Incidence of Gastrointestinal Events Among Bisphosphonate Patients in an Observational Setting

August 1, 2004
Redonda G. Miller, MD

Michael Bolognese, MD

Karen Worley, PhD

Amy Solis, BS

Richard Sheer, BA

Supplements and Featured Publications, Gastrointestinal Events and Medical Costs Associated With Bisphosphonate Therapy for Osteoporosis, Volume 10, Issue 7

Objective: To compare the incidence of gastrointestinal(GI) events among patients initiating alendronateor risedronate therapy.

Study Design: Retrospective observational studyof an administrative claims database.

Methods: Patients aged 65 years and older whoreceived a new prescription for risedronate (n = 865)or alendronate (n = 5255) between November 1,2000, and May 31, 2002, were selected for analysis.Preexisting GI conditions and medication use for the2 treatment groups in the 6 months before initiationof bisphosphonates were also determined. A Mantel-Haenszel relative risk estimate was used to comparethe incidence of GI events within the first 4 monthsof treatment.



Results: In both the alendronate and risedronatetreatment groups, the mean age was approximately 76years and 93% were female. Treatment groups had asimilar overall health status at baseline with the exceptionthat proportionally more individuals who initiatedrisedronate had preexisting GI conditions comparedwith alendronate users (13.8% vs 11%, odds ratio =0.77, = .02). In the first 4 months following initiationof treatment, 8.2% of alendronate patients and 5.5% ofrisedronate patients had a documented GI-relatedevent. Adjusting for age, sex, preexisting GI conditions,and number of concomitant medications in thepretreatment period, the alendronate patients exhibiteda 44% higher risk of GI events compared with risedronatepatients (relative risk = 1.44; 95% confidenceinterval, 1.03-2.00; = .03).

Conclusion: This analysis of administrativeclaims from a large managed care database supportsa difference between alendronate and risedronatewith respect to GI tolerability in the first 4 months oftherapy.

(Am J Manag Care. 2004;10:S207-S215)

The bisphosphonate therapeutic classhas become an important group ofmedications for the prevention and/ortreatment of postmenopausal and glucocorticoid-induced osteoporosis.

Risedronate and alendronate are bothnitrogen-containing compounds, but alendronateis a second-generation bisphosphonateand risedronate is a third-generationbisphosphonate with a nitrogen atom thatforms part of a pyridine ring.1 Structural differencesaffect not only the antiresorptivepotency of these compounds, but also thenature and extent of side effects, with thegastrointestinal (GI) tract drawing the mostconcern.1,2 Although the labels of both alendronateand risedronate instruct patients toremain upright for at least 30 minutes aftertaking the tablet (to reduce the potential foresophageal/gastric irritation), the clinicaldata suggest that GI side-effect profiles of the2 products may be different.

Data from the pivotal randomized controlledclinical trials of alendronate suggestedthat the drug was well tolerated and hadno greater clinical evidence of adverseeffects than placebo, although patients withcertain preexisting GI problems (eg, pepticulcer disease or dyspepsia) were excluded,3-5as were those recently treated with agentsthat irritate the GI tract.3 Soon after alendronate'srelease, an unexpectedly highernumber of cases of esophagitis and esophagealstrictures were encountered when thedrug was prescribed to the general population,which resulted in changes to thealendronate label.1,2 In the pivotal trials,risedronate was well tolerated, showing anincidence of side effects similar to placebo.6-10These trials included patients with preexistingand ongoing GI conditions. A pooledanalysis of the tolerability experiences ofmore than 10 000 patients (>98% postmenopausalwomen) from the risedronateclinical trials demonstrated that the incidenceof upper GI events was similar toplacebo, even in subpopulations withunderlying upper GI disease, nonsteroidalanti-inflammatory drug (NSAID) use, oracid-secretion blocker use.11

