Frequency and Cost of Chemotherapy-Related Serious Adverse Effects in a Population Sample of Women With Breast Cancer

Abstract

Background: The number, nature, and costs of serious adverse effects experienced by younger women receiving chemotherapy for breast cancer outside of clinical trials are unknown. Methods: From a database of medical claims made by individuals with employer-provided health insurance between January 1998 and December 2002, we identified 12 239 women 63 years of age or younger with newly diagnosed breast cancer, of whom 4075 received chemotherapy during the 12 months after the initial breast cancer diagnosis and 8164 did not. Diagnostic codes for eight chemotherapy-related adverse effects were identified. Total hospitalizations for all causes, hospitalizations or emergency room visits for adverse effects that are typically related to chemotherapy, and health care expenditures were compared between the two groups of women. All statistical tests were two-sided. Results: Women who received chemotherapy were more likely than those who did not to be hospitalized or to visit the emergency room for all causes (61% versus 42%; mean difference = 19%, 95% confidence interval [CI] = 16.7% to 21.3%, P <.001) and for chemotherapy-related serious adverse effects (16% versus 5%, mean difference = 11%, 95% CI = 9.6% to 12.4%, P <.001). The percentages of chemotherapy recipients who were hospitalized or visited the emergency room during the year after their breast cancer diagnosis were 8.4% for fever or infection; 5.5% for neutropenia or thrombocytopenia; 2.5% for dehydration or electrolyte disorders; 2.4% for nausea, emesis, or diarrhea; 2.2% for anemia; 2% for constitutional symptoms; 1.2% for deep venous thrombosis or pulmonary embolus; and 0.9% for malnutrition. Chemotherapy recipients incurred large incremental expenditures for chemotherapy-related serious adverse effects ($1271 per person per year) and ambulatory encounters ($17 617 per person per year). Conclusions: Chemotherapy-related serious adverse effects among younger, commercially insured women with breast cancer may be more common than reported by large clinical trials and lead to more patient suffering and health care expenditures than previously estimated.

A drug-related serious adverse effect has been defined as any untoward medical occurrence that is related to drug use and results in death or significant disability/incapacity, requires hospital admission or prolongation of existing hospital stay, or is life threatening ( 1 , 2 ) . Population-based data suggest that drug-related adverse effects are common and cause morbidity and mortality accounting for approximately 6.5% of all hospitalizations ( 3 ) and as many as 100 000 deaths per year ( 4 , 5 ) in the United States. Although clinical trials of new drug therapies provide some information regarding the number and nature of serious adverse effects, reports of these complications are frequently inadequate ( 6 , 7 ) and may not accurately reflect the experiences of the general population ( 8 ) . Indeed, recent and widely publicized cases have demonstrated that serious adverse effects that are not fully appreciated during early clinical trials can appear after a drug is approved by the US Food and Drug Administration (FDA) and used by the public ( 9 ) . In fact, one study of serious adverse effects identified after FDA approval found that 22 cancer drugs had been linked with 25 serious adverse effects between 2000 and 2002 ( 10 ) .

Breast cancer is the most common indication for chemotherapy among women in the United States ( 11 ) , and chemotherapy drugs are the leading cause of serious drug-related adverse effects among women with breast cancer. Although large clinical trials of adjuvant chemotherapy for breast cancer have reported that hospitalizations for serious adverse effects are uncommon ( 12 ) , only one study ( 13 ) has investigated the frequency of chemotherapy-related serious adverse effects in the general population of women with breast cancer. That study, which analyzed data from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) database linked with Medicare claims data found that, for women with breast cancer who were treated in actual practice, the rates of hospitalization for eight chemotherapy-related toxicities were higher than had been expected based on results of large clinical trials ( 13 ) . However, this study included women diagnosed from 1991 to 1996, when medications that treat the adverse effects of chemotherapy (i.e., growth factors and newer antiemetics) were less available than they are today. Also, it included only women over age 65, a group that may be more likely than younger women to experience chemotherapy-related toxicities, in part because comorbid conditions are more common in older women. To more comprehensively characterize the risks and consequences of serious adverse effects among women in the general population treated with chemotherapy for breast cancer, we used insurance claims data to assess the frequencies and costs of eight chemotherapy-related adverse effects in women who were diagnosed with breast cancer when they were under 64 years of age and who were treated during the period from April 1, 1998, to December 31, 2002.

