What impact does pharmaceutical promotion have on behaviour?

This is both the most difficult area to research and the most important. Doctors may not be aware of how much promotion they are exposed to. Thus, as much as possible, research on the effect of promotion on behaviour should avoid relying on self-report data to show causal relationships. Self-report data are appropriate for finding out what people think is happening, or how they want to present themselves to others, but in this area, that may be far from the reality.

This review looks at the evidence for several different possible effects of promotion on behaviour. These are the impact of promotion on individual prescribing behaviour, on overall drug sales, and on requests for formulary additions; the effect of direct to consumer advertising (DTCA) on consumers’ decisions, the effect of promotion on the content of continuing medical education (CME) courses, and the impact of industry funding on research outcomes.

3.1 Impact of promotion on individual prescribing practices

The ideal study would use data on actual prescribing before and after documented exposure to promotion whilst keeping other influences on prescribing constant. This situation is very difficult to create in real life. Instead researchers have often relied on self-assessments of exposure to or reliance on promotion, and self-reported prescribing. They have also found it hard to measure changes over time, and have often used data on different practitioners to guess at changes over time. This has serious limitations.

This section examines various approaches to the question of how promotion affects individual prescribing.

The first approach uses self-reported reasons for changes in prescribing, and investigates whether promotion is one of these self-reported reasons. (3870), (2940), (3560), (1200), (1100). The prescribing changes might be measured (i.e. externally verified) or they might be self-reported. Ideally they are specific changes in prescribing particular drugs. It is inherent in this approach that the exposure to, and relative influence of promotion is self-assessed. Consequently with this approach it is difficult to do more than give practitioners opportunities to present researchers with their self-image as people who are, or are not, influenced by promotion.

Stronger evidence for some kind of association between promotion and individual prescribing decisions comes from studies that look for associations between variations in prescribing decisions and variations in reliance on promotion. In these studies doctors are asked general questions like how reliable or useful promotional information is, and/or whether it is important in their prescribing decisions. Prescribing by those who give more positive assessments of promotion is then compared with prescribing by those who are more sceptical. Prescribing data are either self-assessed, elicited in response to hypothetical situations, or real prescribing data are used. There is strong consensus from these studies that doctors who rely more on promotion are heavier or less rational prescribers, or adopt new medicines earlier than those who rely less on promotion (760), (3740), (3510), (780), (750), (2070), (3970), (16460), (2000), (4050). However, this kind of research cannot show a causal connection between promotion and prescribing. The results may be due to other doctor characteristics. They do not prove that if these doctors relied less on promotion their prescribing would improve.

A third group of studies look at different levels of exposure to promotion (between doctors or over time), and prescribing. These studies look at specific drugs, and the promotion related to them. These are the best kind of evidence that promotion actually causes changes in individual prescribing behaviour. The studies described by Peay & Peay (4500), Orlowski & Wateska (1720) and Gonul et al. (21650) are rather convincing and worth replicating in other situations and with other drugs. This would considerably strengthen the argument that exposure to promotion causes prescribing changes. Other studies of this kind, such as (10430), (1650) (15670), (15970), (15650) are also somewhat suggestive, but have methodological shortcomings or the methods are not described fully enough to allow evaluation.

This section ends with a discussion of the effect of samples on prescribing. This is discussed separately because it presents different methodological challenges, so different approaches have been used.

3.2 Self-reported reasons for prescribing changes

Taylor and Bond used real prescribing data (3870). They asked 201 doctors in Scotland to fill out duplicate prescriptions, which included details about perceived influences on prescribing. Of the 161,266 prescriptions most were either repeat prescriptions or drugs that the prescribers had prescribed in the past. New drugs formed a median of 3.5% of the prescription items per doctor. Sales representatives were mentioned as influences for 20% of new drugs added to doctors’ prescribing repertoires during the research period. Sales representatives were more likely to be listed as an influence on the prescribing of drugs used short-term. It is difficult to know how generalisable these findings might be. They may depend on the type of drugs that are being heavily marketed at the time, and other influences on prescribing at the time (Taylor and Bond note the concurrent introduction of a ‘limited list’).

