Costs associated with adverse events in patients with metastatic renal cell carcinoma

Abstract

Objective:

To estimate adverse event (AE) costs in patients receiving targeted therapies for the first line treatment of metastatic renal cell carcinoma (mRCC).

Methods:

Retrospective study based on healthcare claims data for patients with mRCC, aged ≥18 years, receiving first-line treatment with targeted therapies. AEs of interest comprised of abdominal pain, back pain, diarrhea, dyspnea, extremity pain, fatigue/asthenia, hand-foot syndrome, hypertension, lymphopenia, nausea/vomiting, neutropenia, proteinuria, and thrombocytopenia. Healthcare encounters for AEs were based on ICD-9-CM diagnosis/procedure codes on healthcare claims. AE costs were examined over a 30-day period, beginning with the date of first mention of AE, and were estimated based on the difference in total costs between patients with and without events. Drug costs of targeted agents were excluded from the analysis. Multivariate generalized linear models with a log-link function and gamma response probability distribution were utilized to control for differences in baseline characteristics between patients with and without evidence of AEs.

Results:

A total of 533 patients were included in this analysis: 418 patients with AE and 115 patients without AE. Baseline characteristics were generally similar between patients in the two groups. The GLM-based estimate of incremental 30-day post-event costs among patients with evidence of any adverse events was $9807 (95% CI = $4386–$22,947). For all types of adverse events examined, the estimated difference in costs between evented and non-evented patients was positive; the 95% CI did not include zero for all of the adverse events considered, except hypertension and proteinurea. Study limitations include errors of commission/omission, especially as they may affect case-finding methods that rely on ICD-9-CM diagnosis and procedure codes, as was the case in the current study.

Conclusion:

Costs associated with AEs of first-line targeted therapies are substantial in patients with mRCC. Efforts to prevent and/or better manage these events may reduce overall healthcare costs.

Keywords::

• Renal cell carcinoma
• Cost
• Adverse events

Introduction

The American Cancer Society estimates that in 2013 there will be 65,150 new cases of kidney cancer and 13,680 deaths due to the disease1. Renal Cell Carcinoma (RCC) accounts for ∼90% of these kidney cancers. While 5-year survival rates for local and regional disease are 90.8% and 63.3%, respectively, the prognosis is poor for metastatic/distant disease, with 5-year survival of 11.1%2.

Since RCC remains clinically occult for most of its course, it often gets diagnosed in advanced stages. As a result ∼25–30% of patients have metastatic disease at diagnosis3.

Metastatic RCC (mRCC) is among the most treatment resistant malignancies3, with nephrectomy, metastectomy, and cytotoxic chemotherapy having limited or no effectiveness4. Cytokine therapy (interleukin-2 [IL-2] and/or interferon-α [IFN]) have been the mainstay course of treatment for the past two decades, but it is of limited benefit in selected patients only3. Advances in understanding of the biology and genetics of RCC have led to novel targeted approaches to treatment, including tyrosine kinase inhibitors (sorafenib, sunitinib, pazopanib); the monoclonal antibody (mAb), bevacizumab; mammalian target of rapamycin (mTOR) inhibitors; and the kinase inhibitors, temsirolimus and everolimus (second-line therapy only)4.

While these targeted therapies are more effective than the earlier treatment strategies that preceded them and, therefore, have become the standard of care in advanced RCC5, moderate-to-severe adverse events have been reported in clinical trials, including fatigue, diarrhea, hypertension, hand-foot syndrome, stomatitis, bleeding, and gastrointestinal perforation. These adverse events are likely associated with additional treatments used to manage them and/or hospitalizations, thereby increasing demand for healthcare resources6, and ultimately cost of treating a patient with mRCC. There is limited information on the economic costs of these events in patients receiving targeted therapies for mRCC. Hence, the objective of this study was to quantify the costs of managing the most common (≥5%) moderate-to-severe (grade 3 and 4) adverse events associated with targeted therapies used in first line treatment of mRCC.