Several endoscopy studies have examinedthe effect of either alendronate or risedronatecompared with placebo or aspirin-treatedpatients.12-19 Only 2 published studieswere designed to make direct comparisonsbetween alendronate and risedronate (at thedose indicated for osteoporosis). Lanza et alrandomly assigned healthy, postmenopausalwomen to treatment with risedronate (5mg/day) or alendronate (10 mg/day) for 2weeks.20

Esophagogastroduodenoscopy was performedafter 7 and 14 days of dosing, andresults showed that risedronate was associatedwith a significantly lower incidence(4.1%) of gastric ulcers than alendronate(13.2%). Similar results were found byThomson et al where gastric ulcers wereobserved in 6% of the risedronate users and12.1% of the alendronate users, supporting anotable difference between risedronate andalendronate with respect to GI irritation.21This endoscopy research is suggestive of adifference in GI irritation between the 2drugs in selected clinical populations, butthe studies were limited by short duration,small sample sizes, and surrogate measuresof intolerance. Additional supporting evidencewould be advantageous, and observationaldatabase analyses could provideperspective on the GI profiles of both alendronateand risedronate in a population-basedsetting.

The purpose of this retrospective analysisis to examine the differences between alendronateand risedronate patients withrespect to GI conditions occurring beforeand after the initiation of bisphosphonatetreatment in a managed care setting.


Database Description.

We conducted aretrospective cohort study among patientswith medical and pharmacy claims containedin a proprietary administrative claimsdatabase.22 This database contains longitudinaldata, representing healthcare servicesfrom professional, facility, and outpatientpharmacy claims and enrollment data, andhas been used extensively for more than 10years to conduct retrospective studies.23-28These services are provided through healthmaintenance organizations, preferred providerorganizations, and various specialtyproducts to approximately 3 million membersannually. The plans cover a wide geographicdistribution, with members residingin more than 20 states.

International Classification ofDiseases, Ninth Revision, Clinical Modification(ICD-9-CM)

Study Population. Patients selected forthe present study were women and men 65years of age and older. All had a new prescription("index" prescription) for alendronate(5, 10, 35, or 70 mg) or risedronate(5 mg) between November 1, 2000, and May31, 2002 (Figure). Eligible patients wererequired to have medical and pharmacybenefits for the entire study period. Patientswith a bisphosphonate prescription in the 6months before the index date were not consideredto be new users and were thereforenot included in the analysis. Althoughpatients were not required to have a diagnosisof osteoporosis, all patients with a Paget'sdisease diagnosis code (731.0)were excluded from the analysis. Patientstaking risedronate 30 mg and patients takingalendronate 40 mg were also considered tobe Paget's disease patients and were notincluded in the study. Finally, individualswho switched products during the studyperiod (ie, risedronate to alendronate or viceversa) were excluded. These switchers representedless than 5% of the study population.


Because weekly alendronate was commerciallyavailable during the study period(and weekly risedronate was not), considerationwas given to combining weekly anddaily alendronate for the analyses. To determineif this was appropriate, alendronatepatients were initially excluded if theyswitched from daily to weekly or vice versa.Among the remaining individuals, weeklyalendronate patients were compared withdaily alendronate patients on the basis of GIevents in the follow-up period. The 2 groupsshowed no significant difference in incidenceof GI events, with 8.3% of the alendronateweekly and 7.6% of the alendronatedaily patients having a diagnosis code for aGI-related event (relative risk [RR] = 1.10, = .40). Therefore, the decision was made tocombine all alendronate users in the analysis,including those who switched dosestrength during the treatment period.

Study Period. The pretreatment periodwas defined as the 6-month period beforeinitiating treatment with the index therapy.The 4-month period after the index prescriptionwas considered to be the treatmentperiod. Rates of GI-related adverse eventstend to be highest in the first few months oftherapy, thus a 4-month treatment periodwas optimal for capturing the majority ofevents without compromising sample size.29



