Objective: To examine the effect of prescription step-therapyprograms in terms of plan-sponsor savings and member experienceat the point of service.
Study Design: Plan-sponsor savings were measured using aquasi-experimental, case-control design. Member experience withstep therapy was measured using a self-administered mailed survey.
Methods: A 20 000-member plan implemented 3 step therapyprograms in September 2002: proton pump inhibitors, selectiveserotonin reuptake inhibitors, and nonsteroidal anti-inflammatorydrugs. Pharmacy claims from September 1, 2001, through June 30,2003, were examined to compare changes in per-member-per-month(PMPM) net cost between the intervention group and arandom sample of members from commercial plans without thestep therapy programs. A mailed, self-administered survey wassent to members with a step edit from September 1, 2002 toDecember 31, 2002.
Results: The employer experienced a decrease of $0.83 in netcost after implementing step therapy, while the comparison grouphad an upward trend of $0.10 PMPM for these therapy classes.Member-reported outcomes indicated that approximately 30% ofpatients received a generic, 23% were granted a medical exceptionfor the brand, 17% received no medication, and 16% paid the fullretail price for the brand. If the pharmacist vs the patient contactedthe physician, members were 8 times more likely to receive a medicationcovered by the health plan (OR, 8.10; 95% CI, 2.94-22.33vs OR, 8.23; 95% CI, 3.11-21.93). Compared with those whoreceived first-line therapy, those who paid out of pocket for thebrand medication vs those who did not receive any medicationwere less likely to be satisfied with their pharmacy benefit (OR,0.25; 95% CI, 0.08-0.80 vs OR, 0.12; 95% CI, 0.04-0.41).
Conclusions: Step therapy produces significant drug savings.However, there appear to be opportunities to further members' andproviders' understanding of these programs.
(Am J Manag Care. 2004;10:457-464)
Faced with continued double-digit growth in prescriptiondrug costs,1,2 plan sponsors are continuallylooking for ways to promote cost-effective useof medications. The growing availability of generic alternativesin many therapy classes has created anunprecedented opportunity for plan sponsors to bettermanage pharmacy benefits to the advantage of plansponsors and their members.
One method of encouraging generic use is throughstep therapy. Step therapy is a pharmacy benefit programthat promotes cost-effective use by requiring atrial of a first-line medication, often a generic alternative,before coverage is granted for a more expensivesecond-line agent, typically a brand. With the recentavailability of generic alternatives in many therapyclasses, the use of step therapy has grown dramatically.3Generic alternatives available for nonsteroidal anti-inflammatorydrugs (NSAIDs), gastroprotective agents,and selective serotonin reuptake inhibitors (SSRIs) renderthese appropriate classes for step therapy. Steptherapy programs have been offered for other classes,such as angiotensin-converting enzyme inhibitors anddisease-modifying antirheumatic drugs.
Research supports the clinical appropriateness ofstep therapy programs. For example, the SSRI step-therapyprogram requires a trial of generic fluoxetinebefore coverage of brand SSRI step therapy. Supportingthis strategy are conclusions from randomized controlledtrials suggesting no difference in primary depressionoutcome measures across SSRIs.4-7 In addition,retrospective analyses using medical chart review oradministrative claims data have found no difference inswitch rates among SSRIs8 and no difference in depression-related outpatient and hospitalization costs basedon initial choice of SSRI.9,10 These findings, togetherwith the fact that the mean cost for a brand SSRI wasnearly double that of generic fluoxetine in 2002,2 supportan SSRI step-therapy policy with generic fluoxetineas first-line therapy.
Step therapy programs for NSAIDs require a trial oftraditional or nonselective NSAIDs (eg, ibuprofen andnaproxen) before granting coverage for the higher-cost,selective NSAIDs (ie, cyclooxygenase [COX] 2 agents).Research has shown that, in the management of acutepain and other conditions associated with pain, COX-2inhibitors and nonselective or traditional NSAIDs areequally effective at equipotent doses.11-22 AlthoughCOX-2 therapy has been shown to reduce the risk ofgastrointestinal adverse events,23,24 recent pharmacoeconomicanalysis suggests that COX-2 agents are notcost effective for the average-risk patient, having a costper quality-adjusted life-year gained of $275 809.25
The impetus for proton pump inhibitor (PPI) steptherapy is research showing that 30% to 70% of patientswith painful reflux symptoms do not have erosive gastrointestinalconditions for which PPIs are indicated.26Given that relief of heartburn symptoms occurs in up to70% of patients taking lower-cost histamine2 (H2) receptorantagonists,27 "stepping up" in those who have notachieved adequate symptom control has been shown tobe a cost-effective alternative.28
Given the more recent popularity of step therapy, itis not surprising that no research, to our knowledge, hasempirically examined the effect of step therapy programs.Therefore, the objective of this study was toexamine the economic effect from the plan sponsor'sperspective, as well as the member's experience withstep therapy programs. Some key questions includedthe following: (1) What are the savings associated witha step therapy program from the plan sponsor's perspective,taking into consideration the administrativecosts of the program? (2) How do members respond toa step therapy program, in terms of contacts withproviders and benefits managers and overall satisfaction?(3) What is the final outcome, in terms of medicationreceived, when members experience steptherapy at the point of service?