S UBJECTS AND M ETHODS

Data Source and Study Population

The MarketScan Commercial Claims and Encounters Research Database served as the data source for this analysis. This database was compiled by Medstat, a medical information company, from claims made to health plans that contract with large employers, state and local governments, and public organizations in the United States. The health plans included in the database provided employer-sponsored private fee-for-service or capitated health insurance to employees and their covered dependents. The database contains inpatient, outpatient, and prescription claims, as well as administrative data at the patient level that is tracked longitudinally as long as patients remain with their employer.

Data for more than 5.6 million individuals were available during the 5-year period from January 1, 1998, to December 31, 2002. The data set analyzed here was restricted to women aged 18–63 with two indications of a breast cancer diagnosis, i.e., code 174.x from the International Classification of Diseases (Ninth Revision) (ICD-9), separated by at least 30 days to exclude patients with ICD-9 codes that were entered by mistake or who were undergoing evaluation for breast abnormalities that were subsequently found to be benign. At least one breast cancer diagnostic code had to appear in conjunction with a face-to-face encounter with a health care provider, and patients with other cancer diagnoses were excluded. To restrict the study to those with new cancer diagnoses, eligible women had to have been enrolled in a health plan for at least 3 months before their first breast cancer diagnosis and to have no evidence of a cancer diagnosis during that time. To ensure that treatments and serious adverse effects were recorded in the data set, women had to be enrolled in a health plan for at least 12 months following the first breast cancer diagnosis (unless they entered hospice or died). Women 64 years of age or older when their breast cancer was first diagnosed were excluded to ensure that claims had not been submitted to Medicare. A total of 12 239 women from the database were eligible for this study.

Data Acquisition and Variable Definitions

For each member of the study population, we extracted, from the MarketScan database, information on age, health plan type (basic comprehensive, health maintenance organization, point of service, preferred provider organization, or point of service with capitation), region of residence (northeast, north central, south, west, or unknown), union status (union or nonunion), work hours (full time or part time), comorbidity score (no comorbid conditions or at least one condition), metastatic status (metastatic or nonmetastatic), claims for biopsies or pathology studies, receipt of cancer-related surgery or radiation therapy, receipt of trastuzumab or hormonal therapy, receipt of chemotherapy, and claims for inpatient, emergency room, and ambulatory encounters. For 83% of the women, data on prescriptions were available. Comorbid diagnoses were considered present if two claims were made at least 30 days apart during the 3 months before and the 12 months after the first breast cancer diagnosis. Comorbid diagnoses excluded cancer diagnoses and were coded using the method developed by Charlson et al. ( 14 ) , with some modifications ( 15 , 16 ) . Claims for cancer treatments, hospitalizations, emergency room visits, and prescriptions were assessed during the 12 months after the first breast cancer diagnosis. The 12-month cutoff was chosen to identify all initial breast cancer treatments and associated serious adverse effects.

Intravenous chemotherapy administration was identified using diagnostic and procedure codes for ambulatory encounters, including Current Procedural Terminology (CPT) codes for chemotherapy services and procedures, J & Q codes for selected chemotherapy agents, and ICD-9 and Diagnosis-Related Group (DRG) codes for chemotherapy administration ( Appendix A ). Women were considered to have received chemotherapy if at least one chemotherapy code was present during the 12 months after the first breast cancer diagnosis. The number of months during which chemotherapy was administered ( 1 – 12 ) and the cost of chemotherapy administration were recorded. The receipt of supportive medications (e.g., antiemetics and growth factors), breast biopsies, breast cancer surgeries, pathology studies, and radiation therapy during the 12 months after the first breast cancer diagnosis were identified using selected medication names and CPT, ICD-9, and DRG codes ( Appendix A ).