Dasta et al. (2940) also had objective evidence of prescribing. Their study, partly supported by Abbott Laboratories, looked at sources of doctors’ information about clarithromycin, a new antibiotic. The study was carried out in one inpatient and several outpatient medical care facilities. In the hospital, doctors who placed an order for clarithromycin were contacted by phone, and in the outpatient facilities doctors were sent a questionnaire when a prescription for clarithromycin, written by them, was presented at the pharmacy. In the hospital 65% of the doctors who prescribed clarithromycin reported not having had contact with a sales representative, and had never received or used samples of the drug at the time of the first interview. Eighteen percent of outpatient prescribers had first heard about clarithromycin from a commercial source.

Peay & Peay (3560) looked at the role of different information sources in specialists’ decisions to adopt new drugs. Each specialist was asked about his or her general drug adoption practices and also about one of eight target drugs. The results suggest that commercial sources of information are relatively unimportant to specialists, with only 4.7% of respondents naming any commercial source as the most influential in their decision to first prescribe the target drug.

These studies are better at identifying the influence of promotion than those that ask for a general self-assessment of the influence of promotion, because they isolate particular prescribing decisions. But they cannot be taken at face value because they rely on doctors’ own assessments of what has influenced their decisions.

Two studies, by Curry & Putnam (1200) and Lurie et al. (1100), relied entirely on self-assessments of reasons for prescribing changes . The former found that only 0.3% of their respondents (practising doctors in Maritime Canada) reported changing their practice in the last year because of discussions with sales representatives. The latter surveyed faculty at seven university teaching hospitals in the US, and house staff in two of the teaching programmes, about their interactions with pharmaceutical representatives. Twenty-five percent of the faculty and 32% of the residents reported that they had changed their practice at least once in the last year as a result of a discussion with a sales representative.

CONCLUSION: Doctors rarely acknowledge that promotion has influenced them to make specific prescribing changes. Specialists tend to report that promotion has less effect on them.

3.3 Prescribing by those who rely on commercial information

One study found no link between prescribing characteristics and self-reported reliance on promotion. Hemminki (1890) found no differences in observed frequency of prescribing psychotropic drugs between doctors who chose journals, textbooks or commercial sources as their main information source.

However most studies have found links. Mapes (760) found that doctors who reported relying on pharmaceutical industry literature were more likely to prescribe three or more drugs that frequently cause side effects. Conservative doctors, who did not endorse the industry as a source of post-graduate knowledge, prescribed drugs that were newer, more effective and safer. This study used prescribing data routinely collected by the Department of Health in the UK. Bower and Burkett (3740) found that family physicians who reported relying less on drug sales representatives for information were likely to prescribe more generic drugs, as were residency trained doctors, and regular readers of the New England Journal of Medicine. The self-assessed ability to recognise generic drug names was also highest amongst these doctors, those who relied least on journal advertising, and regular readers of The Medical Letter. Caudill et al. (3510) found, among primary care doctors in Kentucky, that those who rated information provided by sales representatives highly (as credible, available, and applicable) and reported using it more, chose more expensive prescribing options in response to three clinical vignettes. This study had a low response rate. The study reported in Becker et al. (780) and Stolley et al. (750) used self-report data on attitudes to and reliance on promotion; expert ratings of responses to questions about prescribing for certain conditions, and knowledge about certain drugs; and analysis of actual prescribing of chloramphenicol (an antibiotic which should not be widely used). They found that doctors who relied on journal articles and tended to be disdainful of journal advertisements, sales representatives, and retail pharmacists as sources of information received higher ratings from the experts and prescribed less chloramphenicol. Better prescribers were more positive about generics, and gave other indications of a less positive attitude towards the industry and promotion, than other doctors. A single question, about whether sales representatives were good sources of prescribing information about new drugs, produced the highest correlation with prescribing appropriateness. Berings et al. (4600) found that Belgian doctors in their study who felt that commercial sources of information were more important, prescribed more benzodiazepines than those who rated these sources as less important. Their prescribing was observed through the use of special prescription forms provided by the researchers.