This research is not subject to Department of Healthand Human Services regulations and therefore isexempt from institutional review board approval(Â§46.101 of the Federal Policy for the Protection ofHuman Subjects, August 19, 1991). We follow the principlesoutlined in the Declaration of Helsinki and therecently approved Health Insurance Portability andAccountability Act regulations regarding use of personalhealth information for program evaluation.
On September 1, 2002, an employer located in theMidwest with approximately 20 000 enrollees anddependents implemented 3 step therapy programs:PPIs, NSAIDs, and SSRIs. The employer had a 20% coinsurancebenefit in place at the time of the study, and noother benefit design or clinical program changes weremade during the study. The employer did not communicatewith employees about the change before programimplementation because of the inability to target theinformation to those likely to experience the edit.
The step criteria were automated and administered atthe point of service by the employer's pharmacy benefitmanagement (PBM) company. For all 3 step programs,prior users of the medications, as evidenced by their prescriptionclaims history, were not subject to the steptherapy program (ie, they were grandfathered).Therapy-specific criteria for PPIs required patients to tryan H2 receptor antagonist before receiving coverage for aPPI. The NSAID program required previous trial of 2generic NSAIDs before receiving coverage for a brandNSAID, and the SSRI program required previous use offluoxetine or fluvoxamine maleate before coverage for abrand SSRI would be granted. The program criteria werecommunicated to pharmacies at the time of adjudication,including instructions to call the physician anddocumentation of the covered first-line medications.Medical exceptions could be granted for those patientswho had previously tried a generic or were already stabilizedon the brand but for whom the claim had notbeen captured by the PBM (eg, because the patient usedhis or her spouse's insurance). Medical exceptions couldalso be granted for other clinical reasons (eg, failure withfirst-line agents not captured in the pharmacy claimsdata; history of a gastrointestinal bleed, perforation, orobstruction [NSAID step therapy]; and erosive gastrointestinalconditions [PPI step therapy]). To request amedical exception for brand coverage, the physiciancould call or fax the PBM. The employer paid $20 foreach medical exception reviewed. No other programcosts were incurred by the employer.
Two data sources were used in the study, membersurvey data and pharmacy claims data. Pharmacyclaims from September 1, 2001, through June 30, 2003,for all members were examined to assess changes inper-member-per-month (PMPM) net cost (ie, ingredientcost plus dispensing fee, minus the member copay) forthe 3 therapy classes of interest (intervention group).To allow for comparison with plans that did not implementstep therapy, a random sample of members fromcommercial plans that did not have the 3 step therapyprograms during the study was selected (comparisongroup). This comparison group included approximately1.9 million members, representing 1021 different healthplans.
A mailed survey was sent in February 2003 to adultmembers who had received a step therapy edit betweenSeptember 1, 2002, and December 31, 2002. Memberswith more than 1 edit, including other edits such asrefill too soon, were excluded from the sampling frameto avoid possible confounding. A presurvey postcardmaking members aware of the survey was sent approximately3 days before the self-administered survey, anda $1 incentive was included with the survey.
The survey contained 22 questions based on sociodemographics(4 questions), satisfaction (3 questions),member's experience with the step therapy program(12 questions), outcome of step edit (1 question), and 2screener questions. Satisfaction with the pharmacybenefit was measured on a 5-point Likert scale (1, verysatisfied; 5, very dissatisfied). "Satisfied" was defined asa response of "1" or "2." Satisfaction with the medicationreceived was also measured. Questions were relatedto satisfaction with the member's pharmacy benefit,satisfaction with the medication received, and satisfactionwith the pharmacy at which most prescriptionswere filled. Data reported herein concern satisfactionwith the pharmacy benefit and medication received.Questions related to the member's experience with steptherapy ascertained what efforts were made to obtaincoverage and the member's understanding of theprocess.