Patients and methods

This study used a retrospective cohort design and data from a large private health insurance claims database to estimate the costs associated with adverse events in patients with mRCC receiving selected targeted agents indicated for first-line treatment of this disease. Study subjects comprised all persons, aged ≥18 years, with evidence (based on algorithms using ICD-9-CM diagnosis and procedure codes) of mRCC and receipt of targeted therapy between January 1, 2000 and December 31, 2009. All analyses were descriptive in nature, as there were no a priori study hypotheses.

Data for this study was obtained from the Optum Impact National Manage Care Benchmark Database (a.k.a. Integrated Heathcare Information Services (IHCIS)), which contains facility, professional-service, and retail (i.e., outpatient) pharmacy claims from more than 46 health plans throughout the US, providing healthcare coverage to ∼12 million (with both pharmacy and medical eligibility) persons annually. Demographic and eligibility data are available, including age, gender, geographic region, coverage type, and dates of eligibility for healthcare coverage. In addition to selected demographic and eligibility information, data available for each facility and professional-service claim include date and place of service, diagnoses (in International Classification of Diseases, Ninth Edition, Clinical Modification [ICD-9-CM] format), procedures performed/services rendered (in Health Care Financing Administration Common Procedure Coding System [HCPCS], ICD-9-CM, and Uniform Bill-92 [UB-92] formats), discharge disposition (inpatient facility claims only), and quantity of services (professional-service claims only). Data available for each retail pharmacy claim include the drug dispensed (in National Drug Code [NDC] format), dispensing date, quantity dispensed, and number of days of therapy supplied. All claims include paid (i.e., reimbursed) amounts, including patient deductibles, copays, and/or coinsurance amounts. All data can be arrayed chronologically to provide a detailed longitudinal profile of all medical and pharmacy services used by each beneficiary. All patient-identifying information is fully encrypted and the database is compliant with the 1996 Health Insurance Portability and Accountability Act (HIPPA).

For possible inclusion in the study sample, patients had to have one or more days of eligibility for comprehensive health benefits between January 1, 2000 and December 31, 2009, two or more medical encounters with a diagnosis of kidney cancer (ICD-9-CM 189.0) or malignant neoplasm of the renal pelvis (189.1), and two or more medical encounters with a diagnosis of distant secondary malignant neoplasm (ICD-9-CM 197.XX-199.0, excluding 198.0 [kidney metastasis]) on different days less than 120 days apart (the date of the earliest such encounter was designated the ‘index date’). As the same ICD-9-CM diagnosis codes are used for RCC and Transitional Cell Carcinoma (TCC) of the renal pelvis, patients selected via this initial algorithm, however, could have either of these two diseases.

Differences in the diagnosis and treatment of these two diseases therefore were used to differentiate patients with advanced RCC vs TCC. While cytotoxic chemotherapy (e.g., methotrexate, vinblastine, doxorubicin, and cisplatin [M-VAC], paclitaxel, permetrexed, etc.) is a key component in the treatment of TCC, it is ineffective (and hence only rarely used) in patients with advanced RCC. Moreover, while cystoscopy is very often undertaken to ascertain the scope of tumor spread in patients with TCC, it is not performed in patients in whom advanced RCC is suspected. Finally, while radical cystectomy is sometimes performed in patients with TCC involving the bladder, this intervention is not customarily performed in patients with RCC. Accordingly, patients were excluded from the study sample if they (1) had evidence of receipt of chemotherapeutic agents indicated or used in the treatment of advanced TCC (i.e., bleomycin [J9040], carboplatin [J9045], cisplatin [J9060], doxorubicin [J900], 5-fluorouracil [J9190], gallium nitrate [J1457], gemcitabine [J9021], ifosfamide [J9208], methotrexate [J9260] permetrexed [J9305], docetaxel [J9328], paclitaxel [J9265], or vinblastine [J9360]) during the 30-day period preceding their index date or anytime thereafter; (2) had evidence of cystoscopy or biopsy of the bladder (57.3X) in the 30-day period preceding their index date or any time thereafter; (3) had evidence of radical cystectomy (57.7X) at anytime; (4) had evidence of any other primary cancer (ICD-9-CM 140–188, 189.2–195, 200–208) during the 365-day period preceding their index date (‘pre-index period’) at a site that is different than an established site of distant metastases (adjacent lymph nodes, the lungs, the liver, the bones, and the brain); (5) were not continuously eligible for health benefits during the 365-day period preceding their index date; (6) were aged <18 years as of the index date; (7) were aged ≥65 years and not enrolled in Medicare HMO; or (8) were a Medicaid beneficiary.