Direct medical expenditures were obtained for hospitalizations, emergency room visits, ambulatory encounters, and prescriptions; co-payments incurred by patients were reported for ambulatory encounters and prescriptions. Expenditures for chemotherapy medications and their administration were considered ambulatory expenditures but were also analyzed separately. Expenditures for hospitalizations or emergency room visits attributable to chemotherapy-related serious adverse effects were reported separately as well.

Matching the Cohorts

The population of eligible patients was divided into two cohorts: those who received chemotherapy within 12 months of their first breast cancer diagnosis and those who did not (4075 and 8164 women, respectively). Women who did not receive chemotherapy were on average older than those who did, were more likely to have had comorbidities, and were less likely to have received breast cancer–related surgery or radiation therapy. Therefore, propensity score–based matching was used as follows to designate a nonchemotherapy group with a distribution of measured covariates similar to that of the chemotherapy cohort ( 17 , 18 ) . First, a logistic regression model predicting receipt of chemotherapy was created, with age, health plan type, region of residence, union status, work hours, comorbidity score, receipt of cancer-related surgery or radiation therapy, and metastatic status as predictors. Next, each chemotherapy recipient was paired with a nonchemotherapy patient who was randomly selected from the subset of nonchemotherapy patients whose predicted logits were within a specified small distance of the chemotherapy recipient's predicted logit. Using a small distance between paired logits was possible because there were many nonchemotherapy patients from whom to select a match, and this small distance ensured that matched patients were similar (i.e., had nearly identical propensities to receive chemotherapy). After matching, the predicted probabilities corresponding to the logits were less than 0.01 apart for 80% of the pairs. Of the 4075 women in the chemotherapy group, 3526 (87%) were matched. Although the analysis reported here reflects only those patients who were matched, a sensitivity analysis incorporating all chemotherapy recipients yielded similar results (data not shown).

Defining and Identifying Adverse Effects of Therapy

For the purposes of this study, a serious adverse effect was defined as harm in a patient resulting in an emergency room visit or hospitalization. A patient was considered to have had a serious adverse effect if a diagnostic code for that event was associated with a hospitalization or emergency room visit during the 12 months following the first breast cancer diagnosis.

Based on their association with chemotherapy in previous clinical trials ( 12 ) , the following chemotherapy-related serious adverse effects were selected: 1) infections or fever; 2) neutropenia or thrombocytopenia; 3) anemia or transfused packed red blood cells; 4) nausea, emesis, or diarrhea; 5) dehydration or electrolyte abnormality; 6) malnutrition; 7) constitutional symptoms or nonspecific complications of treatments; and 8) deep venous thrombosis or pulmonary embolus. Chemotherapy-unrelated events (i.e., conditions that are never or only rarely [<1% of the time] associated with chemotherapy use in trials) were also selected: 1) fracture or dislocation, 2) asthma or chronic obstructive pulmonary disease, 3) renal failure, 4) thyroid disorder, and 5) headache. ICD-9, DRG, and CPT codes were identified and grouped together to form each chemotherapy-related serious adverse effect and each chemotherapy-unrelated event ( Appendix B ). Diagnostic codes associated with outpatient encounters were not considered when identifying either type of outcome, because they reflected less severe conditions and would have generated oversampling bias, given that women who received chemotherapy had more outpatient visits than those who did not.

