Physicians' Behavior and Their
`Interactions With Drug Companies
`A Controlled Study of Physicians Who Requested Additions
`to a Hospital Drug Formulary
`Mary-Margaret Chren, MD, C. Seth Landefeld, MD
`
`Objective.\p=m-\Itis controversial whether physicians'
`interactions with drug com-
`panies affect their behavior. To test the null hypothesis, that such interactions are
`not associated with physician behavior, we studied one behavior: requesting that
`a drug be added to a hospital formulary.
`Design.\p=m-\Nestedcase-control study.
`Setting.\p=m-\Universityhospital.
`Participants.\p=m-\Full-timeattending physicians. Case physicians were all 40 phy-
`sicians who requested a formulary addition from January 1989 through October
`1990. Control physicians were 80 randomly selected physicians who had not made
`requests.
`Main Exposure Measure.\p=m-\Physicianinteractions with drug companies, as de-
`termined by survey of physicians (response rate, 88% [105/120]).
`Results.\p=m-\Physicianswho had requested that drugs be added to the formulary
`interacted with drug companies more often than other physicians; for example, they
`were more likely to have accepted money from companies to attend or speak at
`educational symposia or to perform research (odds ratio [OR], 5.1 ; 95% confidence
`interval [CI], 2.0 to 13.2). Furthermore, physicians were more likely than other phy-
`sicians to have requested that drugs manufactured by specific companies be added
`to the formulary if they had met with pharmaceutical representatives from those
`companies (OR, 13.2; 95% CI, 4.8 to 36.3) or had accepted money from those
`companies (OR, 19.2; 95% CI, 2.3 to 156.9). These associations were consistent
`in multivariable analyses controlling for potentially confounding factors. Moreover,
`physicians were more likely to have requested formulary additions made by the
`companies whose pharmaceutical representatives they had met (OR, 4.9; 95% CI,
`3.2 to 7.4) or from whom they had accepted money (OR, 1.7; 95% CI, 1.0 to 2.7)
`than they were to have requested drugs made by other companies.
`Conclusion.\p=m-\Requestsby physicians that drugs be added to a hospital formu-
`lary were strongly and specifically associated with the physicians' interactions with
`the companies manufacturing the drugs.
`(JAMA. 1994;271:684-689)
`
`From the Departments of Dermatology (Dr Chren)
`and Medicine (Dr Landefeld) and the Program in Health
`Care Research (Drs Chren and Landefeld), Cleveland
`(Ohio) Veterans Affairs Medical Center, University
`Hospitals of Cleveland, and Case Western Reserve
`University School of Medicine, Cleveland (Drs Chren
`and Landefeld).
`Presented in part at meetings of the American Fed-
`eration of Clinical Research, May 6, 1991 (Seattle,
`Wash), and May 3, 1992 (Baltimore, Md).
`Reprint requests to Department of Dermatology 11\x=req-\
`G(W), Cleveland Veterans Affairs Medical Center,
`10701 East Blvd, Cleveland, OH 44106 (Dr Chren).
`
`PHYSICIANS interact with drug
`these
`companies in many ways, but
`interactions are controversial.111 For
`example, according to the American
`Federation for Clinical Research, "[I]n-
`dustry-sponsored lectureships, sympo¬
`sia, and support for research are an in¬
`tegral part of the biomedicai research
`and educational system that would be
`severely damaged by disallowing such
`
`activities."12 In contrast, some see in
`such arrangements "the essence ofgood
`bribery."1 This controversy is due,
`in
`part, to the paucity of studies of phy¬
`sicians' behavior and factors that affect
`it.
`
`Many types of physician behavior re¬
`late to drugs. Not only do physicians
`prescribe the drugs that patients buy,
`but physicians who speak about drugs
`at medical symposia may influence their
`listeners' prescribing behavior,5 faculty
`physicians shape the prescribing behav¬
`ior of their students, and physicians' re¬
`search affects the availability of and in¬
`dications for drugs. An additional phy¬
`sician behavior—requesting that a drug
`be added to a hospital formulary—is also
`important to drug sales,13 because the
`formulary regulates the drugs available
`for physicians to prescribe for hospital
`patients. Furthermore, in teaching hos¬
`pitals, the formulary determines which
`drugs are most easily prescribed by phy¬
`sicians in training, who are developing
`prescribing styles likely to persist in
`their careers.