GI Health. In both the pretreatment andtreatment periods, patients were assessedwith respect to 3 components: GI events,prescription GI medication use, and prescription NSAID/salicylate use. In the pretreatmentperiod, a broad list of diagnosiscodes was evaluated to distinguishindividuals who have preexisting GI conditionsand thus might be more prone to developingGI problems after initiating bisphosphonatetherapy (Table 1). Based on thepresence of preexisting GI conditions, anadditional stratified analysis was conductedfor the treatment period. In the 4-monthtreatment period, a more restricted list ofcodes was used to define GI events,including only those conditions that couldpossibly be associated with bisphosphonates.GI events were examined in 2 ways:(1) those recorded as a primary diagnosisonly, and (2) those recorded as a primarydiagnosis and up to 4 secondary diagnosesfor a particular office visit or hospitalization.GI medication use was defined as at least 1prescription for an H2 receptor blocker, protonpump inhibitor (PPI), and/or cytoprotectiveagent. NSAID/salicylate use wasevaluated to assess possible differencesbetween the alendronate and risedronatepatients that could potentially influencetheir susceptibility to further GI irritation.

Statistical Methods. The Wilcoxon ranksum test was used to compare alendronateand risedronate patients on age, sex, numberof concomitant medications (quantifiedby the number of therapeutic classes30represented by prescriptions in the pretreatmentperiod), milligrams of oral glucocorticoids,number of hospitalizations, andnumber of specialist visits in the pretreatmentperiod because the variables were continuousand not normally distributed. Thechi-square test was used to compare the 2groups receiving oral glucocorticoid (percentwith at least 1 prescription). Mantel-Haenszel odds ratios (OR)31 and 95%confidence intervals were used to comparethe alendronate and risedronate patients onpreexisting GI conditions, GI medicationuse, and NSAID/salicylate use during thepretreatment period.

For the 4-month treatment period, the 2populations were compared on these variablesusing both crude and adjusted Mantel-Haenszel RR29 estimates and 95% confidenceintervals. Age, sex, and preexisting GI conditionswere included in the adjusted models.Additional variables were then selected onlyif they significantly added to the multivariatemodel. Mantel-Haenszel RRs were also usedto compare alendronate and risedronatepatients based on use of GI medications andNSAIDs/salicylates in the treatment period,with adjusted models including age, sex, anduse of GI medications or NSAIDs/salicylatesin the pretreatment period.

To provide an additional perspective, thebisphosphonate patients were also classifiedas having a "GI history" or "no GI history"based on the presence of 1 or more GI conditionsin the pretreatment period. A stratifiedanalysis was then carried out by usingMantel-Haenszel RRs to compare the alendronateand risedronate users and adjustingfor age and sex.


A total of 865 risedronate patients and5255 alendronate patients were eligible foranalysis. The mean age was approximately76 years for both the risedronate and alendronategroups, and 93% of the patients werewomen, with no statistically significant differencesobserved (Table 2).

The risedronate and alendronate patientswere very similar with respect to their overallhealth status at baseline, as indicated bythe number of concomitant medications,hospitalizations, and specialist visits. Nostatistically significant differences wereobserved between the groups (Table 2). Aslightly higher proportion of risedronatepatients had filled a prescription for an oralglucocorticoid, but the difference was minimaland neither the percentage of patientswith a glucocorticoid prescription nor theaverage cumulative 6-month exposure inprednisone equivalents among patients withat least 1 glucocorticoid prescription for the2 groups (900 vs 865 mg, respectively, forrisedronate and alendronate) was significantlydifferent. Only "number of concomitantmedications" was included in the finalmodel to compare alendronate and risedronateon GI events, because it significantlycontributed to the explanatory power ofthe model.




In the 6 months before treatment, a statisticallysignificant higher percentage of risedronatepatients (13.8%) compared withalendronate patients (11.0%) had a GI-relatedevent (Table 3). Alendronate patients were23% less likely (OR = 0.77, = .02) to havehad GI events in the pretreatment period.The 2 groups were very similar with respectto use of potential gastric irritants; 22.2% ofrisedronate patients and 22.4% of alendronatepatients had prescription NSAIDs or salicylates(OR = 1.01, = .91) during the pretreatmentperiod. Use of medicationstypically prescribed to treat GI problems wasalso comparable, with 19% of the risedronateand 18.7% of the alendronate patients usingprescription H2 receptor blockers, PPIs, orcytoprotectives (OR = 0.98, = .87).