There were insufficient clinical data in the MarketScan data set to definitively establish a cause–effect relationship between chemotherapy administration and the chemotherapy-related serious adverse effects. Therefore, odds ratios (ORs) for each of the chemotherapy-related serious adverse effects and each of the chemotherapy-unrelated events were derived to assess their association with chemotherapy administration. To further assess the strength of the association between the chemotherapy-related serious adverse effects and chemotherapy administration, odds ratios for each of the serious adverse effects were calculated for patient groups defined by receipt and nonreceipt of radiation therapy and receipt and nonreceipt of breast cancer surgery. The percentages of women who experienced any of the eight chemotherapy-related serious adverse effects were calculated for chemotherapy recipients and nonrecipients and reported as the primary outcome. The mean number of serious adverse effects per person per year was calculated because women could have experienced more than one serious adverse effect during the observation period. The percentage of women who experienced hospitalizations or emergency room visits for any reason, the mean duration of all hospitalizations in days, and the mean expenditure per patient per year were also reported.

Statistical Analysis

Statistical analyses were performed using SAS software version 9.1 (Cary, NC). Binary and categorical variables were compared using the Fisher's exact and chi-square tests, respectively. Distributions of continuous variables were summarized by their means and standard deviations and compared using t tests. All tests for statistical significance were two-sided. A multivariable logistic regression analysis for the odds of experiencing each chemotherapy-related serious adverse effect was performed to adjust for residual confounding and identify independent associations. Anthracycline chemotherapy use was entered as an independent binary predictor for women who received chemotherapy because its addition to chemotherapy regimens may be associated with additional short-term adverse effects ( 19 ) .

R ESULTS

Characteristics of Eligible Breast Cancer Patients

After matching, the study population included equal numbers of women who did and did not receive chemotherapy (3526 in each group). However, matching left some small imbalances ( Table 1 ). Chemotherapy recipients were younger than nonrecipients (by an average of 10 months), more likely to have metastatic disease (8.3% versus 6.9%, mean difference = 1.4%, 95% confidence interval [CI] = 0.16% to 2.64%, P = .03), and more likely to work full time at the time of diagnosis (74.1% versus 71.8%, mean difference = 2.3%, 95% CI = 0.23% to 4.37%, P = .005). Compared with women who did not receive chemotherapy, those who did were slightly more likely to have a breast biopsy or pathology claim and to have had radiation therapy. These differences persisted when the 534 women with metastatic disease were excluded from the analysis. Chemotherapy recipients and nonrecipients did not show differences in comorbidity score, type of health care plan, region of residence, or union status. Rates of breast cancer surgery and receipt of trastuzumab/hormonal agents were also similar for the two groups.

Women in the study population received a range of chemotherapy treatments; among those treated, 58% received alkylating agents, 51% received anthracyclines, 25% received taxanes, and 18% received antimetabolites. For 78% of the chemotherapy recipients, the duration of chemotherapy was 3–7 months, the approximate time needed to administer a course of adjuvant chemotherapy ( 12 ) ; 15% received chemotherapy for less than 3 months and 7% received it for more than 7 months. Women who had undergone chemotherapy were more likely than those who had not to receive prescriptions for antiemetics (84% versus 23%, mean difference = 61%, 95% CI = 59.0% to 63.0%, P <.001) or growth factors (12% versus 1%, mean difference = 11%, 95% CI = 9.7% to 12.2%, P <.001).

Odds Ratios of Serious Adverse Effects

The association between chemotherapy-related serious adverse effects and chemotherapy administration was assessed in two ways. First, we compared the odds of serious adverse effects experienced by women who did and did not receive chemotherapy for two categories of adverse effects—“chemotherapy related” and “chemotherapy unrelated” ( Table 2 ). Women who received chemotherapy had statistically significantly greater odds of experiencing each of the eight chemotherapy-related serious adverse effects and 3.6-fold (95% CI = 3.0 to 4.3) greater odds of experiencing at least one of the eight chemotherapy-related serious adverse effects. However, they were more likely to experience only one of the five chemotherapy-unrelated events. We also computed the odds ratios of each of the chemotherapy-related serious adverse effects for recipients of radiation therapy compared with nonrecipients and for those who underwent breast cancer surgery compared with those who did not ( Table 2 ). The lingering effects of chemotherapy given immediately before radiation therapy may have caused radiation therapy recipients to have greater odds of experiencing neutropenia or thrombocytopenia compared with those who did not receive radiation therapy (OR = 1.4, 95% CI = 1.0 to 1.8). The odds ratios for all other chemotherapy-related serious adverse effects among women who received radiation therapy relative to those who did not were not statistically significantly greater than 1. Finally, the odds ratios of chemotherapy-related serious adverse effects among women who received breast cancer surgery relative to those who did not were also not statistically significantly greater than 1.0.

