Comparative Effectiveness & Technology Assessment Call for Papers Session Comparative Effectiveness Research: The Increasing Elegance of ‘‘Coke vs. Pepsi’’ Studies Chair: Tanisha Carino, Ph.D. Tuesday, June 5 • 9:00 a.m.-10:30 a.m. Improving HIV Screening with Nurse Rapid Testing and Streamlined Counseling Henry Anaya, Ph.D., Steven M. Asch, M.D., M.P.H., Tuyen Hoang, Ph.D., Matthew Goetz, M.D., Allen Gifford, M.D., Candice Bowman, Ph.D. Presented By: Henry Anaya, Ph.D., Research Scientist, U.S. Department of Veteran's Affairs, 11301 Wilshire Boulevard, 111G, Los Angeles, CA 90073, Phone: (310) 478-3711, Email: henry.anaya@va.gov Research Objective: Testing for HIV has been shown to be cost effective in unselected general medical populations, yet rates of testing among those at risk remain far below optimal, even among those with regular primary care. The specific aims of this project are: oTo determine whether nurse-based referral for traditional HIV testing and counseling will improve screening rates compared to current testing procedures. oTo determine whether nurse-based rapid testing with streamlined counseling improves screening rates more than nurse-based referral for traditional testing and counseling alone. Study Design: A parallel-group, controlled study was conducted in the primary/urgent care clinics of the West Los Angeles VA. Eligibility was based on same-day appointment; age (18-65); no prior HIV test in past year; unknown HIV status. One hundred sixty six patients were randomized to one of three screening models: Model A: patients urged to discuss testing with their physician (control). Model B: nurses offered traditional counseling/testing. Model C; nurses offered streamlined counseling/rapid testing. Interventions were performed by nurses in addition to their regular clinic duties. Population Studied: honorably discharged veterans between the ages of 18-65 with unknown HIV status Principle Findings: Model A: 22 patients (40.7%) had test ordered; Model B: 48 (84.2%) had test ordered; Model C: 51 (92.7%) had test ordered. Of 22 patients in Model A with a test order, 9 (40.9%) received results; of 48 patients in Model B with test order, 25 (52.1%) received results; of 51 patients in Model C with test order, 46 (90.2%) received results. Conclusions: Results show that both interventional models will likely result in higher screening rates than traditional HIV testing models in primary care. Implications for Policy, Practice or Delivery: HIV rapid testing has been shown to be an effective means by which to convey results to patients, which is especially salient given the approximately 300,000 persons in the US alone who are unaware of their HIV-positive status. Increased rates of testing could lead to earlier identification of disease, increased treatment and reduced morbidity and mortality. Reduced intensity of counseling might free staff resources. As the VA is the largest HIV care provider in the US, it would be beneficial for policymakers to use this project and associated findings as a model when considering implementing rapid testing on a regular basis. Funding Source: VA Constructive Technology Assessment of Microarray Testing in Breast Cancer Treatment Willem Van Harten, M.D., Ph.D., Jolien M. Bueno-deMesquita, M.D., Valesca P. Rètel, M.Sc., Kim Karsenberg, M.Sc., Marjan M. Hummel, Ph.D. Presented By: Willem Van Harten, M.D., Ph.D., Member Ex. Board of Directors and Prof. Quality Management of Health Care Technology, Organisation & Management, School of Management and Governance, The Netherlands Cancer Institute and University of Twente, Plesmanlaan 121, Amsterdam, 1066 CX, Netherlands, Phone: 0031 20 5122860, Fax: 0031 20 6691449, Email: w.v.harten@nki.nl Research Objective: Performing technology assessment in an early stage of the controlled implementation of the prognostic 70-gene signature, a genomic test using microrrays, in the treatment of node-negative breast cancer patients. Study Design: As the technology was in its earliest stage of clinical implementation the Constructive Technology Assessment approach was chosen. This method is related to theories on Technology Dynamics and was developed in the field of Public Policy. All aspects as defined by the Institute of Medicine as well as technology related issues, such as juridical aspects, were covered. Method used: -Documentation analysis of internal process dynamics of the organisations before and after introduction. -Multidisciplinary team interviews before and after introduction. -Patient consultation recordings and patient questionnaires. -Descriptive registration and comparison of prognosis assessment using traditional clinical. guidelines and the genomic 70-gene microarray test. Scenario drafting and revision. Population Studied: In this implementation study, 812 eligible patients from 16 Dutch hospitals participated and 425 70-gene microarray test were performed. Principle Findings: - Introducing and implementing this new technology in clinical practice took on average about 6 months per participating hospital.