Incidence of GI Events. Within the first4 months of initiating therapy, 8.2% of thealendronate patients and 5.5% of the risedronatepatients experienced a GI eventbased on a primary diagnosis code (Table 4).In the final model, adjusting for age, sex,preexisting GI conditions, and number ofconcomitant medications, the alendronatepatients showed a 44% higher risk of GIevents during the treatment period comparedwith risedronate patients ( = .03). Inthe analysis examining all GI events, regardlessof coding in the primary position, alendronatepatients had a statisticallysignificant 34% higher risk of GI events inthe first 4 months of therapy ( = .02).



In a stratified analysis, where patientswere grouped based on their GI history,findings were similar. Among those with nohistory of GI events in the 6-month pretreatmentperiod, the alendronate patientsexhibited a statistically significant 49% higherrisk of GI events in the first 4 months oftreatment compared with the risedronatepatients, adjusting for age and sex (RR =1.49, = .03). An even higher elevation inadjusted risk (57%) was observed for thepatients with a known history of GI events(RR = 1.57, = .05). A descriptive analysisof only esophageal events coded as the primarydiagnosis revealed that 2.1% of thealendronate patients and 1.6% of the risedronatepatients had an esophageal event.Given the small number of esophagealevents, this difference was not statisticallysignificant.



NSAID/Salicylate and GI MedicationUse. A slightly higher percentage of alendronate(17.4%) versus risedronate (16.9%)patients had a prescription for an NSAID orsalicylate during the treatment period(Table 5). Adjusting for previous use of thesemedications, the risk was comparable for the2 groups (RR = 1.05, = .71). Alendronatepatients also had a somewhat higher proportionof GI medication use (18% vs 17.3%),but the adjusted RR estimate (RR = 1.02)was not statistically significant ( = .74).


In the present analysis of a fully insuredelderly managed care population, alendronatepatients exhibited a significantlyhigher risk of GI events in the first 4 monthsof therapy compared with risedronatepatients. This increased risk was observedregardless of whether the GI diagnosis wascoded in the primary position or in any position.These observational results support thefindings of previous head-to-head comparativeendoscopy trials in that the 2 bisphosphonatesdid not exhibit identical profileswith respect to GI irritation. The endoscopystudy by Lanza et al showed healthy, postmenopausalrisedronate patients to be at significantlylower risk (4.1% incidence) forgastric ulcers compared with alendronatepatients (13.2% incidence).20 Similarly,Thomson et al demonstrated more potentialfor gastric irritation by alendronate thanrisedronate, with rates of gastric ulcer foundto be 12.1% and 6%, respectively.21

These results also offer a unique perspectiveof clinical practice that cannot beexplored via randomized clinical trials orendoscopic studies. For example, patientswith preexisting GI problems were excludedfrom the alendronate clinical trials andtreated patients appeared to have a GI eventrate similar to placebo.3-5 Once alendronatewas available on the market, however, real-worlduse of the drug was associated with anincreased number of esophagitis andesophageal strictures.2 Observational databasestudies have several advantages overcontrolled clinical trials in that real-worldpractice patterns can be observed over avariety of health plans and physician specialties,a large number of patients can befollowed over time, and the patients may bemore typical of those seen in actual practice,because they are not excluded by restrictiveclinical trial exclusion criteria. Additionally,it is possible to generate comparative datafor conditions with relatively low incidencebecause a sizable patient population can bereadily evaluated.




There are several limitations of administrativeclaims database research as well.Patient charts cannot be reviewed to confirmthe accuracy of diagnosis codes,unknown variables (such as compliancewith pill-taking instructions or others notcaptured in a database) could influencetreatment and outcomes, and only clinicallyreported events can be evaluated.Specifically, with respect to the evaluation ofGI events, only those conditions severeenough to warrant medical service (eg,physician visit) and subsequently receive andiagnosis code are counted. If thepatient merely mentions the GI complaintas a secondary concern, then it is less likelyto be coded. Another limitation is the accuracyof the specific codes assigned tothe patient. The individual responsible forcoding the patient's visit might select onecode over another because of reimbursementassociated with a specific code.Additionally, a physician's preconceptionsabout a particular therapy can potentiallylead to over- or underdiagnosis of a certaincondition among those treated patients.

Patients included in this analysis are representativeof a fully insured elderly population,which is eligible to receive prescriptionsfor osteoporosis therapy in a managed caresetting. This study did not explore bisphosphonateuse among patients who lackhealth insurance (medical and/or prescription)benefits or the differences in patternsof care among patients with differing copaymentstructures because the data were notavailable.


While the primary focus of this analysiswas to evaluate differences in the incidence ofGI events (via codes), use of medicationsto treat such GI events was alsoassessed. Most GI events are treated withPPIs, H2 receptor blockers, or cytoprotectives.Because many forms of these therapies areavailable over the counter, it is difficult totruly assess their use in an administrativeclaims database that captures only prescriptionproducts. Furthermore, any medicationsthat were dispensed in a hospital setting arerolled into a single claim for the hospital visitand cannot be identified individually. Thus,in the present study, use of GI medications islikely to be underreported for both bisphosphonates.Whether or not there is differentialunderreporting is unknown.


In observational database analyses, it isalso of interest to establish the comparabilityof patient populations at baseline or beforetreatment. For the present study, it wasimportant to understand whether the treatmentgroups were similar in their underlyingrisk for incurring a GI problem. For the 6-month pretreatment period, we consideredpatients to have a GI history if they had atleast 1 code for a GI-related condition.This served as a surrogate for chart review(which typically is not available in a claimsdatabase) and for more definitive yet invasiveevaluations, such as endoscopy. As seen inTable 3, risedronate patients were notablymore likely to have had a GI condition beforeinitiating therapy. This was accounted for inthe analysis, however, via both statisticaladjustment and stratification based on GI history.Other more general measures of healthstatus were also considered, including concomitantmedication use, number of specialistvisits, and hospitalizations. In this study,treatment group comparability does notappear to be an issue affecting study results.The risedronate and alendronate groupsshowed no significant differences withrespect to these health status indicators,lending support to the assumption that the 2populations were fairly comparable overall.

Another issue that arises in databaseresearch is the appropriate classification ofpatients who switch therapies during the follow-up period. In the present study, weexcluded any patient who switched productsduring the 4-month treatment period. Thiseliminates the potential for confusion inassignment of GI events to the current versusbefore therapy. Noteworthy is that byexcluding these patients it was not possibleto examine those who switched therapies,potentially as a result of a GI event. Evaluatingswitching patterns is important inunderstanding tolerability with therapiesused in a naturalistic setting, as seen in astudy by Hamilton et al, to determine compliancewith risedronate dosing instructions.32 The authors concluded that onethird of patients who discontinued risedronatetherapy because of adverse eventsbegan therapy again. Of those who didresume therapy, adverse events subsided inhalf of the patients, and they were able tocontinue their full course of therapy. In thepresent study, less than 5% of the potentialsample was eliminated on the basis of aproduct switch during follow-up.


In summary, the present study showedthat within the first 4 months of treatment,patients taking alendronate had a significantlyhigher risk of incurring a GI eventcompared with patients taking risedronate.No difference in GI tolerability was observedbetween the weekly and daily doses of alendronate.Additionally, patients with preexistingGI problems were significantly morelikely to initiate therapy with risedronatethan alendronate. Tolerability issues, includingdifferences in the incidence of GI-relatedevents, have important implicationsnot only for the patient, but also the managedcare organization and clinical pharmacist.Opportunities exist in a managedcare setting to improve utilization managementas well as reduce the risk of GIevents in patients receiving bisphosphonatetherapy. To ensure treatment effectiveness,physicians must be deliberate inproviding accurate instructions to thepatient on administration techniques andcarefully monitor patient adherence to treatmentregimens. In turn, the patient mustfully understand the implications of nonadherenceto these instructions to reduce thelikelihood of GI events. Utilization managementstrategies must direct appropriate prescribingto patients without a documentedhistory of at-risk conditions to preventfuture GI events as well. Excess events maylead to a substantial increase in healthcarecosts (ie, medications, hospitalizations,physician visits) for patients receiving a particulartherapy. Additional research, includingan economic analysis of the medicalcosts and resource utilization associatedwith each therapy, can further illustrate theimportance of such tolerability differencesand quantify the impact for managed care.



We would like to thank David Graham,MD, Veterans Affairs Medical Center,Houston, Texas, for his input in selectingappropriate codes for analysis, andMichael Steinbuch, PhD, section head, Epidemiologyand Pharmacoeconomics, Procter& Gamble Pharmaceuticals, Cincinnati,Ohio, for his assistance with manuscriptdevelopment and revisions.

Dig Dis Sci.

1. Graham DY. What the gastroenterologist shouldknow about the gastrointestinal safety profiles of bisphosphonates. 2002;47:1665-1678.

N Engl J Med.

2. de Groen PC, Lubbe DF, Hirsch LJ, et al. Esophagitisassociated with the use of alendronate. 1996;335:1016-1021.

N Engl J Med.

3. Liberman UA, Weiss SR, Broll J, et al. Effect of oralalendronate on bone mineral density and the incidenceof fractures in postmenopausal osteoporosis. The AlendronatePhase III Osteoporosis Treatment Study Group. 1995;333:1437-1443.


4. Black DM, Cummings SR, Karpf DB, et al. Randomisedtrial of effect of alendronate on risk of fracturein women with existing vertebral fractures. FractureIntervention Trial Research Group. 1996;348:1534-1541.


5. Cummings SR, Black DM, Thompson DE, et al. Effectof alendronate on risk of fracture in women with low bonedensity but without vertebral fractures: results from theFracture Intervention Trial. 1998;280:2077-2082.

Calcif Tissue Int.

6. Wallach S, Cohen S, Reid DM, et al. Effects of risedronatetreatment on bone density and vertebral fracturein patients on corticosteroid therapy. 2000;67:277-285.

N Engl J Med.

7. McClung MR, Geusens P, Miller PD, et al. Effect ofrisedronate on the risk of hip fracture in elderly women.Hip Intervention Program Study Group. 2001;344:333-340.


8. Harris ST, Watts NB, Genant HK, et al. Effects ofrisedronate treatment on vertebral and nonvertebral fracturesin women with postmenopausal osteoporosis: arandomized controlled trial. Vertebral Efficacy withRisedronate Therapy (VERT) Study Group. 1999;282:1344-1352.

Osteoporos Int.

9. Reginster J, Minne HW, Sorensen OH, et al. Randomizedtrial of the effects of risedronate on vertebralfractures in women with established postmenopausalosteoporosis. Vertebral Efficacy with Risedronate Therapy(VERT) Study Group. 2000;11:83-91.

J Clin Endocrinol Metab.

10. Mortensen L, Charles P, Bekker PJ, Digennaro J,Johnston CC Jr. Risedronate increases bone mass in anearly postmenopausal population: two years of treatmentplus one year of follow-up. 1998;83:396-402.

Mayo Clin Proc.

11. Taggart H, Bolognese MA, Lindsay R, et al. Uppergastrointestinal tract safety of risedronate: a pooledanalysis of 9 clinical trials. 2002;77:262-270.

Am JGastroenterol.

12. Graham DY, Malaty HM, Goodgame R. Primaryamino-bisphosphonates: a new class of gastrotoxicdrugs—comparison of alendronate and aspirin. 1997;92:1322-1325.

Am J Gastroenterol.

13. Lanza F, Rack MF, Simon TJ, Lombardi A, Reyes R,Suryawanshi A. Effects of alendronate on gastric and duodenalmucosa. 1998;93:753-757.

AlimentPharmacol Ther.

14. Marshall JK, Rainsford KD, James C, Hunt RH. Arandomized controlled trial to assess alendronate-associatedinjury of the upper gastrointestinal tract. 2000;14:1451-1457.

Am J Gastroenterol.

15. Lanza F, Sahba B, Schwartz H, et al. The upper GIsafety and tolerability of oral alendronate at a dose of 70milligrams once weekly: a placebo-controlled endoscopystudy. 2002;97:58-64.

Aliment Pharmacol Ther.

16. Graham DY, Malaty HM. Alendronate gastric ulcers. 1999;13:515-519.

Am J Gastroenterol.

17. Lowe CE, Depew WT, Vanner SJ, Paterson WG,Meddings JB. Upper gastrointestinal toxicity of alendronate. 2000;95:634-640.

Aliment Pharmacol Ther.

18. Lanza FL, Rack MF, Li Z, Krajewski SA, Blank MA.Placebo-controlled, randomized, evaluator-blindedendoscopy study of risedronate vs. aspirin in healthypostmenopausal women. 2000;14:1663-1670.

ArchIntern Med.

19. Graham DY, Malaty HM. Alendronate and naproxenare synergistic for development of gastric ulcers. 2001;161:107-110.


20. Lanza FL, Hunt RH, Thomson AB, Provenza JM,Blank MA. Endoscopic comparison of esophageal andgastroduodenal effects of risedronate and alendronate inpostmenopausal women. 2000;119:631-638.

Helicobacterpylori status.

J Rheumatol.

21. Thomson AB, Marshall JK, Hunt RH, et al. 14 dayendoscopy study comparing risedronate and alendronatein postmenopausal women stratified by 2002;29:1965-1974.

22. Constella Health Strategies. Databases at a Glance.Available at: Accessed June 28, 2004.

Am J Manag Care.

23. Dubois RW, Alexander CM, Wade S, et al. Growthin use of lipid-lowering therapies: are we targeting theright patients? 2002;8:862-867.

Am J Kidney Dis.

24. London R, Solis A, Goldberg GA, Wade S, Ryu S.Health care resource utilization and the impact of anemiamanagement in patients with chronic kidney disease. 2002;40:539-548.

Drug Benefit Trends.

25. Lee DW, Fowler RB, Kadlubek PJ, Haberman M.No significant difference in diabetes risk during treatmentwith typical versus atypical antipsychotics: resultsfrom a large observational study. 2002;14:46-52.

J Manag Care Pharm.

26. Arocho R, Solis A, Wade S, Goldberg G, Tang S.Exploring the methodological challenges of investigatingcomparison groups with different underlying characteristics:a case study. 2002;8:353-359.

J Bone Joint Surg Am.

27. Freedman KB, Kaplan FS, Bilker WB, Strom BL, LoweRA. Treatment of osteoporosis: are physicians missing anopportunity? 2000;82-A:1063-1070.

Am J Manag Care.

28. Ascher-Svanum H, Kennedy JS, Lee D, HabermanM. The rate, pattern, and cost of use of antiparkinsonianagents among patients treated for schizophrenia in amanaged care setting. 2004;10:20-24.

Am J Manag Care.

29. Ettinger B, Pressman A, Schein J. Clinic visits andhospital admissions for care of acid-related upper gastrointestinaldisorders in women using alendronate forosteoporosis. 1998;4:1377-1382.

AHFS: Drug Information.

30. McEvoy GK, ed. Bethesda,Md: American Society of Health-System Pharmacists;2003.

Epidemiologic Research: Principles and QuantitativeMethods (Industrial Health and Safety).

31. Kleinbaum DG, Kupper LL, Morgenstern H. New York, NY: John Wiley & Sons, Inc; 1982.

Osteoporos Int.

32. Hamilton B, McCoy K, Taggart H. Tolerability andcompliance with risedronate in clinical practice. 2003;14:259-262.