A multivariable logistic regression model was developed to predict the odds ratio for a breast cancer patient experiencing at least one of the eight chemotherapy-related serious adverse effects. Only receipt of chemotherapy (OR = 3.7; 95% CI = 3.0 to 4.3) and the presence of metastatic disease (OR = 2.1; 95% CI = 1.6 to 2.7) were independently associated with a statistically significant increase in the odds of experiencing a serious adverse effect. During the 12 months after the initial breast cancer diagnosis, each additional month of chemotherapy was associated with a 20% increase in the odds of experiencing a chemotherapy-related serious adverse effect (95% CI = 17% to 23%). Among the women who received chemotherapy, anthracycline use was not associated with greater odds of experiencing any individual chemotherapy-related serious adverse effect (data not shown).

Effects of Chemotherapy on Rates of Hospitalization and Emergency Room Visits

More than half (51%) of the women in our cohort had at least one hospitalization or emergency room visit during the year after their first breast cancer diagnosis; the percentages among those who did and did not receive chemotherapy were 61% and 42%, respectively (difference = 19%, 95% CI = 16.7% to 21.3%) ( Fig. 1 ). The majority of all hospitalizations (57%) were related to breast cancer surgery. Chemotherapy recipients experienced more hospitalizations than nonrecipients (1.41 per person per year versus 1.25 per person per year, difference = 0.16, 95% CI = 0.10 to 0.22, P <.001), and they had longer hospital stays (5.0 versus 3.8 days, difference = 1.2, 95% CI = 0.6 to 1.7, P <.001). In the overall cohort, 10.5% of the patients had a hospitalization or emergency room visit for a chemotherapy-related serious adverse effect; for chemotherapy recipients and nonrecipients the corresponding percentages were 16% and 5%, respectively (difference = 11%, 95% CI = 9.6% to 12.4%, P <.001 [ Fig. 1 ]). Chemotherapy-related serious adverse effects were identified for 22% of all hospitalizations among chemotherapy recipients but only 12% of all hospitalizations among women who did not receive chemotherapy.

The rates of each of the chemotherapy-related serious adverse effects and the chemotherapy-unrelated events are presented in Fig. 2 . Fever or infection was the most common chemotherapy-related serious adverse effect and was associated with hospitalizations or emergency room visits in 8.4% of chemotherapy recipients ( Fig. 2 ). When we limited the cohort to women who had breast cancer surgery as a way to focus on incident cases and exclude women with metastatic disease (who may have been symptomatic because of their disease), we obtained similar results (data not shown). Because some women experienced multiple serious adverse effects, analyzing the percentage of women who experienced at least one underestimates the impact of these effects on the population. To address this issue, we calculated the number of serious adverse effects per person per year. Chemotherapy recipients had four times the number of chemotherapy-related serious adverse effects per person per year (0.25 versus 0.06, difference = 0.19, 95% CI = 0.16 to 0.21, P <.001).

Economic Costs of Breast Cancer Chemotherapy and Chemotherapy-Related Serious Adverse Effects