With regards to treatment-related adverse events (AEs), only Grade 3 or 4 AEs (i.e., moderate or severe in nature) that were reported with a frequency ≥5% in randomized controlled trials of sunitinib, sorafenib, bevacizumab, or pazopanib were included in the current study; this was ascertained based on a review of product label information for these therapies. Adverse events that met these criteria for one or more targeted therapies included: Abdominal pain (789.0); Back pain (724.1, 724.2, 724.5); Diarrhea (787.91) or receipt of anti-diarrheal or anti-motility agents based on NDC codes; Dyspnea (786.05); Extremity pain (729.5); Fatigue/asthenia (780.7X); Hand-foot syndrome (693.0); Hypertension (405.XX, 997.91); Lymphopenia (288.51); Nausea and/or vomiting (787.0X) or receipt of antiemetics based on J codes and NDC codes; Neutropenia (288.0, 288.03); Proteinuria (791.0),); and thrombocytopenia (287.4). Patients were deemed to have experienced any of the above-mentioned events if they had any evidence thereof during follow-up, based on a scan of all inpatient and outpatient medical claims, and all outpatient pharmacy claims (as appropriate). Grade 3 or 4 category adverse events that were reported with a frequency ≥5% in randomized controlled trials but which could not be identified by means of coding information in healthcare claims data, and, therefore, not included in the current study, were lab abnormalities for alanine transaminase (ALT), aspartate transaminase (AST), lipase, amylase, uric acid, phosphorus, blood glucose, serum sodium, hemoglobin, and leukocytes.

Patients with evidence of the selected adverse events during follow-up were examined based on medical and pharmacy claims. Patients were counted once only for each event, and time since initiation of treatment was determined in relation to the first event. The impact of adverse events on healthcare costs was examined among all patients with mRCC who received the targeted therapies; the costs of targeted therapy were excluded from these analyses. Since it is possible for patients to experience more than one type of adverse event, and since attribution of care to specific causes can be problematic, the analyses were conducted for each adverse event separately. For each adverse event under consideration, healthcare costs over a period of 30 days following first mention of the event (‘post-event period’) was examined, comparing costs between patients with evidence of the event to patients without such evidence; for non-evented patients, follow-up time began as of a ‘shadow’ date that was assigned based on mean time to event among evented patients.

To control for differences in baseline characteristics between patients with and without evidence of adverse events, healthcare costs during the 30-day post-event period were analyzed using multivariate generalized linear (GLM) models with a log-link function and gamma response probability distribution. A separate model was estimated for each adverse event. Independent variables in each model included a binary variable indicating presence/absence of the event in question and the following covariates: age (as of index date), gender, calendar year in which index date fell, time from diagnosis to initial receipt of targeted therapy, third-party payer, US census region, and Charlson comorbidity index at index date. Using the estimated parameter (β) for the binary (i.e., 0/1 dummy) variable for the event in question and mean total costs among patients who did not experience the adverse event of interest (X), the increase in total costs of medical-care services among patients who experienced the event was estimated as X*[exp(β) − 1]. All analyses were conducted using SAS Proprietary Software, Release 9.2 (SAS Institute Inc., Cary, NC). Costs were estimated using paid (i.e., reimbursed) amounts. Payment amounts were not adjusted for inflation, and cost estimates therefore reflected the experience of study subjects between 2001 and 2009.