Basic univariate and bivariate statistics were used todescribe the survey sample, to compare respondentsand nonrespondents, and to examine relationshipsbetween variables. Logistic regression analysis was usedto assess if any sociodemographic characteristics werepredictors of whether the patient contacted thephysician. Logistic regression analysis was alsoused to assess predictors of whether the patientreceived a medication subsidized by his or heremployer (ie, generic or medical exception forthe brand) vs no medication or a medication notsubsidized by the employer (over-the-counter[OTC] products, samples, etc). Logistic regressionanalysis was also used to assess what factorspredicted patient satisfaction with the pharmacybenefit. Finally, linear regression analysisassessed the immediate effect of the step therapyprogram on PMPM net cost, while adjusting forthe time trend and periodicity of the data. Thetime series consisted of 22 values for monthlycosts.
Interrupted time series showed a $0.29 PMPMdecrease in drug expenditures ( < .001, adjusted = 0.83) in the month following implementationof the step therapy program for NSAIDs(Figure 1). No statistically significant change inPMPM SSRI drug costs was observed, while PPIsshowed a $0.48 decrease in net drug cost ( <.05, adjusted = 0.42). There were no significanttime trends for any of the therapy classes.
Across all 3 therapy classes, there was an immediatedecrease of $0.93 in PMPM costs ( < .01, adjusted =0.66), representing a savings of 19% off net cost relativeto the mean monthly preperiod expenditures for these3 therapy classes (Figure 2). Plan costs for medicalexception review totaled approximately $23 000 duringthe study, or $0.10 PMPM (number of medical exemptioncalls times $20 per call, divided by the total membership, divided by 12). The comparison group had anincreasing trend of $0.10 PMPM for the 3 therapy classesduring the study, with no significant change inexpenditures in the month following step therapyimplementation for the intervention group ( < .001,adjusted = 0.88).
Between September 1, 2002, and December 31,2002, there were 874 members with 1 or more steptherapy edits. Of those, 217 had edits in more than 1therapy class, leaving 657 members eligible for the survey.Seven surveys were returned undeliverable, and212 completed surveys were returned, for a 33%response rate. Of the 212 returned surveys, 3 wereexcluded because of incomplete information, and 33members (16% of respondents) indicated they did notrecall the situation, leaving 176 useable responses forthe analysis.
Approximately 60% of respondents were female, and82% were 35 years of age or older (Table 1). The mostcommonly reported (45%) annual income level was$40 000 to $59 999. About 75% of respondents said theywere in excellent or good health. There was no statisticallysignificant difference (α = .05) in the mean age,sex, month of edit, therapy class, or medication received(ie, generic vs brand) between respondents and nonrespondents(data not shown).
Only 43% of respondents reported that the pharmacistcalled the physician after the step therapy edit tofacilitate a new prescription (Table 2), while 62% ofmembers contacted their physician. Nearly 76% ofrespondents indicated that they or their pharmacistcontacted the physician. Approximately 42% of respondentsreported contacting the PBM call center, and 14%called their employer's human resource office after thestep therapy edit. Logistic regression analysis showedthat no sociodemographic characteristics were significantlyrelated to whether the patient contacted thephysician (data not shown). When asked whether theirpharmacist or pharmacy staff told them why the drugwas not covered by their employer, 58% reported thatthey had been told, and 6% were not sure or did notremember.
The most common outcome (29%) from a step therapyedit was having the prescription switched to ageneric medication (Table 3). Approximately 23% ofrespondents reported getting a medical exception toreceive coverage for the brand medication, and another16% paid out of pocket for the brand medication. Nearly17% reported getting no medication, and about 10%received a sample or an OTC alternative. Finally, 5%could not remember or had another outcome (eg, theirspouse's insurance paid for the medication). No significantdifferences were seen across income categories inthe percentage receiving no medication or the percentagepaying out of pocket (data not shown).
The percentage paying out of pocket for the brand(Pearson χ22 = 9.8, < .01) and the percentage purchasingan OTC product (Pearson χ22 = 7.3, < .05)varied across therapy classes. The percentage payingout of pocket for PPIs was lower (5%), while the percentagegetting an OTC product was higher for PPIs(11%), relative to the other 2 classes.