To estimate breast cancer–related medical expenditures for the whole population, we calculated expenditures per person per year across all women in each group regardless of whether they had a particular service ( Table 3 ). For the whole cohort of 7052 women with breast cancer, mean expenditure was $26 928 per person per year for ambulatory encounters (includes chemotherapy), $5594 per person per year for hospitalizations, $148 per person per year for emergency room visits, and $2056 per person per year for prescriptions. Among chemotherapy recipients, incremental expenditure for chemotherapy-related serious adverse effects (i.e., expenditure in excess of those made on behalf of the matched nonchemotherapy recipients) was $1271 per person per year ( Table 3 ). Chemotherapy recipients also had large incremental expenditures for inpatient care ($3213 per person per year), ambulatory encounters ($17 617 per person per year), and ambulatory co-payments ($299 per person per year). On an individual basis, the costs of chemotherapy-related serious adverse effects were striking: those who experienced chemotherapy-related serious adverse effects had large incremental expenditures for hospitalizations ($12 907 per person per year), prescriptions ($1908 per person per year), and prescription co-payments ($120 per person per year) compared with those who did not have chemotherapy-related serious adverse effects.

D ISCUSSION

This study is the first, to our knowledge, of chemotherapy-related serious adverse effects in a population-based sample of younger women with breast cancer. We found that eight chemotherapy-related serious adverse effects may be more common than reported by large clinical trials ( Table 4 ), and, therefore, that these adverse effects may be responsible for more patient suffering and higher health care expenditures than currently predicted. There are several reasons why the rates of serious adverse effects in the general population could be greater than the rates reported by large clinical trials. First, clinical trials are designed and powered primarily to identify the benefits of chemotherapy, not to characterize the risks of serious adverse effects. If only a few thousand patients are exposed to a chemotherapy regimen, then a clinical trial may not detect uncommon serious adverse effects or it may detect them but generate inaccurate risk estimates for rare events ( 10 , 20 ) . Second, there are differences—in terms of age, race, disease severity, comorbidities, socioeconomic status, and educational status—between the small fraction of women with cancer who participate in clinical trials and the general population ( 21 – 28 ) . Participants in a clinical trial typically have a better prognosis than nonparticipants ( 29 – 33 ) and therefore may be less likely to experience serious adverse effects. Third, some ( 34 , 35 ) , although not all ( 36 , 37 ) , studies suggest that participation in a clinical trial could be associated with improved outcomes independent of the effects of the treatment under investigation. Fourth, clinical trials may not detect all serious adverse effects; a recent study investigating the ability of physicians to identify adverse events in the setting of a phase II clinical trial of chemotherapy found physician reports to be neither sensitive nor specific in identifying common adverse effects ( 6 ) . Fifth, clinical trials frequently do a poor job of reporting the frequency and severity of serious adverse effects identified during the course of a trial. A review of 192 randomized drug trials ( 7 ) found that 25% did not report the number of discontinuations due to toxicity and 54% of those that did failed to give specific reasons for the discontinuations. Moreover, 50% of the trials inadequately reported the severity of clinical adverse effects, and 63% failed to properly report the severity of lab-determined toxicities.

One other study has analyzed the frequency of chemotherapy-related serious adverse effects in the general population of women with breast cancer ( 13 ) . It differed from this one in that it used SEER–Medicare data to identify hospitalizations for serious adverse effects among women aged 65 years and older, whereas this study analyzed insurance claims for hospitalizations and emergency room visits among women aged 18–63 years. Our finding that rates of serious infections, fevers, and low blood counts were greater than reported in large clinical trials agrees with the results of the SEER–Medicare study ( Table 4 ). Our study also found that the rate of pulmonary embolus or deep venous thrombosis was greater than reported in clinical trials (these adverse effects were not examined in the prior study). The rates of several chemotherapy-related serious adverse effects, including anemia, neutropenia, thrombocytopenia, and dehydration, were lower in our study compared with those observed in the SEER–Medicare study ( 13 ) . One explanation for this difference could be that our study included younger patients who may have had greater marrow reserve, fewer comorbidities, or better baseline performance status. Also, our study included patients treated when the use of growth factors and newer antiemetics may have been more widespread than during the time frame of the previous study.

Few studies have examined the economic outcomes associated with chemotherapy-related serious adverse effects. We found that, for breast cancer patients, incremental expenditure for chemotherapy-related serious adverse effects was $1271 per chemotherapy recipient per year. Data from our population-based sample suggest that at least 35 000 women under age 64 receive chemotherapy for breast cancer each year in the United States. If this estimate is correct, then incremental expenditure for hospitalizations or emergency room visits due to chemotherapy-related serious adverse effects could reach $45 million per year. We found that expenditure for inpatient care was $3213 per person per year higher among chemotherapy recipients than for nonrecipients and $12 907 per person per year higher among women who experienced serious adverse effects compared to those who did not experience them. These costs are higher than previous estimates of incremental costs for cancer patients, which found that expenditure for hospitalizations was $2716 higher among metastatic colorectal cancer patients treated with 5-FU ( 38 ) and $2725–$5565 higher per chemotherapy cycle among cancer patients with chemotherapy-induced mucositis who had been treated for a variety of different cancers with chemotherapy cycles of various lengths ( 39 ) .

Our study also showed that mean monthly health care expenditure for women with breast cancer was $2894. This value compares well with those from a retrospective analysis of insurance claims for patients with seven different types of cancer, which found that mean monthly health care costs ranged from $2187 to $7616 ( 40 ) . Combining ambulatory co-payments, prescription co-payments, and deductible payments, out-of-pocket expenditures averaged $91 per month. However, this figure underestimates actual out-of-pocket expenditures because it includes only some of the costs incurred by women with breast cancer. Interviews with 156 breast cancer patients found that out-of-pocket expenditures and lost income costs averaged $1455 per month ( 41 ) .

The increased odds of chemotherapy-related serious adverse effects reported by our study may be biased, in part, because women who do not receive chemotherapy tend to be older and to have more comorbid conditions and because women who do receive chemotherapy tend to have more advanced disease. To limit the extent of this bias, a propensity score–based matching algorithm was used to identify a group of women who did not receive chemotherapy but were relatively similar to those who did. However, even after matching, heterogeneous distributions of unmeasured covariates could still have confounded odds ratio estimates. In any event, our estimates of the actual frequencies of chemotherapy-related serious adverse effects among women who received chemotherapy would not be affected by the composition of the control group or the matching process. Although matching did remove 549 women from the chemotherapy group, the rates of chemotherapy-related serious adverse effects in the original chemotherapy cohort (4075 patients) were nearly identical to those in the matched chemotherapy cohort (3526 patients), so this group's ability to reflect the general population of chemotherapy recipients was not compromised by the matching process.

Because the insurance claims analyzed in this study contained limited amounts of clinical data, it was not possible to control for stage or to establish a direct link between chemotherapy administration and particular serious adverse effects. Using only diagnostic and procedure codes to identify serious adverse effects could have overestimated their frequency. However, even if one considered only the additional serious adverse effects experienced by the chemotherapy group, the rates of chemotherapy-related serious adverse effects would have been 6% for infections or fevers and 11% for all causes––values that are still greater than those reported by large clinical trials. Identifying serious medical conditions that were common among women under 64 and completely unrelated to all chemotherapy medications used for breast cancer was particularly challenging; however, in support of our classification, the data showed that most of the chemotherapy-unrelated events were not more common in chemotherapy recipients than in nonrecipients ( Table 2 ).

In conclusion, our results suggest that breast cancer chemotherapy may cause more patient suffering and higher health care costs than previously estimated. The findings support assertions that clinical trials often lack external validity ( 8 ) and underscore the fact that health care providers must carefully decide whether trial results can be applied to routine clinical practice. Although our findings may not substantially alter estimates of breast cancer patients' overall survival, they have important implications for quality of life and could affect decisions regarding therapy.

To help patients make informed decisions, the risks and benefits of therapy must be presented in ways that are relevant to them. Compared with the NCI's common terminology criteria, hospitalization rates may be a less objective and systematic method of estimating the frequencies of treatment-related serious adverse effects. However, they do confer important information that patients may find valuable. Improving efforts to estimate the frequencies of treatment-related serious adverse effects and presenting this information in ways that patients find meaningful would generate a more complete picture of the impact treatments have on patients' lives and facilitate informed decision making.

References