In the Netherlands Haayer (2070) presented 8 case studies of hypothetical patients to general practitioners and asked them if they would prescribe medication for this patient, and if so, what they would prescribe. An expert panel assessed the rationality of their prescriptions. The GPs were later interviewed and asked about their use of different sources of information about medicines. Less than half (48%) of the prescribing decisions made were rated as ‘entirely rational’. Differences between doctors accounted for more variance than differences between cases: that is, doctors seem to be more or less rational prescribers, over a range of different conditions. Haayer found that reliance on information provided by the pharmaceutical industry was negatively associated with prescribing rationality. That is, doctors who relied on promotional information wrote less rational prescriptions for the case studies than those who reported relying less on promotion.

Cormack and Howells (3970) surveyed GPs in the UK before and after they attended a course on benzodiazepine prescribing. Their prescribing, adjusted by their number of patients (‘list size’) and the number over 65 years of age, was also analysed using Prescription Pricing Authority data. This produced a very wide range of scores. Doctors were classified as high or low prescribers of benzodiazepines. In interviews low prescribers rated information from drug companies more sceptically than high prescribers.

Williams and Cockerill (16460) found that doctors who reported writing higher numbers of prescriptions per week had more contact with the drug industry (ie, interacting with sales representatives, receiving benefits such as meals or conference fees) and were more likely than others to rate sales representatives and industry-sponsored seminars as important sources of drug information. The first result may have been partly due to higher prescribers being likely to spend more time in medical practice per week than lower prescribers. However Williams et al. note that high volume prescribers reported writing more prescriptions per patient, which adds weight to the idea that these are doctors who prescribe heavily. These results are also presented in Williams, Cockerill and Lowy (950).

There is also evidence that those who rely more on promotion may be older, and are earlier adopters of new drugs. Stross (2000) investigated the reasons for changes in the management of chronic airways obstruction between 1978 and 1983 in small community hospitals. Using chart audits he identified a significant change in management of the condition during this period. He interviewed doctors who had treated patients at these hospitals in the study years. Older doctors reported relying more on sales representatives as a source of information for changing patient management. Stross looked at decisions to adopt three types of medicines (single-agent bronchodilators, beta-sympathomimetic agents, and corticosteroid aerosols). For the last two, around 35% of doctors said sales representatives were their most important source of information in decisions to adopt the drugs. Early adopters of the changes were more likely than late adopters to list sales representatives as a major source of information. This study is useful in that it relies on significant observed changes in prescribing, which the researcher identified.

Strickland-Hodge and Jepson (4050) compared the characteristics of the first and last 100 doctors to prescribe cimetidine in one area in the UK. Although their response rate was rather low, they found that earlier prescribers rated commercial sources of information (sales representatives, advertisements in medical journals, direct mail, MIMS, and controlled circulation journals) significantly higher as information sources than late prescribers. Early prescribers reported reading more of their direct mail than late prescribers and reading fewer journals.

Together these studies provide convincing evidence that doctors who regard promotion more highly, and report relying on it more as a source of information about drugs, prescribe more drugs, prescribe less rationally, and prescribe new drugs earlier than other doctors. However they can only provide circumstantial evidence for a causal link between promotion and individual prescribing. Other doctor characteristics, such as attitudes to risk, beliefs about clinical experience and evidence, views of new technologies, and academic inclination or ability may be behind these results. For example, doctors who believe that their clinical experience is more important than scientific evidence may be less likely to respond to evidence presented in journals, therefore be more dependent on other sources of information such as promotion, and less likely to prescribe rationally (i.e., according to the evidence). Alternatively less academically inclined doctors may not read journals, may rely on advertising because it is very accessible, and may also prescribe in less than optimal ways. The main problem with these studies is that they cannot show that doctors who report relying on promotion would prescribe differently or more rationally, if they did not rely on promotion.

