Overview Amy Davidoff, Ph.D., M.S. Associate Professor Pharmaceutical Health Services Research Department, Peter Lamy Center on Drug Therapy and Aging University of Maryland School of Pharmacy Clinical Background Treatment of MDS Framing the MDS Question within CER Specific Aims Retrospective observational studies using secondary data Project Data & Methods Grantsmanship May 24, 2010 Study team UMB Amy Davidoff, PI PHSR/SOP Sheila Weiss Smith PHSR/SOP Maria Baer UMGCC/SOM Pharmaceutical Research Computing - PHSR Xuehua Ke & Franklin Hendrick - PHSR JHMI Steve Gore - JHSKCCC/SOM Myelodysplastic Syndromes (MDS) Hematopoietic stem cell neoplasms characterized by ineffective hematopoiesis. Median diagnosis age is 71 Incidence rate increases with age Peak incidence at 36.3 per 100,000 (age 70-89) Annual US incidence >10,000 cases Common Clinical Features of MDS Heterogeneity of disease characteristics, natural history, progression MDS may worsen, progress to acute myeloid leukemia; death Symptomatic anemia common => diminished quality of life Treatment of Low Risk MDS Therapeutic focus = palliation of symptomatic anemia Erythropoietic stimulating agents (ESAs) Overall response rate = 20%, targeting ti increases => 50% Median 2 year response GCSF may augment response ESA costly, injected in physician office ESA associated with adverse events, progression in solid tumors 1
Alternative therapy is RBC transfusion High cost Lengthy time to administer Risk of alloimmunization, transfusion reaction, iron overload and dtransfusion-associated i t dinfection Symptom relief oscillates CER Comparison for low risk MDS MDS diagnosis w/ anemia Serum epo levell ESA Assess response 5 Aza use RBC transfusion = default tx if ESA fails RBC use Transfusion dependence Newer therapies address broader spectrum of MDS risk groups Azacytosine nucleoside analogues 5-azacitidine (2004), decitabine (2006) palliate anemia in 40 50% of low-risk patients who did not respond or stopped responding to ESA. treatments costly, associated with significant complications, relatively brief response duration Lenalidomide (2005) Oral medication Patients with deletions of chromosome 5q31.1 improves anemia in 25% of low-risk MDS patients Preliminary Data: MDS Treatment Patterns, Medicare w/ Private Ins, 2007 100% 90% 80% 70% 60% 50% 40% 30% 20% 0% 100% 28% 6% 65% 74% 65% 25% 9% 32% 7% 4% 3% % Cases ESA GCSF 5AC, DAC Lenalid RBC All Low Risk High Risk 7% 12% 19% Framing as a CE Problem Two key treatment arms for low risk MDS Uncertainty about clinical benefit Both palliate anemia, ESA response limited ESA may delay disease progression Uncertainty about risk of adverse events RBC known ESA FDA warning related to solid tumors, risk in MDS unclear Both high cost, concern about inappropriate use CMS coverage decisions Potential for RCT in US very limited CE: Erythropoietic Stimulating Agents in Treatment of MDS Specific Aims: Describe MDS pt characteristics, treatment patterns, outcomes Determine role of pt sociodemographic, health status, provider characteristics on treatment Examine effect of ESA exposure on health outcomes, including time to transfusion (RBC) dependence, 5AC/DAC use, thrombosis, other adverse events disease progression, Death 2
Specific Aims (Cont.) Measure costs associated with ESA use, other treatments from Medicare s perspective. Identify costs associated with continued ESA use after presumptive treatment failure Describe change in patterns of ESA use with addition of new oral therapies for patients with MDS in 2006; release of the FDA advisory and Medicare NCD in 2007. Estimate change in MDS related treatment costs associated with measured changes in ESA use Methods Retrospective observational study using secondary data Characteristics of classical RCTs Focus on development of new therapies, new applications under ideal conditions Careful selection of subjects into study Commonly exclude older patients, with comorbid conditions Random assignment to tx arms Controlled delivery of intervention Measurement of pre-defined outcomes Strong internal validity Ability to demonstrate causal relationships Clinical trials have limitations High cost => may limit sample size, statistical power Difficult to examine unexpected clinical outcomes Limited external validity/generalizability Can t examine treatment patterns Retrospective Observational Studies Using 2 o Data Common sources = survey, insurance enrollment, claims Examine treatment patterns and clinical outcomes Include patients generally excluded from clinical trials Compare treatments or regimens not otherwise approved for study in patient population Key limitations of using secondary data Measurement of key variables Original purpose not research Need to validate measurement, assess error Treatment selection unobserved prognostic factors can confound relationship between treatment and outcomes 3