Results

A total of 1670 patients (aged ≥ 18 years) were identified with mRCC. Table 1 reports sample selection. As is evident from Table 1, principal reasons for exclusion (not mutually exclusive) included: (1) evidence of another primary cancer (44.6%)—possibly due to miscoding of primary vs secondary cancers since ICD-9 codes were used as proxy for cancer diagnosis; (2) age ≥65 years and not on Medicare risk plan (44.2%); (3) gaps in eligibility for comprehensive health benefits during the 12-month period preceding the diagnosis date (34.6%); and (4) evidence of TCC (13.9%). Among the 1670 who met the study inclusion/exclusion criteria, 533 mRCC patients had evidence of receiving targeted therapy and were thereby included in the current analyses. Among the 533 patients in the final study sample, none had received pazopanib (this result was not unexpected, since pazopanib received FDA approval in October 2009 and our study period ended in December 2009).

Table 1. Sample selection—patients with advanced renal cell carcinoma (RCC) in IH CIS (2000–2009).

Criteria	Count
Two or more diagnosis of kidney cancer (kidney or pelvis)	58,210
Two or more diagnosis of secondary malignant neoplasm, not more than 120 days apart	10,292
Exclusions:
Evidence of TCC	1431
Evidence of other primary cancer during 12-month  period prior to index date	4616
No continuous eligibility for comprehensive benefits  during 12-month prior index date	3561
Age <18 as of index date	40
On Medicaid	37
Age >64 and not on Medicaid Risk Plan	4546
Total excluded	8622
Patients remaining	1670

Percentages of patients who experienced the adverse events of interest are reported in Table 2. Abdominal pain (50.1%), fatigue/asthnea (47.3%), and dyspnea (44.3%) were among the most common AEs that were observed in this study.

Table 2. Number (percentage) of patients with metastatic RCC receiving targeted therapy with evidence of selected adverse events.

Adverse events	No. of cases	%
Abdominal pain	267	50.1
Fatigue/asthenia	252	47.3
Dyspnea	236	44.3
Back pain	199	37.3
Nausea and/or vomiting	194	36.4
Extremity pain	186	34.9
Diarrhea	97	18.2
Proteinuria	19	3.6
Hand–foot syndrome	18	3.4
Hypertension	11	2.1
Neutropenia	11	2.1
Lymphopenia	1	0.2

Baseline characteristics were generally similar between patients with evidence of adverse events versus those without such evidence. The ratio of male/female patients with evidence of adverse event was 70.3%/29.7% compared to 69.6%/30.4% among patients with no evidence of adverse event. Mean (SD) age for patients with adverse event was 60 years (6.9 years) compared to 63 years (7.1 years) in patients without evidence of adverse event. Among patients with evidence of adverse events, 77.5% were below 65 years of age, 19.6% were between 65–74 years, and 2.9% were ≥75 years. In comparison, among patients without adverse events, 65.2% were below 65 years of age, 26.1% were between 65–74 years, while 8.7% were ≥75 years. The Charlson index was similar (9; SD = 1.7) in both groups, indicating a comparable level of co-morbidities across the two groups.

Total medical-care costs during the 30-day post-event period, excluding the cost of targeted therapy for mRCC, are reported in Tables 3 and 4. Mean time to first adverse event ranged from 169 days for abdominal pain to 326 days for diarrhea; mean time to first event of any type was 202 days (Table 3). Mean (95% CI) post-event costs over 30 days among patients with evidence of adverse events was $10,101 (95% CI = $8462–$13,770); it was $4519 (95% CI = $2381–$6936) among those without any evidence of adverse events.

Table 3. Mean total cost (US$) of medical-care services (excluding cost of targeted therapy) during 30-day, post-event period* among patients with metastatic renal cell carcinoma receiving targeted therapy.