- Especially pathologists had to change their working routine as they were requested to process the tissue directly after surgical removal of the tissue.Health care professionals had to take more than average time than before to properly explain to the patient the pro’s and con’s of this test.- In 30% of the cases the patients’ prognosis based on clinical guidelines was discordant with the prognosis based on the 70-gene microarray test. - In general, patients were accepting the specialists’ treatment advice. This was even the case if the patients’ prognoses based on clinical guidelines was discordant with genomic prognosis based on the 70-gene microarray test. - Scenario development proved feasible and provided options in the early stage that were considered unlikely by professionals, but nevertheless became reality. Conclusions: - The implementation of new genomic tests like the 70-gene microarray test in node-negative breast cancer patients is a complex process. - Constructed Technology Assessment is a promising method to analyse technologies in their early stage of development and implementation that are introduced in a controlled way. - Health care professionals have to anticipate to 30% discordance between prognosis assessed by traditional clinical guidelines versus genomic testing. Implications for Policy, Practice or Delivery: - The controlled introduction of promising genomic tests is feasible. - Constructive Technology Assessment is a promising broad assessment method for the implementation of new technologies in an early stage of their development. - Based on the experiences of this controlled implementation trial, a large European randomised trial (MINDACT-trial) was designed and will start in January 2007. Looking for Modifiers of Treatment Effects in the General Medical Literature: Room for Improvement Nicole Bloser, M.H.A., M.P.H., Naihua Duan, Ph.D., Diana Liao, M.P.H., Elizabeth Yakes, M.S., Kiavash Nikkhou, B.S., Richard L. Kravitz, M.D., M.S.P.H. Presented By: Nicole Bloser, M.H.A., M.P.H., Graduate Student Researcher, University of California, Davis, 2103 Stockton Boulevard, Grange Building, Suite 2224, Sacramento, CA 95817, Phone: (916) 734-2399, Fax: (916) 734-8731, Email: nrbloser@ucdavis.edu Research Objective: Randomized controlled trials (RCTs) generate average treatment effects, but patients want to know which treatments will work for them. Individualizing care for the complex patient requires knowledge of treatment impact in similar individuals or subgroups, which in turn depends on identifying moderators of treatment effects (MTEs). In an effort to avoid the appearance of ‘‘data dredging,’’ clinical investigators may be missing opportunities to explore MTEs, thus slowing accrual of evidence for treating ‘‘patients like me.’’ This study was undertaken to determine current practice in evaluating MTEs and to elucidate trends. Study Design: We examined a probability sample of 227 RCTs. Articles were independently reviewed and coded by 2 investigators with adjudication by a third. Studies were classified as having: a) MTE analysis utilizing a formal test for heterogeneity or interaction; b) subgroup analysis only, involving no formal test for heterogeneity or interaction, or c) no subgroup or MTE analysis. Chi-square tests and multiple logistic regression analysis were used to identify study characteristics predictive of MTE reporting. Population Studied: 227 RCTs published in 5 journals (Ann Intern Med, BMJ, JAMA, Lancet, and NEJM) during odd numbered months of 1994, 1999, and 2004. Principle Findings: Of the 227 RCTs, 101 (44%) performed no subgroup or MTE analysis, 62 (27%) examined subgroups but without MTE analysis, and 64 (28%) performed MTE analysis. MTE analysis gained currency with time (18%, 29%, and 34% of studies in 1994, 1999, and 2004, respectively). Among the 64 studies reporting MTE analysis, major covariates examined included age (30%), sex (28%), study site or center (17%), and race/ethnicity (8%). Using multiple logistic regression to