Logistic regression analysis showed that the probabilityof receiving a medication subsidized by theemployer varied by therapy class (Table 4). First,patients with an SSRI step-therapy edit were more likelyto receive a medication subsidized by their employerthan those with an NSAID edit ( < .05). Second, thepharmacist calling the physician was associated with710% greater odds of receiving a covered medication (< .01), and the patient calling the physician increasedthe odds of receiving a medication covered by theiremployer to 720% ( < .01). The model classified 77% ofthe cases correctly.
Medication satisfaction was greater for brand users(including medical exception and out-of-pocket brandpayers) vs generic users at 95% vs 53% (Pearson χ21 = 28.0, < .001). Just over 50% of respondents reportedthey were satisfied with their overall pharmacy benefit.However, pharmacy benefit satisfaction varied based onthe medication received ( < .01). Controlling forsociodemographics, logistic regression analysis foundthat paying out of pocket for the brand and receiving nomedication were associated with significantly lowerpharmacy benefit satisfaction compared with those whoreceived a generic (odds ratio [OR], 0.25; 95% confidenceinterval [CI], 0.08-0.80 vs OR, 0.12; 95% CI, 0.04-0.41) (Table 5). The model classified 68% of the casescorrectly. In addition, compared with members withhousehold annual incomes less than $40 000, the oddsof being satisfied with their pharmacy benefit was 22%lower among those in the highest annual income category(≥$60 000).
To our knowledge, this is the first study to examineplan-sponsor savings and member effect of a step therapyprogram. The findings suggest that step therapy producessignificant savings for the employer, results inmultiple drug product selections (many of which are notcaptured in administrative pharmacy claims data), andis associated with lower pharmacy benefit satisfactionfor those who do not receive a medication subsidized bytheir health plan.
Limitations of the study should be considered. Thisstudy reflects the experience of 1 employer; while thereis no reason to believe that other employers would havedramatically different outcomes, some variation isexpected because of patient, provider, and other plan-specificfactors. Similarly, the sample size at the therapyclass level was small, making it impossible to reportmany results by therapy class and emphasizing theneed to view therapy class—level results with caution.
This analysis did not include evaluation of plan-sponsormedical claims costs. Research measuringthese costs, in addition to patient health—related outcomes,is critical to understanding the full economicand clinical consequences of these and other PBM programs.A portion of program savings was achieved inpart due to members paying out of pocket or obtainingmedication from sources not subsidized by the healthplan (ie, OTC products and samples). For those payingout of pocket, it is not known whether these were memberswho were unaware of the lower cost alternatives orwhose time was valued at a rate greater than the cost ofthe medication. Similarly unknown is whether thosewho did not obtain a medication had any negativehealth consequences.
Some respondents indicated that they didnot remember the step therapy edit. Thiscould be due to the time lapse between whenthe edit took place and the survey administration(although we saw no relationshipbetween date of edit and remembering theevent). More likely, it is a reflection of randompatient variability in remembering theevent and, to some extent, the fact thatpatients do not always pick up the prescriptionfrom the pharmacy.29
The step therapy program was associatedwith a mean decrease of $0.83 PMPM in drugcosts for this employer after factoring in theadministrative program costs. This figureunderestimates savings because, as seen inthe comparison group, costs for these therapyclasses were increasing at a rate of $0.10PMPM during the same period among planswithout step therapy. Therefore, savings at 10months following implementation approximated$1.83 PMPM ($0.83 + [$0.10 Ã— 10]months) or 38% of the total net cost for these3 classes. Comparisons with savings fromother step therapy programs are not available.Savings were not limited to the employer.
Plan-sponsor savings resulted from differentmedication alternatives after the steptherapy edit, some of which were appropriateand others that were not the intended outcome.It is assumed that physicians were prescribingthe generic alternative when theyconsidered it clinically appropriate, and wererequesting a medical exception when theybelieved the brand medication was medicallynecessary. For patients who paid full price forthe brand medication, it is unknown whetherthe patient simply chose to do so rather than to contactthe physician, or whether the patient did not understandthat alternative medications were covered.Similarly, those who received no medication or an OTCalternative could have done so because of the "hasslefactor," because they had a less severe condition forwhich an OTC alternative or no medication seemedappropriate, or for other reasons. Income did notappear to affect whether the patient received no medication,as there were no differences in the percentagereceiving no medication across income categories, andthe percentage paying out of pocket was higher amonglower-income groups.