`Physicians generally perceive their be¬
`havior regarding prescription drugs as
`motivated mainly by drug performance
`data, yet their beliefs about selected
`drugs more closely matched advertising
`claims.14 Although the influence of drug
`companies on physician behavior has not
`been extensively studied, available evi¬
`dence suggests that physician behavior
`with respect to drugs may be associated
`with drug company promotional activi¬
`ties.1519 Most studies have been limited,
`however, in that they relied solely on
`physicians' reports of their own behav¬
`ior rather than observing it directly and
`did not relate the behavior of individual
`doctors to their interactions with spe¬
`cific companies.
`
`
`
`Downloaded from at University of California - San Diego on October 8, 2009 www.jama.com
`
`
`
`ARGENTUM Exhibit 1191
`Argentum Pharmaceuticals LLC v. Research Corporation Technologies, Inc.
`IPR2016-00204
`
`

`
`STUDY GOAL AND HYPOTHESES
`Our goal was to determine if physi¬
`cians' interactions with drug companies
`were associated with a physician behav¬
`ior that was independently observed,
`namely, requesting that a drug be added
`to our hospital formulary. To accomplish
`this goal, we compared two groups of
`physicians, physicians who had requested
`that drugs be added to a hospital formu¬
`lary and other physicians who had not.
`We studied interactions between doctors
`and drug companies in general and be¬
`tween doctors and the specific compa¬
`nies manufacturing drugs that had been
`requested. For each interaction, we
`tested the null hypothesis, that interac¬
`tions were not associated with the be¬
`havior. In particular, we tested three dis¬
`crete null hypotheses: (1) Physicians who
`interacted with drug companies in gen¬
`eral were no more likely than other phy¬
`sicians to have made formulary requests;
`(2) physicians who interacted with spe¬
`cific companies were no more likely than
`other physicians to have requested drugs
`manufactured by those companies; and
`(3) physicians who interacted with spe¬
`cific companies were no more likely to
`have requested drugs manufactured by
`those companies than drugs manufac¬
`tured by other companies.
`
`METHODS
`Setting
`The study was conducted at the Uni¬
`versity Hospitals of Cleveland (Ohio),
`an 874-bed teaching hospital affiliated
`with Case Western Reserve University
`School of Medicine. A request for the
`addition of a drug to the formulary may
`be made by any health care professional
`at the hospital who completes a stan¬
`dard form about the drug's pharmaco¬
`logie properties. The form stipulates that
`requests may not be made for drugs if
`the requester is involved in research on
`the requested drug that is sponsored by
`the company making the drug. All forms
`are included in the minutes of the For¬
`mulary Review Subcommittee of the
`Pharmacy and Therapeutics Committee.
`During the period of this study, phar¬
`maceutical sales representatives were
`permitted access to most patient care
`the emergency depart¬
`areas, except
`ment, nurseries, and intensive care, la¬
`bor, delivery, operating, recovery, and
`dialysis units. Institutional review board
`approval was obtained for this investi¬
`gation.
`Subjects
`The study had a nested case-control
`design. Subjects were physicians on the
`full-time attending staff at University
`Hospitals of Cleveland. Case physicians,
`
`who were identified by review of the
`minutes of the Formulary Review Sub¬
`committee of the Pharmacy and Thera¬
`peutics Committee, were all 40 physi¬
`cians who had requested that a drug be
`added to the formulary from January 1,
`1989 to October 31, 1990; all requests
`during that time period had been made
`by full-time attending physicians. Con¬
`trol physicians were those who had not
`made formulary requests and were mem¬
`bers ofdepartments directly involved in
`selecting medications for patients; phy¬
`sicians in the departments of pathology,
`nuclear medicine, diagnostic radiology,
`and radiation therapy were excluded.