Adverse events	Mean time to event (days)	With adverse event				Without adverse event
		n	Mean	SD	95% CI	n	Mean	SD	95% CI
Abdominal pain	170	267	12,236	19,726	6361–16,208	221	3482	6281	1893–5191
Fatigue/asthenia	169	199	8415	15,844	5961–12,615	219	4214	6319	2236–4612
Dyspnea	326	97	6561	6983	2392–7492	308	3714	6846	2681–5119
Back pain	220	236	14,348	22,342	7458–24,326	178	3593	7127	1897–4382
Nausea and/or vomiting	228	186	12,386	18,008	6672–21,980	228	4746	9167	2437–5729
Extremity pain	229	252	7902	9782	4921–11,324	153	3340	4574	3146–4273
Diarrhea	217	18	11,629	17,532	−2854–25,384	351	4622	9413	2601–6397
Proteinuria	298	11	2827	1284	−1292–7361	372	5173	11,706	3646–7195
Hand–foot syndrome	277	194	10,192	15,014	5304–17,742	171	3244	4889	1915–3962
Hypertension	190	11	3207	1545	1314–6299	385	5150	10,618	4150–6918
Neutropenia	309	19	9894	12,156	2505–17,292	178	4378	8221	2115–6841
Lymphopenia	202	418	10,101	16,290	8462–13,770	115	4519	9519	2381–6936

*For patients with adverse events, the 30-day post event period started on the first date of the adverse event. For patients without the particular adverse event under consideration, the period started on the mean event day among patients with adverse events. Patients without adverse event under consideration with a follow-up period ending prior to the mean event day were excluded.

Table 4. Estimated increase in mean total cost of medical-care services (excluding cost of study medications) during 30-day post-event period* due to evidence of treatment-related adverse events among patients with advanced renal cell carcinoma receiving sunitinib, sorafenib, bevacizumba, or pazopanib (estimates adjustment for differences in baseline characteristics).

Adverse events	Estimated increase in cost associated with adverse event $, (95% CI)
	$	95% CI
Abdominal pain	9462	3818–18,927
Fatigue/asthenia	6404	2294–15,473
Dyspnea	4847	1292–16,948
Back pain	11,355	5195–22,618
Nausea and/or vomiting	10,040	3817–19,937
Extremity pain	8562	3971–16,295
Diarrhea	12,607	2849–68,419
Proteinuria	5853	−1194–33,947
Hand–foot syndrome	9057	1899–21,947
Hypertension	1920	1130–14,196
Neutropenia	3115	−2195–28,632
Lymphopenia	9807	4386–22,947

*For patients with adverse events, the 30-day post event period started on the first date of the adverse event. For patients without adverse events, the period started on the mean event day among patients with adverse events. Patients without adverse events with a follow-up period ending prior to the mean event day were excluded.

In GLM models, costs associated with each adverse event during the 30-day post-event period were estimated based on the parameter estimate for the binary event flag and mean total cost among patients who did not have evidence of the adverse event (Table 4). The model-based estimate of the increase (95% CI) in post-event costs over 30 days among patients with evidence of any adverse events is $9807 (95% CI = $4386–$22,947). For all types of adverse events examined, the estimated difference in costs between evented and non-evented patients was positive; the 95% CI did not include zero for hypertension and proteinurea.

Discussion

Although cytokine therapy (interleukin-2 [IL-2] and/or interferon-α [IFN]) has been the standard of care for mRCC treatment for the past two decades, advances in understanding of the biology and genetics of RCC have led to novel targeted approaches to treatment4. These targeted agents exhibit different tolerability profiles, associated with different targets that they inhibit7. Tolerability is an important consideration when considering first-line treatment options for patients with metastatic RCC7. Targeted agents related adverse events are likely associated with additional treatments used to manage them and/or hospitalizations, thereby increasing cost of treating a patient with mRCC.

This study estimated the costs associated with most common (frequency > 5%) moderate-to-severe adverse events (Grade 3 or 4) reported in randomized clinical trials of sunitinib, sorafenib, bevacizumab, or pazopanib in patients with mRCC receiving targeted agents. The Optum Impact National Manage Care Benchmark Database, a large administrative claims dataset containing facility, professional-service and retail (i.e., outpatient) pharmacy claims from more than 46 health plans throughout the US was utilized in this study to estimate these costs.