The significant variation in medication receivedacross therapy classes (ie, SSRIs were associated withincreased likelihood of receiving a covered medication)could be due to greater clinical need for SSRIusers or fewer OTC alternatives relative toNSAIDs and PPIs. This could also explain thenonsignificant trend in PMPM costs for SSRImedication. It is not surprising that pharmacybenefit satisfaction was lower for patientspaying full price and for those who did notreceive a medication. To that end, efforts toincrease member understanding and satisfactionwith these programs are needed. Furtherresearch is needed to assess the reasons whypatients do not receive a medication coveredby their health plan, the extent to which theyare informed about their alternatives, theextent to which patients consult with theirphysician and pharmacist about this decision,and the clinical outcomes across the variousdrug agents received.
In addition, we believe this to be the firststudy to look at pharmacist and member contactsafter receiving a utilization managementedit. There are multiple explanations for whyfewer than half of the respondents said theirpharmacists contacted the physician. First,patients may not have necessarily known orremembered that the pharmacist contactedthe physician (reported by 10%). Second,there may be pharmacies in which the computersystem does not show the message fromthe PBM (eg, the message appears temporarily)and the pharmacist does not understandthe reason for coverage denial. It may be thatpharmacists consult with patients and thencall the physician if the patient says he or sheis willing to take the generic, or that pharmacistscall only during nonpeak hours whenthey have more time for such activities. Pharmacistsmay also request that the member shoulder the administrativeburden and follow up with his or her physicianor health plan with additional questions. As found inthis study, pharmacist or member contact with thephysician significantly increases the likelihood ofreceiving a medication subsidized by the employer.The hassle factor and administrative burden placed onpharmacists support the use of electronic prescribing,in which physicians are made aware of plan design featuresand can make the appropriate changes at thepoint of care.
This study demonstrated that step therapy programscan produce savings for plan sponsors but identifiedopportunities to further members' understanding ofthese programs and pharmacist intervention, potentiallyimproving drug selection and member satisfaction.More generally, this study highlights the importance ofongoing evaluation of the economic, clinical, andhumanistic effect of pharmacy benefit design tools,such as step therapy, to ensure that intended objectivesare being achieved without untoward consequences.
We thank Kathi Fairman, MA, for developing the survey, andJagat Sheth, PhD, for his statistical analysis.
From Outcomes Research, Express Scripts, Inc, Maryland Heights, Mo.This study was funded by Express Scripts, Inc.
Address correspondence to: Brenda R. Motheral, PhD, Outcomes Research, ExpressScripts, Inc, 13900 Riverport Drive, Maryland Heights, MO 63043.
Health Aff (Millwood)
1. Levit K, Smith C, Cowan C, Lazenby H, Sensenig A, Catlin A. Trends in UShealth care spending, 2001. . 2003;22(1):154-164.
2002 Express Scripts Drug Trend Report
2. Teitelbaum F, Martinez R, Parker A, Kolling B, Svirnovskiy Y, Peterson C.Drug therapy class review. In: . St Louis,Mo: Express Scripts Inc; 2003:68, 75, 80.
Pharmacy Benefit Guide
3. Motheral B, Teitelbaum F, Frear R. Plan design: a stepwise approach to trendmanagement. In: . St Louis, Mo: Express Scripts Inc;2003:84.
J Clin Psychiatry
4. Newhouse PA, Krishnan KRR, Doraiswamy PM, Richter EM, Batzar ED, ClaryCM. A double-blind comparison of sertraline and fluoxetine in depressed elderly outpatients.. 2000;61:559-568.
5. Kroenke K, West SL, Swindle R, et al. Similar effectiveness of paroxetine, fluoxetine,and sertraline in primary care. 2001;286:2947-2955.
J Affect Disord
6. Fava M, Rosenbaum JF, Hoog SL, Tepner RG, Kopp JB, Nilsson ME. Fluoxetineversus sertraline and paroxetine in major depression: tolerability and efficacy inanxious depression. . 2000;59:119-126.
J Clin Psychopharmacol
7. Aberg-Wistedt A, Agren H, Ekselius L, Bengtsson F, Akerblad AC. Sertralineversus paroxetine in major depression: clinical outcome after six months of continuoustherapy. . 2000;20:645-652.
J Clin Psychiatry
8. Nurnberg HG, Thompson PM, Hensley PL. Antidepressant medication changein a clinical treatment setting: a comparison of the effectiveness of selective serotoninreuptake inhibitors. . 1999;60:574-579.
9. Berndt ER, Russell JM, Miceli R, Colucci SV, Xu Y, Grudzinski AN. ComparingSSRI treatment costs for depression using retrospective claims data: the role of nonrandomselection and skewed data. . 2000;3:208-221.