`Eighty control physicians were selected
`randomly from the 330 eligible physi¬
`cians to avoid overmatching.20
`Data Collection and Measures
`Drugs.—There were 55 requests for
`52 commercially available drugs to be
`added to the hospital formulary; 28 phy¬
`sicians requested that one drug be added,
`10 requested two drugs, one requested
`three drugs, and one requested four
`drugs. The company manufacturing the
`drug at the time of the request or the
`company specifically noted by the re¬
`questing physician (if more than one
`company manufactured the drug) was
`considered to be the manufacturing com¬
`pany. There were 35 manufacturing com¬
`panies.
`The requested drugs were classified
`by a doctor of pharmacy who was fa¬
`miliar with the University Hospitals for¬
`mulary but not associated with this
`study, using a modification of the Food
`and Drug Administration therapeutic
`rating.21 The requested drugs were clas¬
`sified as representing a major thera¬
`peutic advantage over therapies exist¬
`ing in the formulary (n=13), a modest
`advantage (n=13), or little or no advan¬
`tage (n=29); this classification was spe¬
`cific to the formulary (Table 1). The drugs
`were also classified as having been re¬
`quested for addition to the formulary in
`the year following Food and Drug Ad¬
`ministration approval (n=23) or after that
`year (n=23) (this information was not
`available for one drug).
`Demographic Characteristics and
`Publications of Physicians.—Demo¬
`graphic information about all physicians
`was obtained from the hospital staff of¬
`fice, the medical school, and the Com¬
`pendium ofMedical Specialists.22 MED-
`LINE-indexed publications by each phy¬
`sician in 1989 were classified as reflect¬
`ing laboratory-based or patient-based
`research.
`Survey Instrument.—To measure in¬
`teractions with drug companies, a sur¬
`vey instrument was developed,
`to be
`completed by all physicians. This instru-
`
`Table 1.—Drugs for Which Requests Were Sub¬
`mitted for Addition to a Hospital Formulary, Classi¬
`fied According to Their Therapeutic Advantage
`Over Drugs Already in the Formulary*
`Major Advantage
`Clofazimine
`Clozapine
`Epoetìn alfa
`Fluconazole
`Flurbiprofen sodium ophthalmic drops
`Ganclclovir
`Levamisole
`Mefloquine
`Methoxsalen
`Nimodiplne
`Octreotide
`Pulmonary surfactant, synthetic
`Typhoid vaccine, oral
`Modest Advantage
`Adenosine
`Buproprion
`Carboplatin
`Flutamide
`Ifosfamide/mesna
`Ketorolac tromethamine
`Moriclzine
`Nafarelin acetate
`Omeprazole
`Pergolide mesylate
`Permethrin 1%
`Permethrln 5%
`Selegillne hydrochloride
`Little or No Advantage
`Albuterol Repetabs
`Albuterol Rotacaps
`Acebutolol hydrochloridef
`Ceflxime
`Ciclopi roxt
`Dezocine
`Diclofenac
`Diltiazem hydrochloride sustained release
`Disopyramide
`Esmolol concentrate
`Estradiol, transdermal
`Flecainide acetate
`Goserelin acetate
`Hydroxypropyl methylcellulose drops
`Lidocaine spray
`Lisinopril
`Magnesium gluconate
`Methotrexate Dosepack
`Multivitamlns, renal
`Oxycodone with aspirin
`Nifedipine extended release
`Pancrelipase
`Propafenone hydrocloride
`Propofol
`Scopolamine, transdermal
`Ticarcillin/potassium clavulanatef
`*Drugs in bold type were not accepted for addition to
`formulary.
` fDrugs for which two requests were submitted.
`
`ment was pretested by seven physicians
`to determine ambiguous items and was
`revised accordingly; the format of spe¬
`cific questions is available on request.
`The instrument inquired about the
`number of patients cared for by each
`physician. It also inquired about four
`types of interactions with drug compa¬
`nies in the past year: traditional detail¬
`ing by pharmaceutical company repre¬
`sentatives, acceptance of money from
`drug companies to support attendance
`at educational symposia, acceptance of
`money to speak at educational sympo¬
`sia, and acceptance of money for re¬
`search.
`Each physician was asked about these
`interactions with drug companies in gen-
`
`Downloaded from
`
`www.jama.com
`
` at University of California - San Diego on October 8, 2009
`
`

`
`Table 2.—Physician Interactions in the Past Year With Drug Companies in General, According to Self-reports
`by Physicians Who Submitted Formulary Requests (Cases) and by Those Who Did Not (Controls)
`No. (%) of Cases
`No. (%) of Controls
`(Physicians Submitting
`(Physicians Not Submitting
`Requests) (n=69)t
`Requests) (n=36)t
`
`
`
`Type of Interaction*
`Shared meals worth > $10
`Never
`17(47)
`42(61)
`11 (31)
`24 (35)
`Rarely
`3 (8)
`Occasionally
`2 (3)
`Often_5(14)_1_0)_
`Accepted money for travel or lodging
`to attend educational symposia
`Never
`Rarely
`Occasionally
`Often
`Accepted money to speak at
`educational symposia
`Never
`Rarely
`Occasionally
`Often
`Accepted money to perform research
`No
`Yes
`
`63(91)
`5(7)
`0 (0)
`1(1)
`
`55 (80)
`9(13)
`4(6)
`1(1)
`
`49(71)
`20(29)
`
`_
`
`_
`
`.06
`
`<.001
`
`.002
`
`the same specific companies (n=82). Mul¬
`tivariable logistic regression models23
`were used to determine whether any
`observed associations were confounded
`by other physician characteristics. Al¬
`though this analysis used unmatched
`controls,20 the consistency of these re¬
`sults was examined with an alternative
`analysis comparing case physicians and
`physicians matched for types ofpatients
`seen. In this analysis, for each potential
`interaction between a case physician and
`the company making the drug he or she
`had requested (n=47), a matched com¬
`parison physician was selected who had
`not requested a drug made by that com¬
`pany but who had been surveyed about
`interactions with that company. Match¬
`ing of physicians in this way by type of
`patient seen was possible for 32 ofthe 47
`formulary requests; in 21 the physicians
`were in the same department, and for
`the remaining 11 requests the physi¬
`cians were in different departments but
`cared for patients with similar problems
`(for example, a rheumatologist was
`matched to an orthopedic surgeon). Of
`the 32 matched comparison physicians,
`24 were control physicians and eight
`were case physicians who had not re¬
`quested drugs made by the manufac¬
`turer of the drug requested by the case
`to whom they were matched. The com¬
`parison involved a matched analysis in
`which the odds ratio was calculated as
`the ratio of discordant pairs.20
`The third hypothesis was also tested
`by examining associations between re¬
`questing a formulary addition and phy¬
`sician interactions with specific drug
`companies; for this analysis, case phy¬
`sicians were studied with respect to their
`interactions with the specific companies
`making the drugs they had requested,
`compared with their interactions with
`other specific companies. This compari¬
`son also involved a matched analysis in
`which the odds ratio was calculated as
`the ratio of discordant pairs.20
`Data analyses were conducted using
`SAS (Statistical Analysis System, Ver¬
`sion 6).2:! Associations between dichoto-
`mous variables were evaluated by Fish¬
`er's Exact Test. Associations between
`other categorical variables were tested
`by the 2 statistic, with the modification
`for linear trend when appropriate. Con¬
`tinuous variables in two groups were
`compared by the Wilcoxon Rank-Sum
`Test.
`
`RESULTS
`Demographic and Professional
`Characteristics
`Case and control physicians were simi¬
`lar in age, academic rank, tenure status,
`and mean numbers of laboratory- and
`
`27(75)
`7(19)
`2 (6)
`0(0)
`
`17 (47)
`5(14)
`5(14)
`9(25)
`
`14(39)
`22(61)
`
`Types of
`interactions are defined in the text. "Rarely" was defined as one to two times in the past year,
`"occasionally" as three to five times in the past year, and "often" as more than five times in the past year. One column
`does not total 100% due to rounding.
`tOf the 120 physicians, 36 cases and 69 controls responded to the survey instrument.
`
`eral and with seven specific drug com¬
`panies. The 35 companies whose drugs
`had been requested were randomly di¬
`vided into groups of seven companies
`each. Each case physician was asked
`about the group containing the company
`or companies whose drugs he or she had
`requested. For each case physician two
`unmatched control physicians were
`asked about interactions with the same
`seven companies.
`Case and control physicians were sent
`the survey instrument with a letter ask¬
`ing them to participate in a study of
`drug companies' interactions with phy¬
`sicians; the study hypotheses were not
`specified in the letter. Physicians were
`informed that all answers would be coded
`and kept confidential.
`Survey data were available for 105
`(88%) of 120 physicians: 36 (90%) of 40
`case physicians, who had requested 47
`(85%) of the 55 drugs, and 69 (86%) of 80
`control physicians. Of the 36 cases on
`whom survey data were available, 26
`had requested that one drug be added,
`nine had requested two drugs, and one
`had requested three drugs. For specific
`companies, this response rate permit¬
`ted comparison of 47 potential interac¬
`tions between case physicians and the
`specific companies whose drugs they had
`requested and 82 potential interactions
`between the unmatched control physi¬
`cians randomly assigned to each case
`and the same companies.
`Independent Information About Drug
`Company-Sponsored Research.—Phy¬
`sician responses about participation in
`
`research sponsored by drug companies
`were validated by data from hospital de¬
`partments about such research in 1989
`and 1990. This information was provided
`for 92 (77%) of the 120 physicians. In
`every instance in which departmental in¬
`formation indicated that a physician who
`had completed the survey had performed
`drug company-sponsored research, the
`physician had provided this information
`in the survey.
`Analytic Strategy
`The overall strategy for analysis was
`to test each of the three specific null
`hypotheses by determining associations
`between the physician behavior (re¬
`questing that a drug be added to the
`hospital formulary) and types of physi¬
`cians' interactions with drug companies,
`and then to examine any observed as¬
`sociations for consistency.
`The first hypothesis was tested by
`examining associations between request¬
`ing a formulary addition and physician
`interactions with drug companies in gen¬
`eral (data available on 36 cases and 69
`controls from the completed survey in¬
`struments).
`The second hypothesis was tested by
`examining associations between request¬
`ing a formulary addition and physician
`interactions with specific drug compa¬
`nies. The rates of interaction of case
`physicians with each specific company
`whose drug the case physician had re¬
`quested (n=47) were compared with the
`rates of interaction of the randomly as¬
`signed control physicians asked about
`
`Downloaded from
`
`www.jama.com
`
` at University of California - San Diego on October 8, 2009
`
`

`
`Table 3.—Rates of Interaction With Specific Drug Companies Reported by Physicians Who Submitted
`Formulary Requests (Cases) and Those Who Did Not (Controls)
`Interaction Rates, %,f Between
`
`I
`
`Cases and Specific
`Company Manufacturing
`the Drug Requested
`(n=47)
`
`Type of Interaction*
`Met with pharmaceutical
`representatives
`Accepted money to attend
`symposia
`Accepted money to speak at
`symposia
`Accepted money to perform
`company-sponsored research
`Accepted money to attend
`symposia, speak at symposia,
`or perform research
`
`51 (24/47)
`
`4 (2/47)*.
`
`11 (5/47)*.
`
`11 (5/47)*.
`
`19(9/47)*.
`
`I
`
`Controls
`and Specific
`Companies
`(n=82)
`
`7 (6/82)
`
`0 (0/82)
`
`0 (0/82)
`
`1 (1/82)
`
`Odds Ratio
`(95% Confidence
`Interval)
`
`
`
`<.0Q1
`
`13.2(4.8-36.3)
`
`.13
`
`<.Q1
`
`9.1 (0.4-192.9)§
`
`21.4(1.2-395.4)§
`
`9.6(1.1-85.2)
`
`1 (1/82)
`
`<.001
`
`19.2(2.3-156.9)
`
`from drug companies to speak, the odds
`ratios (with 95% confidence intervals
`[CIs]) for submitting a formulary re¬
`quest were 1.8 (0.5 to 6.1), 4.0 (1.0 to
`16.8), and 29.1 (3.4 to 246.6), respectively,
`compared with physicians who had never
`accepted money from drug companies
`to speak. Cases were also much more
`likely than controls to have performed
`research sponsored by drug companies:
`seven (70%) of 10 cases who submitted
`two or more requests had performed
`such research, compared with 15 (58%)
`of 26 cases who had submitted only one
`request and 20 (29%) of 69 controls
`(P=.001).
`
`Physician Interactions With Specific
`Drug Companies: Comparisons of
`Case and Other Physicians
`Physicians who had a variety of in¬
`teractions with specific companies were
`more likely than other physicians to have
`requested that drugs manufactured by
`those companies be added to the for¬
`mulary (Table ß24): odds ratios for re¬
`questing a drug made by a specific com¬
`pany were 13.2 (P<.001), 21.4 (P<.01),
`and 9.6 (P=.02), respectively, for phy¬
`sicians who met with pharmaceutical
`representatives from that company, ac¬
`cepted money from that company to
`speak at symposia, or accepted money
`from that company to perform research.
`Too few physicians accepted money from
`specific companies to attend symposia
`to achieve statistical significance. Phy¬
`sicians who accepted money from spe¬
`cific companies for any of the above edu¬
`cational or research activities were much
`more likely than other physicians to have
`requested that drugs manufactured by
`those companies be added to the for¬
`mulary (odds ratio, 19.2; P<.001).
`All nine case physicians and the one
`control physician who had accepted
`money for educational or research pur¬
`poses from the companies making the
`drug requested also had met with phar¬
`maceutical representatives from those
`companies. Thus, compared with phy¬
`sicians who neither met with represen¬
`tatives nor accepted money from the
`specific companies, the odds ratios for
`requesting that that company's drug be
`added to the formulary were 9.9 (95%
`CI, 3.3 to 30.2) for physicians who met
`with representatives from the company
`but did not accept money and 29.7 (95%
`CI, 3.6 to 247.3) for physicians who both
`accepted money and met with repre¬
`sentatives from the company.
`In five multivariable logistic regres¬
`sion models controlling for four physi¬
`cian characteristics (age, gender, ap¬
`pointment in the department of medi¬
`cine, and number of patients seen per
`
`*The types of interactions are defined in the text.
`tFor each type of interaction with specific companies, responses were available on 47 potential
`interactions
`between case physicians and the companies making the 47 drugs they had requested and 82 potential interactions
`between the randomly selected control physicians assigned to each case and the same specific companies. For
`cases, the numerator was the number of case physicians who responded that they had experienced a given type
`of interaction with that specific company, and the denominator was the total number of case physicians responding
`about interactions with the specific companies making the drugs they had requested (n=47). For controls,
`the
`numerator was the number of controls who responded that they had had that type of interaction with the specific
`company making the drug requested by the case to which the control physician had been assigned, and the
`denominator was the total number of controls responding about that type of interaction with the specific companies
`(n=82). These comparisons used a physician's response to a question about an interaction with a company as the
`unit of analysis. When parallel analyses were performed using data comparing the 47 potential interactions of cases
`and the specific companies whose drugs they had requested with all potential
`interactions between controls and
`specific companies (n=483, since 69 controls were each asked about seven specific companies), the results were
`consistent.
`tThe numerator represents different physicians, not the same physician responding to questions about different
`formulary requests he or she had made.
`§Odds ratios were calculated by the method of Fleiss24 for situations in which one of the observations is zero.
`
`Table 4.—Results of Five Multivariable Models of Cases' and Controls'
`Companies*
`
`Interactions With Specific Drug
`
`Odds Ratio (95% Confidence Interval)
`for Submitting a Formulary
`Request for a Drug Made
`by the Specific Company!.
`3.4(1.8-6.6)
`7.9(1.1-55.6)
`3.9(1.2-12.7)
`9.5 (2.6-35.7)
`
`<.001
`.04
`.03
`<.001
`
`<.001
`
`Type of Interaction*
`Met with pharmaceutical representatives
`Accepted money to attend symposia
`Accepted money to speak at symposia
`Accepted money to perform company-sponsored research
`Accepted money to attend symposia, speak at
`symposia, or perform research
`5.7(2.2-14.7)
`*The comparisons used data for the 47 potential interactions of case physicians and the specific companies whose
`drugs they had requested and the 483 potential interactions between control physicians and specific companies (69
`controls were each asked about seven specific companies). Each model
`included four potentially confounding
`factors: physician age, gender, appointment in the department of medicine, and number of patients cared for per
`week. When parallel analyses were performed using only the 82 potential interactions of controls assigned to each
`case with specific companies as the unit of analysis (Table 3), the results were consistent, except that the analyses
`for attending and speaking at symposia could not be done because none of the controls had interacted in that way
`(see Table 3).
` fThe types of specific interactions are defined in the text.
`jwhen the interaction rate of each company with controls was also included in each model (because companies'
`interaction rates differed greatly, from 0% to 60% for meeting with pharmaceutical representatives, for example), the
`odds ratios were 10.8 (4.4, 26.3), 20.4 (1.8, 250.0), 5.2 (1.5,18.5), 13.0 (3.1, 52.6), and 7.4 (2.7, 20.4), respectively.
`patient-based publications indexed in
`many sorts of detailing activities spon¬
`MEDLINE in 1989. Case physicians,
`sored by drug companies in general, but
`however, were more likely to be male
`case physicians reported sharing meals
`(34 [85%] of 40 vs 53 [66%] of80; P=.03),
`worth more than $10 paid for by drug
`in the department of medicine (18 [45%]
`companies more often than did control
`of 40 vs 20 [25%] of 80; P=.04), and re¬
`physicians (Table 2).
`ported to be caring for more patients
`Case physicians were somewhat more
`likely (P=.06) than control physicians to
`per week on average (median, 44 vs 31,
`respectively; P=.01).
`have accepted money from drug com¬
`panies to pay for their personal travel or
`Physician Interactions With Drug
`lodging to attend educational symposia,
`Companies in General
`and they were much more likely than
`Most physicians (85 [81%] of 105) met
`control physicians (P<.001) to have ac¬
`with pharmaceutical
`cepted money to speak at such symposia
`representatives.
`Case and control physicians did not dif¬
`(Table 2). For physicians who rarely,
`fer in the frequency of their exposure to
`occasionally, or often accepted money
`
`Downloaded from
`
`www.jama.com
`
` at University of California - San Diego on October 8, 2009
`
`

`
`Table 5.—Matched Analysis of Potential Interactions of Case Physicians With Companies Manufacturing the Drug Requested and With Other Pharmaceutical
`Companies
`
`No. of Concordant Pairs*
`
`I
`
`Case Physicians
`Interacted With
`Neither Company
`
`No. of Discordant Pairs*
`
`Case Physicians
`Interacted Only With
`the Company Making
`Drug Requested
`
`-1
`
`Case Physicians
`Interacted Only With
`Other Companies
`
`Odds Ratio*.
`(95% Confidence
`Interval)
`
`I
`Case Physicians
`Interacted With
`Both Companies
`
`34
`
`Type of Interaction*
`Met with pharmaceutical
`representatives
`Accepted money to attend
`symposia
`Accepted money to speak at
`symposia
`Accepted money to perform
`company-sponsored research
`Accepted money to attend
`symposia, speak at symposia,
`or perform research
`
`101
`
`218
`
`218
`
`185
`
`102
`
`26
`
`23
`
`42
`
`21
`
`12
`
`15
`
`25
`
`4.9 (3.2-7.4)
`
`2.2 (0.8-6.2)
`
`2.2(1.1-4.2)
`
`1.5(0.8-2.9)
`
`1.7(1.0-2.7)
`
`*The types of specific interactions are defined in the text.
`tFor each of the 26 case physicians submitting the one formulary request, there were six possible pairs (one specific company making the drug and six specific companies
`not making the drug); for each of the nine case physicians submitting two requests, there were 10 possible pairs (two specific companies making the drugs and five specific
`companies not making the drug); and for the case physician submitting three requests, there were 12 possible pairs (three specific companies making the drug and four specific
`companies noi making the drug).
`JThe odds ratio was calculated as the ratio of discordant pairs.20 The numerator was the number of instances (pairs) in which a case physician interacted with the company
`making the drug requested but not with other specific companies not making the drug. The denominator was the number of instances in which a case physician did not interact
`with the company making the drug requested but did interact with another specific company not making the drug.
`
`week), each type of interaction with the
`drug companies making the drugs re¬
`quested was independently associated
`(P<.05) with submission of a formulary
`request (Table 4).
`Case physicians who had met with
`pharmaceutical
`representatives from
`specific companies were more likely than
`matched comparison physicians to have
`requested that drugs manufactured by
`those companies be added to the for¬
`mulary (odds ratio, 8.0 [95% CI, 1.4 to
`45.9]). For accepting money from spe¬
`cific companies to attend symposia, speak
`at symposia, or perform research, the
`odds (95% CI) ofrequesting a drug made
`by those companies were 5.0 (0.5 to 46.5),
`2.0 (0.4 to 10.6), and 9.0 (1.0 to 77.5),
`respectively, compared with matched
`comparison physicians.
`Physician Interactions With Specific
`Drug Companies: Comparisons
`Among Case Physicians
`Physicians who had met with phar¬
`maceutical representatives from specific
`companies or who had accepted money
`from specific companies to attend sym¬
`posia, speak at symposia, or perform
`research were more likely to have re¬
`quested drugs manufactured by those
`companies than they were to have re¬
`quested drugs manufactured by other
`companies (Table 5).
`Similar proportions of case physicians
`requested a drug of major or moderate
`therapeutic advantage (as opposed to
`little or no therapeutic advantage) when
`they accepted funding from the drug
`companies manufacturing the drugs or
`met with the company's representatives
`without accepting funding (40% and 44%,
`respectively) as when they did not ac¬
`cept funding from the companies or meet
`
`with the company's representatives
`(52%; P=.8). More cases requested a drug
`approved by the Food and Drug Ad¬
`ministration in the past year (as opposed
`to longer than a year) when they ac¬
`cepted funding from the manufacturer
`than when they met with the manufac¬
`turer's representative without accept¬
`ing funding or when they neither ac¬
`cepted funding nor met with the repre¬
`sentative (89% compared with 40% and
`41%, respectively; P=.03).
`COMMENT
`These results demonstrate a strong
`association between a physician behav¬
`ior, requesting that a drug be added to
`a hospital formulary, and the doctors'
`interactions with drug companies. Four
`aspects of this association deserve em¬
`phasis.
`First, the association was strong: de¬
`pending on the type of interaction, phy¬
`sicians who interacted with drug com¬
`panies were from nine to 21 times more
`likely than other physicians to have re¬
`quested a drug made by the company
`with whom they had interacted.
`Second, the association was consis¬
`tent for many types of physician-com¬
`including sharing
`pany interactions,
`meals paid for by companies, meeting
`with detailing representatives of the
`companies making the drugs requested,
`and accepting money from companies
`for educational or research activities.
`the association was specific:
`Third,
`physicians who had interacted with spe¬
`cific companies were more likely to re¬
`quest drugs made by those companies
`than they were drugs made by other
`companies, and than were physicians
`who had not interacted with those com¬
`panies.
`
`Fourth, the association was indepen¬
`dent of many confounding factors:
`in
`matched and multivariable analyses ac¬
`counting for physician and practice char¬
`acteristics, the association remained con¬
`sistent.
`Comparison With Past Studies
`Past work has suggested that physi¬
`cians' behavior is associated with and may
`be affected by their interactions with drug
`companies. Information presented by 11
`physicians at company-sponsored sym¬
`posia was biased in the company's fa¬
`vor,19 and some physicians have reported
`changing their practice after discussions
`with pharmaceutical representatives or
`attending symposia.16·17 Promotional at¬
`tributes ofpublished company-sponsored
`symposia have been demonstrated.25 Also,
`in-hospital use of two drugs increased
`after physicians were invited to attend
`symposia sponsored by the manufactur¬
`ing companies.w Results of the present
`study are consistent with and amplify
`those ofthe