Stage Migration, Selection Bias, and Survival Associated With the Adoption of Positron Emission Tomography Among Medicare Beneficiaries With Non–Small-Cell Lung Cancer, 1998-2003

Data are from the Surveillance, Epidemiology, and End Results (SEER) –Medicare linked data. SEER-Medicare is a collaborative effort between the National Cancer Institute and the Centers for Medicare & Medicaid Services that links routinely collected population-based data from SEER cancer registries to Medicare administrative claims data. SEER data include demographic and incident cancer characteristics, including grade and stage, for approximately 25% of the US population with cancer. Medicare provides health insurance for 97% of people age 65 years and older in the United States, and these data reflect health care services used and comorbid health conditions.¹¹ SEER-Medicare data have been used to examine factors that affect cancer care quality, including sociodemographic characteristics, physician and hospital characteristics, surgery, chemotherapy, radiation, comorbid conditions, complications, screening, relapse, and costs.^12–22 This study was approved by the Office of Human Research Ethics at the University of North Carolina at Chapel Hill.

From the 12 SEER registries that were continuously active from 1998 onward, we included all patients who had a diagnosis of cancer of the lung and bronchus with microscopically confirmed NSCLC histology between 1998 and 2003, were age ≥ 66 years at diagnosis, and had Medicare Part A and Part B coverage without participating in a health maintenance organization or Medicare Part C for the year before and the year after diagnosis or until death. We excluded patients who were diagnosed at autopsy or death or had another diagnosis of malignancy in the year before the NSCLC diagnosis. We excluded patients who did not survive at least 2 months from diagnosis to exclude patients with poor performance status for whom PET use was thought to be unlikely. To identify patients likely to have complete Medicare claims related to NSCLC management, patients were required to have a primary diagnosis of lung cancer on an inpatient, outpatient, or carrier-based Medicare claim within 2 months before and 4 months after the SEER-reported month of diagnosis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 162.2-162.9, 231.2).

The primary outcomes were stage at diagnosis and survival at 2 years. Cancer stage was ascertained from SEER data by using the American Joint Committee on Cancer (AJCC) Cancer Staging Manual, Third Edition, from 1998 through 2003, during which staging criteria and stage-related data collection in the SEER registry remained unchanged.²³ Advanced disease was defined as stage IIIB or IV (incurable).²⁴ Survival at 2 years was obtained from SEER. We chose 2 years to provide a clinically meaningful end point that would also allow us to detect changes in 2-year survival across all disease stages (ranges from 5% to 75%).²⁵ We examined survival at 4 years as an additional sensitivity analysis to better detect changes in survival of patients with early-stage disease.

Use of PET was assessed by using outpatient and carrier claims in the period 2 months before and 4 months after the SEER diagnosis.²³ To control for patient distance to a PET facility, we determined the straight-line distance between ZIP codes of patients and the closest location of PET administration at the time of diagnosis.^26,27 We dichotomized as less than 40 miles or ≥ 40 miles from the facility.

All remaining variables were obtained from the SEER Patient Entitlement and Diagnosis Summary File. Demographic variables included age, sex, race, ethnicity, marital status, and local census tract characteristics (metropolitan urban or rural status; proportion of the population not finishing high school; proportion below the poverty line; and proportion with black race). Histology of NSCLC was classified as adenocarcinoma, large-cell carcinoma, squamous cell carcinoma, or NSCLC otherwise undifferentiated by using the International Classification of Diseases for Oncology, Third Edition (ICD-O-3) code for the SEER diagnosis. The 12 SEER registries were grouped by census region: Northeast (Connecticut), Midwest (Detroit, Iowa), South (Atlanta, rural Georgia), and West (San Francisco, Hawaii, New Mexico, Seattle, Utah, San Jose, and Los Angeles).

We used categorical specifications of study variables because they changed more predictably than group averages as members of one group migrated to another.²⁸ To study changes in staging and survival over time, we grouped patients by year of NSCLC diagnosis into cohorts representing the initial phase of PET adoption (1998–1999) and the post-PET phase of adoption (2002–2003). We compared baseline patient characteristics, frequency of PET use, survival, and stage distribution between the initial and post-PET cohorts by using χ² tests to assess differences over time. We plotted overall trends in stage distribution and survival by year of diagnosis from 1998 through 2003 among Medicare beneficiaries with NSCLC.

We used bivariate analyses to examine associations of PET with stage migration and survival. We set the significance at P ≤ .001 to address multiple hypothesis testing. We used SAS version 9.2 (SAS Institute, Cary, NC) for all statistical analyses.

Our study has several limitations. First, the relationships between PET use, stage, selection bias, and survival are complex and, in many cases, bidirectionally causal. A patient's pre-PET stage and chance of survival affects whether a patient receives PET, and receipt of PET affects the patient's post-PET stage and future management. Because of the risk of misinterpretation of conventional multivariable modeling in this context, we focused on aggregate temporal trends and did not present the complex and potentially misleading results of investigational regression analyses. Second, we investigated whether PET use was associated with increased survival and did not examine other potential values of PET in NSCLC, such as reduction of futile thoracotomies and the potential to improve quality of life or result in cost savings. Third, this study was a retrospective, claims-based analysis. Only PET scans paid for by Medicare could be detected in the analysis. To minimize the proportion of missed claims, all analyses were limited to Medicare beneficiaries with both Medicare Part A and Part B coverage and no enrollment in managed care or Medicare Part C for the 12 months before and after diagnosis. Fourth, patients in the SEER registry are more likely to be nonwhite, to live in areas with less poverty, and to live in urban areas,¹¹ which may limit the generalizability of the findings. Fifth, during the study period, disease stage was based on SEER data obtained over 4 months or until first surgery. In 2004, data collection for SEER changed to the collaborative staging system. It is unclear how our results would differ with this newer approach. Finally, we cannot exclude the possibility that survival or stage may have been affected by unexplored changes in patient care. In particular, increased use of chemotherapy and potential changes in the use of early palliative care may have affected survival. A recent randomized trial³⁰ demonstrated that early palliative care may confer a survival advantage, suggesting the possibility that changes in early palliative care during the study period could have influenced observed trends in NSCLC survival.

In conclusion, overall survival of Medicare beneficiaries with NSCLC remained unchanged after widespread adoption of PET. Previous reports of an association between PET use and increased survival among patients with NSCLC during this period may be attributable to a combination of stage migration and preferential administration of PET to patients with less advanced disease. The ability of PET to affect patient management, health care resource use, and costs remains an important area of research that may change as new treatments become available. Emerging screening technologies and therapies should be rigorously evaluated so that physicians, policy makers, insurers, and patients can make informed decisions about health care.