Am J Manag
10. Russell JM, Berndt ER, Miceli R, Colucci S, Grudzinski AN. Course and cost oftreatment for depression with fluoxetine, paroxetine, and sertraline. Care. 1999;5:597-606.
11. Bensen WG, Fiechtner JJ, McMillen JI, et al. Treatment of osteoarthritis withcelecoxib, a cyclooxygenase-2 inhibitor: a randomized controlled trial. . 1999;74:1095-1105.
12. Simon LS, Weaver AL, Graham DY, et al. Anti-inflammatory and upper gastrointestinaleffects of celecoxib in rheumatoid arthritis: a randomized controlledtrial. 1999;282:1921-1928.
13. Emery P, Zeidler H, Kvien TK, et al. Celecoxib versus diclofenac in long-termmanagement of rheumatoid arthritis: randomized double-blind comparison. .1999;354:2106-2111.
14. Dougados M, Behier MJ, Jolchine I, et al. Efficacy of celecoxib, a cyclooxygenase2â€“specific inhibitor, in the treatment of ankylosing spondylitis. . 2001;44:180-185.
15. Acevedo E, Castaneda O, Ugaz M, et al. Tolerability profiles of rofecoxib(Vioxx®) and Arthrotec®. Scand J Rheumatol. 2001;30:19-24.
Arch Intern Med
16. Day R, Morrison B, Luza A, et al. A randomized trial of the efficacy and tolerabilityof the COX-2 inhibitor rofecoxib vs ibuprofen in patients with osteoarthritis.. 2000;160:1781-1787.
Arch Fam Med
17. Saag K, van der Heijde D, Fisher C, et al, Osteoarthritis Studies Group.Rofecoxib, a new cyclooxygenase 2 inhibitor, shows sustained efficacy, comparablewith other nonsteroidal anti-inflammatory drugs: a 6-week and a 1-year trial inpatients with osteoarthritis. . 2000;9:1124-1134.
18. Cannon GW, Caldwell JR, Holt P, et al, Rofecoxib Phase III Protocol 035Study Group. Rofecoxib, a specific inhibitor of cyclooxygenase 2, with clinical efficacycomparable with that of diclofenac sodium. . 2000;43:978-987.
19. Malmstrom K, Daniels S, Kotey P, Seidenberg BC, Desjardins PJ. Comparisonof rofecoxib and celecoxib, two cyclooxygenase-2 inhibitors, in postoperative dentalpain: a randomized, placebo- and active-comparatorâ€“controlled clinical trial.. 1999;21:1653-1663.
20. Morrison BW, Christensen S, Yuan W, et al. Analgesic efficacy of the cyclooxygenase-2â€“specific inhibitor rofecoxib in post-dental surgery pain: a randomized,controlled trial. . 1999;21:943-953.
21. Morrison BW, Daniels SE, Kotey P, et al. Rofecoxib, a specific cyclooxygenase-2 inhibitor, in primary dysmenorrhea: a randomized controlled trial. . 1999;94:504-508.
Am J Orthop
22. Reicin A, Brown J, Jove M, et al. Efficacy of single-dose and multidose- rofecoxibin the treatment of post-orthopedic surgery pain. . 2001;30:40-48.
23. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxibvs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoidarthritis: the CLASS study: a randomized controlled trial. 2000;284:1247-1255.
N Engl JMed
24. Bombadier C, Laine L, Reicin A, et al. Comparison of upper gastrointestinaltoxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. . 2000;343:1520-1528.
25. Spiegel BMR, Targownik L, Dulai GS, Gralnek IM. The cost-effectiveness ofcyclooxygenase-2 selective inhibitors in the management of chronic arthritis. . 2003;138:795-806.
26. Shaheen N, Ransohoff DF. Gastroesophageal reflux, Barrett esophagus, andesophageal cancer: scientific review. 2002;287:1972-1981.
Am J Gastroenterol
27. Howden CW, Henning JM, Huang B, Lukasik N, Freston JW. Management ofheartburn in a large, randomized, community-based study: comparison of 4 therapeuticstrategies. . 2001;96:1704-1710.
28. Eggleston A, Wigerinck A, Huijghebaert S, Dubois D, Haycox A. Cost effectivenessof treatment for gastro-oesophageal reflux disease in clinical practice: a clinicaldatabase analysis. . 1998;42:13-16.
J Manag Care Pharm
29. Lash S, Harding J. "Abandoned prescriptions": a quantitative assessment oftheir cause. . 1995;1:193-199.