examine study year, journal, clinical condition, and sample size, only sample size was a significant predictor of whether MTE analysis was performed; comparing the top quintile of studies (median n=1649) to the bottom quintile (median n=36), the adjusted odds ratio was 4.9 (95% CI 1.6-15.1, p=.0045). However, MTE analysis was performed less than half the time (49%) even in the top quintile. Conclusions: Missed opportunities for MTE analysis abound. In the face of broad NIH mandates for inclusion of subjects by race/ethnicity, the low proportion of studies testing race/ethnicity as a treatment effect moderator is both puzzling and disappointing. Implications for Policy, Practice or Delivery: Accepting Kraemer et al.´s argument (JAMA, 2006) that exploratory moderator analysis is critical for designing appropriate future confirmatory studies, standards are needed to assure that exploratory moderator analysis and reporting become rigorous and routine. Such standards are essential for developing practice guidelines that are appropriate to the needs of the complex patient. Funding Source: Pfizer Inc. Decision Makers' Attitudes Toward Cost Effectiveness Analysis Shoshanna Sofaer, Dr.P.H., Stirling Bryan, Ph.D., Taryn Siegelberg, M.P.A. Presented By: Shoshanna Sofaer, Dr.P.H., Robert P. Luciano Professor of Health Care Policy, School of Public Affairs, Baruch College, One Bernard Baruch Way, Box D901, New York, NY 10011, Phone: (646) 660-6815, Email: shoshanna_sofaer@baruch.cuny.edu Research Objective: While cost-effectiveness analysis (CEA) is part of the policymaker’s toolkit in making insurance coverage decisions in many western nations, it is largely unused in the US. We have little empirical knowledge about what gives decision-makers pause about using CEA. Our research sought to understand receptiveness to CEA among those who influence coverage decisions, and how institutional constraints, individual values and methodological concerns shape decision-maker views. Study Design: This exploratory study collected data through six structured workshops in California. During these 3-4 hour workshops/focus groups, participants were asked to take on the role of ‘‘social decision-maker’’ addressing issues of concern to the Medicare program. CEA methods were explained; ethical/normative issues inherent in CEA were discussed; and participants prioritized 14 treatments for coverage in response to information from published CEA studies. At the end of each session, participants removed their social decision maker ‘‘hat’’ and discussed, from an organizational perspective, advantages and barriers to CEA. A pre-group survey addressed knowledge and attitudes to CEA and gave respondents an opportunity to prioritize the 14 treatments based on effectiveness information alone. A postgroup survey re-asked knowledge and attitude questions adding other questions about the workshop and CEA. Survey data were analyzed descriptively at both points in time. Changes over time were assessed statistically, as were changes in priorities following the presentation of costeffectiveness data. All sessions were audio-taped, transcribed, and coded and analyzed using NVivo software. Population Studied: Participants included senior leaders (both clinical and non-clinical) from different types of health insurance plans, private and public sector health care purchasers, disease management organizations, and state regulators of managed care plans. Principle Findings: In the post-workshop survey, over 90% of participants indicated that CEA should be used as an input into coverage decisions for Medicare and over 70% said it should be used in private insurance plans. Participants also identified reasons to avoid using CEA, including: fears of negative consumer perceptions; litigation risks; inadequate inhouse expertise; worries about biased studies; concerns about ethical issues; and a preference for reducing costs by reducing demand. Many noted that no single entity, particularly in the private sector, could ‘‘go it alone’’ in explicitly using CEA. Finally, when provided with cost-effectiveness information on a variety of condition-treatment pairs, participants changed their priorities to fund treatments with more favorable cost effectiveness ratios, and exclude treatments with higher ratios. Conclusions: Senior California decision-makers believe that CEA is an effective and promising tool to assist cost containment.. Some are not confident in their ability to acquire, assess and apply high quality CEA studies. Cost effectiveness data