Document Type


Date of Degree

Fall 2010

Degree Name

PhD (Doctor of Philosophy)

Degree In


First Advisor

Brooks, John M.


The objective of this research was to estimate prophylactic G-CSF effectiveness among patients on the extensive margin, whose treatment decisions are most likely to be affected by policy changes intended to alter the G-CSF treatment rate.

Using the national Surveillance, Epidemiology and End Results (SEER)-Medicare Linked Database, we studied patients 66 years or older diagnosed with Non-Hodgkin's Lymphoma (NHL) and on anthracycline in one of the 13 SEER registry areas from 1994-2002. Prophylactic G-CSF use was designated if a patient had a G-CSF claim within the first five days of the first chemotherapy cycle. The dependent variable of neutropenia hospitalization (NH) was identified within 6 months of diagnosis and was further specified as NH incidence within 21, 42, 63 and 126 days after anthracycline initiation in sensitivity analyses. Multivariate regression estimates were used to examine whether treated patients actually benefited from G-CSF. Instrumental variable estimates using local area prophylactic G-CSF treatment rates as instruments were used to estimate whether increases in the G-CSF utilization rate could lead to further reductions in the rate of neutropenia hospitalization.

We found only 9.85% of study patients had prophylactic G-CSF. After adjustment for patient demographic and clinical risk factors, multiple regressions indicated prophylactic G-CSF significantly reduced NH events within six months of diagnosis date for the patients who received G-CSF (OR=0.595, 95% CI=0.384-0.922). This estimate of G-CSF's effect may be biased low from the true values of Average Treatment effect on the Treated (ATT) because patients may be selected into treatment based on unobservable risk factors. Chow F-statistics showed our instrumental variable of area prophylactic G-CSF treatment rate described a statistically significant portion in the variation of G-CSF use (F=60.46, P<0.0001). In the base-case analysis, we found instrumental variable estimates of prophylactic G-CSF benefits within 6 months of diagnosis date among marginal patients. The estimated benefits varied with different instrument specifications, regardless of the level of statistical significance. In the sensitivity analyses, the exclusion criteria for the inability to calculate an area reimbursement variable in base-case analysis were removed. We found substantial G-CSF benefits available within first cycle of chemotherapy among marginal patients and the instrumental variable estimates were statistically significant.

Among elderly NHL patients on anthracycline-based chemotherapy, our multiple regression estimates suggest that patients treated with prophylactic G-CSF reduced their neutropenia risk within six months of diagnosis date. The effect of prophylactic G-CSF on neutropenia hospitalization among marginal patients whose choices varied with local area G-CSF treatment rate was negative. Substantial G-CSF treatment benefits within the first cycle of chemotherapy were available for patients on the extensive margin. Higher treatment rates may be guaranteed to improve patient short-term benefits from G-CSF.


xi, 154 pages


Includes bibliographical references (pages 143-154).


Copyright 2010 Xiaoyun Pan