.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Although all authors completed the disclosure declaration, the following author(s) and/or an author's immediate family member(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a “U” are those for which no compensation was received; those relationships marked with a “C” were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors.

Employment or Leadership Position: None Consultant or Advisory Role: Dawn L. Hershman, Genentech (C) Stock Ownership: None Honoraria: None Research Funding: None Expert Testimony: None Patents: None Other Remuneration: None

AUTHOR CONTRIBUTIONS

Conception and design: Dawn L. Hershman, Jason D. Wright, Alfred I. Neugut

Financial support: Dawn L. Hershman

Administrative support: Alfred I. Neugut

Collection and assembly of data: Dawn L. Hershman

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors