Synchronization of Treatment With Erythropoiesis-Stimulating Agents and Chemotherapy

Cancer is associated with an elevated risk of anemia that may be caused by the disease itself and by its treatment. A systematic review of studies providing prevalence estimates of anemia in cancer patients found a high degree of variability, ranging from 7% to 100% depending on cancer type, stage, and defi nition of anemia.¹ The development of anemia in cancer patients predicts shorter survival times² and may lead to fatigue, which has a negative impact on patientreported quality of life.^3,4 Compounding the problem of anemia that occurs with cancer is a high frequency of anemia caused by myelosuppressive chemotherapy.⁵ Management of anemia associated with chemotherapy may include red blood cell transfusions or the erythropoiesis-stimulating agents (ESAs) epoetin alfa (EA) and darbepoetin alfa (DA). Both EA and DA, when used for treatment of chemotherapy-induced anemia, have been associated with decreases in fatigue and some indications of improvements in quality of life.⁶

Recent controversy regarding the use of ESAs in patients on chemotherapy has centered on safety concerns and findings of increased risk of tumor progression and death in some populations on these drugs.^7-9 Nevertheless, use of ESAs for anemia associated with chemotherapy continues to be recommended for some patients.¹⁰ For those patients it is important to understand the degree of burden, for both patients and providers, that results from the combination of chemotherapy and ESA use.

Many chemotherapy regimens are administered in schedules of every 3 weeks (Q3W); examples include doxorubicin + cyclophosphamide,¹¹ paclitaxel + carboplatin, and cisplatin + docetaxel.¹² Epoetin alfa is indicated for dosing weekly (QW) or 3 times per week,⁹ whereas DA is currently indicated for dosing QW or Q3W.⁸ With Q3W chemotherapy administrations, an ESA that is given Q3W will not require any additional visits beyond those already planned for chemotherapy. In contrast, with more frequent ESA dosing schedules, patients may need to visit the clinic more often, which could lead to higher costs and greater disruption of their lives beyond that imposed by the chemotherapy treatment. The synchronization of treatment, such that ESA administrations occur during the same visits as chemotherapy administrations, may help reduce inconvenience and costs relative to ESA administrations given during otherwise-unneeded visits. The hematologic effects of synchronous (ie, same day as chemotherapy) versus asynchronous (ie, at the midpoint of a 3-week chemotherapy cycle) administration of DA have been found to be similar,¹³ which suggests that the benefi ts of treatment synchronization are not offset by any reduction in efficacy.

This retrospective cohort study examined the patterns of ESA and chemotherapy use in a real-world setting between 2002 and 2007. The primary objective was to estimate the proportions of patients who received these 2 kinds of treatment in synchronous versus asynchronous schedules and to identify predictors of greater synchronization of treatment.

METHODS

Data Source

The study data were obtained from the Varian Medical Oncology database of electronic medical records (EMRs) from outpatient oncology practices. The database includes information on more than 185,000 cancer patients from 18 oncology provider organizations comprising 78 clinic locations in the United States. At each patient visit to the clinic, the staff entered diagnoses, treatments, and other relevant information into the database. Treatment data included orders or prescriptions for medications, with specifi cs such as dose and route, as well as duration of supply of oral medications and amount and timing of drugs administered in the clinic. Laboratory results were typically fed directly from the lab into the EMR system; data included the date of the test, lab test name, result, units, and normal range. The data used for the present study were deidentifi ed, as required by Health Insurance Portability and Accountability Act (HIPAA) regulations. When using a HIPAA-compliant deidentifi ed database such as this, separate institutional review board approval is not required.

Cohort Definition

We selected patients with a primary tumor type of breast cancer, lung cancer, colorectal cancer, or lymphoma at any time in the database who received at least 1 conventional myelosuppressive chemotherapy agent administered in the clinic (ie, not self-administered) with a start date between January 1, 2002, and June 30, 2007. The cutoff date permitted a minimum of 3 months of follow-up prior to database cutoff of September 30, 2007. To qualify for the study, patients had to have at least 1 ESA administration in the database during a conventional chemotherapy regimen (ie, within 30 days after a conventional chemotherapy administration) and to be at least 18 years old in 2007.

From this set of cancer patients treated with conventional chemotherapy and an ESA, we excluded patients without an eligible chemotherapy episode of care, defined as an episode containing at least 2 administrations of conventional chemotherapy that occurred no more than 30 days apart, when the fi rst administration date was on or before June 30, 2007. Patients were also excluded if they lacked an eligible ESA episode of care during a corresponding eligible chemotherapy episode of care. An ESA episode of care was eligible if it contained only 1 type of ESA (either EA or DA), with at least 2 administrations occurring no more than 42 days apart and within the start to stop dates of a chemotherapy episode of care. As the EMR database provides data only on individual ESA administrations and not on ESA episodes of care, we followed the model of claims analyses in this field^14,15 in using a 42-day period to defi ne an episode of care, which allows up to twice the indicated interval for DA. Finally, patients under age 18 years at the start of the eligible chemotherapy episode and those with multiple primary cancers during or prior to the chemotherapy episode were excluded. Each included patient was assigned an index date corresponding to the start date of the qualifying chemotherapy episode.

Baseline Characteristics

Baseline characteristics of patients were identified using data from the EMR system from on or before the index date. Age was estimated as year of index date minus year of birth; exact date of birth is prohibited by HIPAA privacy rules from inclusion in deidentified healthcare data. Patients were classified with respect to sex, primary cancer type, geographic region, clinic type (ie, community based vs hospital affi liated), and year of index date. Because hemoglobin levels during chemotherapy may be an important predictor of ESA treatment patterns, we also found the lowest hemoglobin level recorded in the EMR system for each patient during the chemotherapy regimen.

Chemotherapy Administrations

Each administration date of chemotherapy was identified. From these chemotherapy dates, we calculated the interval between sequential chemotherapy administrations for each patient. Schedules were categorized into intervals of

For each patient, we calculated the mode of the distribution of his or her chemotherapy schedules, which for most patients should represent the planned chemotherapy schedule. To investigate how well the ESA administrations fi t into the preexisting chemotherapy schedule, we used the modal schedule to defi ne the patient’s overall chemotherapy schedule.

The cohort was split into primary and secondary cohorts based on chemotherapy schedules. The primary cohort comprised patients on fairly regular chemotherapy schedules (ie, more than 50% of intervals between chemotherapy administrations were classifi ed as the same schedule) ranging from weekly to every 4 weeks. Synchronization with an irregular schedule is qualitatively different from synchronization with a regular schedule, with the former requiring a more complex planning process; hence, the patients with irregular or unusually frequent schedules were followed as a secondary cohort. Patients with no more than 50% of chemotherapy administrations on the same schedule (QW, Q2W, Q3W, Q4W) or with a schedule

ESA Administrations

For each patient, the study period was defi ned as the time on conventional myelosuppressive chemotherapy surrounding the first qualifying episode of ESA use (ie, an episode that included at least 2 ESA administrations separated by no more than 42 days). Dosing schedules for ESAs are indicated at no more than 3 weeks apart; any ESA given after a gap of more than twice this duration was hence assumed to represent a new episode of ESA care. Those ESAs given more than 30 days after most recent chemotherapy administration were not considered for this analysis, as they might not be considered treatment for anemia associated with chemotherapy.

The primary outcome of this study was the synchronization of chemotherapy and ESA administration, defined as the co-occurrence of both treatments on the same day. The date of each administration of an ESA during the qualifying ESA episode was determined. We calculated the median interval between sequential ESA administrations and categorized these median intervals as for chemotherapy (

of >Q4W). Each eligible ESA administration was coded as occurring on the same day as a chemotherapy administration (ie, synchronized) or not.

Because visits to the oncology clinic may occur for reasons other than chemotherapy or ESA administrations, we also found the total number of visits to the clinic during the fi rst month of the ESA episode for all patients with at least 1 full month of follow-up during the ESA episode.

Analysis

Baseline characteristics were compared between patients on EA and DA using t tests for continuous variables (eg, age), Wilcoxon rank sum tests for ordinal variables (eg, year of regimen start), and χ² tests for categorical variables (eg, clinic type). The proportions of patients on EA and DA with each chemotherapy schedule (QW, Q2W, Q3W, and Q4W) were compared using χ² analyses. Chemotherapy and ESA schedules for each patient were crosstabulated to show the distribution of same-schedule versus different-schedule administrations of the 2 types of drugs. Dosing schedules for EA remained the same throughout the study period, but dosing of DA at Q3W schedules was approved by the US Food and Drug Administration in early 2006.¹⁶ We therefore performed the synchronization analyses both overall and stratifi ed by year of regimen start, categorized as 2002-2005 and 2006-2007.

All ESA administrations during the qualifying episode for each patient, which were coded as synchronized with chemotherapy or not, were entered into generalized estimating equation (GEE) models. These models allow multiple observations per patient and account for withinpatient correlation of results. Models were constructed for all patients together, separately by year strata of 2002-2005 and 2006-2007, and stratifi ed by tumor type (breast, lung, colorectal, or lymphoma). All models were adjusted for each of the baseline characteristics described above, as all of these variables were either of interest a priori and/or were known to be associated with treatment patterns in many studies (eg, geographic region, hospital-affi liated vs community-based clinic).

To determine whether greater synchronization of ESA administrations with chemotherapy was related to fewer visits to the clinic overall, we calculated the correlation between proportion of ESA administrations that were synchronized during the fi rst month of follow-up and total number of visits in the same time period. To control for potential confounders, we also constructed a linear regression model of total number of visits in the fi rst month, where having at least 75% of ESAs synchronized with chemotherapy during the month was the primary predictor of interest, and all of the baseline characteristics described above were entered as potential covariates. The final covariates in the model were chosen through backward selection with P <.10. These analyses were conducted among only the patients who had at least a full 30 days of follow-up during the qualifying ESA episode.

For eligible patients in the secondary cohort (ie, those with irregular or unusually frequent chemotherapy schedules), we ran the above analyses but without any stratifi cation by administration schedule. Another important subset of patients were those from the primary cohort who would meet current guidelines for ESA administration (ie, having hemoglobin <10 g/dL on initiation of the ESA). We reran the primary GEE model in this patient subset to determine whether any differences existed in the results relative to the full cohort in the present study.

Because some clinics might have a policy of using only 1 of the 2 ESAs exclusively, preventing any efforts at intentional synchronization through drug choice, we also conducted sensitivity analyses, dropping any clinic with no variability in type of ESA administered (EA or DA). As the study was descriptive in nature rather than hypothesis driven, no Bonferroni or other adjustments were made for the multiple P values that were generated. All data management and analyses were conducted using SAS version 9.1 (SAS, Cary, NC).

RESULTS

The

Figure

shows the disposition of patients, beginning with the initially eligible adult patients with 1 of the 4 tumor types of interest on conventional chemotherapy with ESA treatment. Of 12,977 patients initially identified, 3950 were excluded, leaving 8044 patients in the primary cohort (those whose chemotherapy regimen followed a regular schedule) and 983 in the secondary cohort of patients with irregular chemotherapy schedules. Baseline characteristics, as well as chemotherapy and ESA schedules and the lowest hemoglobin level recorded during the chemotherapy regimen, are shown in

Table 1

. The mean age of DA users was 60.0 years, and the mean age of EA users was 60.9 years (P = .002). The proportions of men and women were similar in the 2 groups (P = .20), with about 71% women overall. Use of DA was infrequent in the early years of its availability on the market and later increased to be more commonly used than EA (P <.001). Darbepoetin alfa was used more often than EA for patients with breast cancer, and EA was used more than DA for patients with lung cancer (P = .002). Darbepoetin alfa was used more often in the Midwest and West, and EA was used more often in the Northeast and South of the United States (P <.001). Hospital-affiliated clinics had more DA use than community-based clinics (P <.001). The lowest hemoglobin level recorded at any time during the qualifying chemotherapy episode for each patient was lower on average for DA than EA users (means of 9.7 and 10.0 g/dL, respectively; P <.001). More EA patients had a QW chemotherapy schedule and more DA patients a Q2W chemotherapy schedule, although the overall pattern did not achieve statistical signifi cance (P = .07). The most common ESA schedules were QW for EA and Q2W for DA (P <.001).

The cross-tabulation of ESA schedule by chemotherapy schedule is shown separately for EA and DA and stratified by year in

Tables 2A

and

2B

. In 2002-2005, the most common DA dosing schedule for patients with Q3W chemotherapy schedules was the Q2W schedule, which was clearly unsynchronized but was the least frequent dose then indicated. In 2006-2007, after Q3W dosing was approved for DA, 45% of patients on Q3W chemotherapy received DA on the same schedule. Overall the most common DA schedule was Q2W, used most often for patients with QW and Q2W chemotherapy.

Epoetin alfa, which is indicated for QW or dosing 3 times per week,⁹ was administered on a QW schedule to 84% of patients in 2002-2005 and 80% in 2006-2007. Dosing of EA Q2W was seen in some patients on Q2W or Q4W chemotherapy, but this schedule was far less common than for DA, as would be expected given the labeled dosing guidelines.

Taking each ESA administration as a separate observation, we found that 70.8% of DA and 63.2% of EA administrations were given on the same day as a chemotherapy administration. The GEE model of synchronization for all patients is presented in

Table 3

. Adjusting for all baseline characteristics, synchronization with chemotherapy was higher for DA than for EA, with an odds ratio (OR) of 1.45 (95% confi dence interval [CI] = 1.38, 1.52). Chemotherapy schedule was the strongest predictor of synchronization, as would be expected; the less frequent the chemotherapy, the less likely the ESA was to be given on the same day as chemotherapy. Odds ratios for chemotherapy schedule went as low as 0.13 (95% CI = 0.12, 0.15) for the least frequent chemotherapy schedule examined (Q4W). Synchronization also was less likely in community-based than hospital-affiliated clinics (OR = 0.53; 95% CI = 0.49, 0.57) and in the Western United States compared with the South (OR = 0.88; 95% CI = 0.82, 0.94). Greater synchronization was seen in the Northeast and Midwest compared with the South, with ORs of 1.21 and 1.47, respectively, and in patients whose lowest hemoglobin level during the chemotherapy regimen was 9 g/dL or above (OR = 1.20; 95% CI = 1.15, 1.26). Patients with colorectal cancer had the greatest synchronization of treatment (OR = 1.27; 95% CI = 1.19, 1.36), with breast cancer the second highest (OR = 1.13; 95% CI = 1.07, 1.20) compared with lung cancer. Lymphoma showed a slight tendency to be associated with lower synchronization than lung cancer (OR = 0.94; 95% CI = 0.88, 1.01), which upon examination of the separate models for 2002-2005 and 2006-2007 appeared to be driven entirely by the earlier year stratum; synchronization in 2006-2007 alone was comparable to that for lung cancer (OR = 1.01; 95% CI = 0.90, 1.13; data not shown).

Examination of the models stratifi ed by tumor type and year revealed that not only was synchronization lower for lymphoma compared with other cancer types in 2002-2005, but also in these earlier years the effect of DA versus EA disappeared in the lymphoma group (OR = 0.99; 95% CI = 0.85, 1.16). In contrast, the OR for DA versus EA in 2006-2007 for the lymphoma patients was an unusually high 2.60 (95% CI = 1.99, 3.39; data not shown). The other tumor types and the model of all tumor types stratified by year showed little difference from the overall model presented in Table 3.

A total of 5392 patients had ESA episodes lasting at least 30 days and hence were eligible for the analyses of total number of clinic visits in the fi rst month. We found the total number of visits to be signifi cantly negatively correlated with the proportion of ESAs that were synchronized with chemotherapy in the fi rst month (r = −0.20, P <.0001). The multivariate linear regression model adjusting for significant baseline predictors of total number of visits is shown in

Table 4

. This model confi rmed the finding in the correlation of greater synchronization associated with fewer visits (P <.0001).

We compared the baseline characteristics of the 983 patients in the secondary cohort (ie, those with irregular or very frequent chemotherapy schedules) with those of the primary cohort and found that the former were older and more likely to be male, and that the majority (52%) were diagnosed with lung cancer. Their proportion of ESAs synchronized with chemotherapy was lower than the proportion for the primary cohort, 62.2% for DA and 56.6% for EA. The GEE analysis within this excluded group showed results similar to those for the model for the primary cohort, with an OR for DA versus EA of 1.31 (95% CI = 1.17, 1.47).

The analysis of patients with hemoglobin <10 g/dL (ie, those meeting the current guidelines for ESA use) at the start of the ESA episode found a pattern of results that was similar to the primary analysis described above. The OR for DA versus EA in this patient subset was 1.35 (95% CI = 1.23, 1.49).

Most clinics showed some variability in the ESA used. A sensitivity analysis excluding the 393 patients from clinics that used only 1 type of ESA (EA or DA) obtained results similar to those for the full primary cohort, with synchronization remaining signifi cantly higher for DA than for EA (OR = 1.44; 95% CI = 1.37, 1.51).

DISCUSSION

The majority of ESA treatment for patients on chemotherapy was found in this sample of outpatient oncology clinics to occur when patients were already in the clinic to receive chemotherapy. This synchronization of treatment may allow reduced utilization of clinic resources, greater effi ciency in medical care, and increased convenience for patients, in that fewer clinic visits may need to be scheduled, travel to and from the clinic can be reduced, and the cost and time taken for additional visits may be avoided. Synchronizing an ESA with chemotherapy is easier with flexible and less frequent ESA dosing, leading to the higher synchronization seen with DA than EA. Greater synchronization of ESA administrations with visits for chemotherapy administrations was found to be associated with fewer visits to the clinic overall.

The primary analyses included only patients on fairly regular chemotherapy schedules. An informal check for selection bias that might thereby be introduced was conducted by investigating synchronization among patients with irregular or unusually frequent chemotherapy schedules. Comparison of the results for this secondary cohort with the results for the primary cohort found a similar effect for DA versus EA but lower synchronization overall. This suggests that not all of our results can be assumed to generalize to patients receiving chemotherapy at intervals other than the standard QW-Q4W schedules. In the present study, lung cancer patients were the most common subgroup found with unusual schedules; chemotherapy treatment may be more complex, aggressive, and variable across time than that for the other cancer types investigated, because of the much poorer success rates with treatment of lung cancer.¹⁷ Synchronization of treatment under such circumstances may not be feasible in many cases.

The present study was carried out using EMRs, which may contain erroneous entries and typically reflect only diagnoses and treatments provided in the oncology clinic. Some of the hemoglobin levels classified as unknown may have been taken but not entered into the EMR system. Red blood cell transfusions, which also are used for treatment of anemia, are largely underreported in the database and hence were not included in the present study. Although cost savings are a hypothesized benefit of greater synchronization of treatment, cost data are largely unavailable in this database and hence could not be examined directly.

The time frame of the present study predated the label changes made in 2008, which restricted initiation of ESA treatment to patients with hemoglobin <10 g/dL among those receiving concomitant myelosuppressive chemotherapy, with the dose adjusted to maintain the lowest hemoglobin level sufficient to avoid red blood cell transfusion. The August 2008 labels also added a statement that ESAs are not indicated in patients receiving myelosuppressive therapy when the anticipated outcome is cure, because of the absence of studies that adequately characterize the impact of ESAs on progression-free and overall survival.^8,9 Under the new labels, many of the patients in our study who were treated with ESAs would presumably not receive ESA treatment. Our sensitivity analysis of patients with hemoglobin <10 g/dL at the start of ESA treatment found results similar to those for the overall cohort of patients treated with ESAs in 2002-2007, although patterns in 2008-2009 could still differ from those seen here.

In conclusion, we found that ESA treatment among patients on chemotherapy often is administered during visits for conventional chemotherapy administration. The more flexible dosing schedules indicated for DA compared with EA allowed for more convenient treatment patterns, in that fewer additional clinic visits were needed. Future pharmacoeconomic analyses should examine the cost savings that may result from synchronized treatment compared with ESA treatment administered in separate visits from chemotherapy.