Among the 533 patients in the final study sample, ∼50% had healthcare encounters with ICD-9-CM diagnosis code of abdominal pain, followed by fatigue/asthnea (47.3%) and dyspnea (44.3%). Although baseline characteristics of patients with and without adverse events were similar, mean (95% CI) total healthcare costs (excluding the cost of targeted therapies) were considerably higher over 30 days in patients with evidence of adverse events—$10,101 (95% CI = $8462–$13,770) compared to $4519 (95% CI = $2381–$6936) for those who did not experience these events. On a covariate-adjusted basis, the estimated increase in cost (95% CI) was substantial and was estimated to be $9807 (95% CI = $4386–$22,947).

A recently published study8 was conducted using the same methodology as was used in the current study, albeit in a different database (PharMetrics database), to ascertain the 30 day total costs associated with selected AEs. The study findings were similar to what was observed in the current study. The PharMetrics study reported the mean (SD) total costs of care during the 30-day, post-event period in patients with adverse events were $12,177 ($19,621) vs $4070 ($8142) in those patients who did not have the evidence of adverse events. Adjusting for differences in baseline characteristics, the estimated cost difference was reported to be $11,373 (95% CI = $5286–$21,419).

Another study9 investigated costs associated with AEs in mRCC patients using the Surveillance, Epidemiology and End Results (SEER) Medicare database. Although the patient population was older (≥65 years of age) in the SEER Medicare database study, the study findings were similar to the current study. Mean (SD) total costs over 30 days were found to be substantially higher among patients with AE [$13,944 ($14,529)] compared to those without evidence of these AEs [$1878 ($52,640)]. When adjusted for differences in baseline characteristics, the mean (95% confidence interval) difference in costs between evented and non-evented patients was $12,410 ($9217–$16,522).

The findings from the current study need to be interpreted in context of its limitations. As with all studies based on health insurance claims data, errors of commission and omission are always a concern, especially as they may affect case-finding methods that rely on ICD-9-CM diagnosis and procedure codes—in the case of the present study, the identification of study subjects with mRCC as well as those who experienced the adverse events of interest. The impact of such errors in coding on study findings is unknown. Moreover, since a unique ICD-9-CM diagnosis code for mRCC does not exist, study subjects had to be selected using the diagnosis codes for kidney cancer (189.0) and malignant neoplasm of the renal pelvis (189.0), followed by exclusion of patients with evidence of TCC based on differences in the diagnosis and treatment of these two diseases. Although this approach may be a reasonable approach, more studies will be needed to understand the sensitive and specific of the administrative coding based algorithm. Moreover, the method that was used to identify patients who experienced the adverse events of interest (i.e., those associated with targeted therapies for mRCC) relied on healthcare encounters with particular ICD-9-CM diagnosis codes. While only ∼1% of patients had healthcare encounters for hypertension, rates reported in package inserts (Grade 3/4 severity) were 13% for sunitinib, 6% for bevacizumab, and 3% for sorafenib. In the current study the costs associated with adverse events were limited to the 30-day period following first mention of the event. To the extent that costs remain elevated beyond 30 days since the occurrence of an adverse event, this study may have under-estimated the costs of adverse events. Lastly, although the study sample was drawn using a large administrative claims database containing claims from health plans throughout the US, findings from this study may not be generalizable to other groups of patients with mRCC.

Conclusion

In summary, notwithstanding these limitations, the study findings indicate that the costs associated with adverse events are substantial in patients with mRCC receiving targeted therapy. Efforts to prevent and/or better manage these events may reduce the economic burden of this disease. Additional research is also warranted to substantiate the study findings.

Transparency

Declaration of funding

This study was funded by GlaxoSmithKline. R Borker was employed by GlaxoSmithKline (which manufactures and commercializes pazopanib) when the study was conducted and the manuscript was developed.

Declaration of financial/other relationships

R. Borker owns GSK stocks or stock options. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

R. Borker would like to thank Larry Knapton for his input during the study research and manuscript development.