was influential in changing hypothetical decisions about coverage in the Medicare program. Implications for Policy, Practice or Delivery: There is a need to identify and motivate a visible policy leader in health care, such as the Medicare program, to move in the direction of making CEA studies an input into coverage decisions. Better promotion of the availability of high-quality CEA studies and more accessible information about the elements of such studies, is needed. Funding Source: California Health Care Foundation Using Observational Data to Extend the Results of a Randomized Controlled Trial: An Application to the HOPE Trial Paul Hebert, Ph.D., Mary Ann McLaughlin, M.D., M.P.H., Jodi M Casabianca, M.S., Anu Lala, M.D. Presented By: Paul Hebert, Ph.D., Assistant Professor, Health Policy, Department of Health Policy, Mount Sinai School of Medicine, One Gustave L Levy Place, Box 1077, New York, NY 10029-6574, Phone: (212) 659-9567, Fax: (212) 423-2998, Email: paul.hebert@mssm.edu Research Objective: The purpose of this study was to explore the use of observational data to extend the results of a randomized controlled trial (RCT). We considered the Heart Outcomes PrEvention (HOPE) RCT, which demonstrated broad clinical benefits of ramipril, an angiotensin converting enzyme inhibitor (ACEI), for patients with risk factors for cardiovascular disease. Study Design: We conducted a retrospective statistical analysis. We first attempted to recreate the findings of the HOPE RCT using observational data and instrumental variable (IV) models, to assess whether IV analyses can address the selection bias inherent to observational data. With the success of this analysis, we then applied these same methods to a sample of non-white patients. Non-white patients accounted for <5% of HOPE trial participants. We also assessed whether other ACEIs showed clinical benefits similar to ramipril. Data came from administrative databases for dually-eligible Medicare and Medicaid beneficiaries from California from 1996-99------the period during which HOPE was conducted. Medicaid claims for prescriptions identified patients taking antihypertensives in 1997. Diagnosis codes on Medicaid/Medicare claims identified risk factors in 1996. The outcome was the composite endpoint of all-cause mortality or hospitalization for stroke or myocardial infarction (similar to HOPE) over the 1997-99 period. Instrumental variables were managed care market penetration in the beneficiary’s county of residence, and the ratio of ramipril to other ACEI prescriptions filled in a beneficiary’s ZIP code in 1996. We estimated linear probability models (LPMs) and IV LPMs with instruments for ramipril use, both with robust standard errors. To assess whether similar clinical benefits could be attributed to other ACEIs, we identified patients who were prescribed other ACEIs (benezapril, captopril, and enalapril) in 1997 and estimated multivariate Cox models of the time to the composite endpoint as a function of the ACEI prescribed. Population Studied: Dually-eligible Medicaid/Medicare patients from California being treated with an antihypertensive. Principle Findings: We identified 108,209 patients who met the HOPE inclusion criteria, of whom 4,789 took ramipril in 1997. LPM on 56,549 White patients suggest a risk ratio (RR) of 0.842 (p<0.001) for ramipril use, which compared favorably to RR=0.75 in HOPE. IV LPMs models yielded RR=0.463 (p=0.032). Applying these same models to 51,660 non-White beneficiaries generated RRs of 0.992 (p=0.869) and 1.535 (p=0.127) for ramipril use from LPMs and IV LPMs, respectively. In Cox models for patients taking any ACEI in 1997, patients taking benezapril (n=2,839; hazard ratio=1.037, p=0.624) had similar risks of the composite endpoint compared to patients taking ramipril, while patients taking enalapril (n=4408; hr=1.17, p=0.018) or captopril (n=1469; hr=1.3, p<0.001) faired worse. Conclusions: IV models applied to observational data matched the results of the HOPE RCT. The same models suggested no benefit of ramipril for non-white beneficiaries. Ramipril and benezepril appeared superior to enalapril and captopril. Implications for Policy, Practice or Delivery: These results can help guide prescribing practices and have implications for proposed formulary policies that restrict access to benezapril and ramipril. These data also provide some support to the hypothesis of racial differences in cardiac drug effectiveness. Funding Source: NIH National Institutes of Diabetes Digestive and Kidney Disease (NIDDK)