Observed Determinants Treatment Unobserved Prognostic Characteristics Problem of Selection Bias, Confounding in Observational Studies Outcome Methods to Address Treatment Selection Methods Improved covariate control Propensity score analysis Sensitivity analysis Instrumental variable analysis Propensity Score Analysis Addresses Selection on Observables Balances characteristics in treatment arms Estimate multivariate model Treatment = f(determinants of outcome, both treatment and outcome) Generate propensity score = predicted treatment probability Use propensity score to Create matched sample Stratify Weight observations Estimate effect of treatment on outcomes Limitation doesn t address confounding, selection based on unobserved factors Sensitivity analysis Instrumental Variable Analysis (IVA) Addresses potential confounding associated with unobserved factors Uses instrument strongly correlated with treatment, but not independently associated with outcome Conceptually - uses alternative comparison based on value of instrument, not observed treatment Use of IVA in MDS project Supplemental Insurance Correlates of supplemental insurance MDS Treatment Outcomes ESA Use in MDS: Data & sample selection Three key sources SEER-Medicare Newly reported MDS cases 2001-2005 (n=~10,000) Medicare claims 2000-2007 Historical cohort of untreated MDS patients Untreated controls from clinical trials (n=850) Medicare enrollment & claims (including Part D) Select MDS cohort using first report of ICD-9 diagnosis codes Newly diagnosed cases 2006-2008, claims => 2009 4
SEER-Medicare combines several sources of secondary data SEER = Surveillance, Epidemiology, and End Results Aggregates data from 16 regional tumor registries For each primary diagnosis SEER reports type of cancer histology, behavior stage (location, size, extent, nodal involvement, metastases) date of diagnosis initial surgical or XRT therapy date of death No data on disease progression, chemotherapy use Limited demographic data SEER linked to Medicare enrollment files, Census data SEER registry data linked for Medicare beneficiaries Medicare enrollment data Census tract level demographics from Census E.g. Income, education Medicare Part A and Part B claims also linked Key measures from SEER, Medicare claims MDS classification =~ risk group Pre-dx comorbidities, performance (functional) status Demographics - Patient, census track level Physician characteristics Treatment ESA, RBC, GCSF, 5AZA/DAC ESA ever, exposure Outcomes Measures Adverse events Those associated with RBC transfusion Associated with ESA use in solid tumors Disease progression Time to transfusion dependence Infections AML Death Cost of treatment Plan to estimate effect of treatment Within SEER-Medicare Examine characteristics of treated vs untreated, determinants of treatment Propensity score analysis to generate matched cohort Logistic models to estimate effect of treatment on discrete outcomes; consider IVA Non-parametric survival models to estimate effect of exposure on time to event Discrete exposure Cumulative exposure Complex combinations Match treated with historic controls Grantsmanship Problem relevant to clinicians, patients, FDA, CMS State of the art analysis plans, but with flexibility, logical progression E ll t biliti f id t d Excellent capabilities for rapid turnaround Study team multi-disciplinary, institutional, cooperative group 5
Grantsmanship (cont.) Demonstrated linkages to users for rapid dissemination Letters of support from 3 advocacy organizations Expert relationship with CMS Advocated with SRO to have MDS expertise on review panel 6