CONCLUSION: Doctors who report relying more on promotion prescribe less appropriately, prescribe more often, or adopt new drugs more quickly.

3.4 Prescribing and exposure to promotion

Peay & Peay in 1988 (4500) clearly showed a relationship between seeing sales representatives and prescribing one new drug, and are often quoted by others. They interviewed 124 doctors in private practice, about their perceptions and use of temazepam, a benzodiazepine hypnotic, and their sources of information about it. The study was done in 1981, approximately a year after temazepam was introduced in Australia. They found that contact with a sales representative about temazepam most consistently predicted a favourable reception of temazepam at various points in the adoption process. Doctors who had seen a sales representative reported earlier awareness of temazepam, prescribed it earlier, were more likely to rate it as a moderate (rather than minor) advance over other drugs, were more likely to have prescribed it, reported prescribing it earlier, and were more likely to prescribe it routinely in preference to other alternatives. Compared to those who saw sales representatives less frequently, those who saw representatives more than once a week were aware of temazepam earlier, prescribed it earlier, and (amongst GPs) were more likely to prescribe it than other alternatives. Peay & Peay found no relationship between doctors’ professional involvement, or involvement in the medical community, and beliefs about temazepam. The study has considerable advantages over those described above. It does not ask doctors to assess themselves whether promotion has affected their decisions. It does not ask them to rate their own level of reliance on commercial information. The question “have you seen a sales representative regarding temazepam?” requests one simple fact that is likely to be easier for doctors to recall than the number of journal advertisements seen, etc. The group of GPs who had seen sales representatives about temazepam may have included more of the commercial information oriented doctors described above, but this is unlikely to account completely for Peay & Peay’s results.

In another important study Orlowski and Wateska (1720) analysed the effect on prescribing of drug company funded, all-expenses paid trips to educational symposia in resort locations. Using the hospital pharmacy inventory, they tracked the use of two drugs within one institution 22 months before and 17 months after each symposium about them. They also collected data on the national usage of these drugs, and informally interviewed the doctors who had gone to the symposia. Most of the doctors said that the symposia would not influence their prescribing, but some said that they might make them think of the drug more and the symposium might convince them of the benefits of the drug. Orlowski and Wateska found a dramatic and statistically highly significant increase in the use of the drugs in the hospital after the relevant symposia. These increases were not reflected in national data, and they did not seem to affect the hospital’s use of alternative drugs. This study provides evidence firstly, that exposure to promotion increases prescribing, and secondly that it can do so whether or not those exposed consider themselves vulnerable to such influence.

A useful study by Gönül et al. (21650) explored the impact of visits by sales representatives and samples, on prescribing. They used data from Scott-Levin Inc. (a company which describes itself as a leading pharmaceutical consulting firm) derived from survey sheets filled in by doctors. These included prescribing, minutes of detailing received for different drugs, and number of samples received, for a ‘typical’ week in each month, from Jan 1989 to December 1994. Gönül et al. looked at one condition and seven drugs used to treat it. Throughout the article it is unclear whether these were different drugs, or different brands of the same drug, and this is a major weakness of the study. Using a multinomial logit model, it appears that exposure to personal selling related to a medicine (visits from sales representatives and samples) increased the probability of that medicine being prescribed (other things being equal). However, the study also showed that excessive detailing or samples did not increase sales further, and that doctors who saw a high proportion of Medicare or Health Maintenance Organisation (HMO) patients were less influenced by promotion. The authors are from marketing schools, and they conclude that the study provides no evidence that personal selling has negative social consequences. There seems little evidence for this in the study. Part of the difficulty in evaluating the conclusion is that it is unclear whether the study examined seven brands of one drug, or seven drugs. The health consequences of changing drug therapy in response to marketing are likely to differ from those of changes in brand.

Research by Walton, a pharmacist and advertising executive, (10430) and (1650) suggests that recall of print advertisements is associated with prescribing. In (10430) results are presented from a study of 1000 doctors in private practice who were shown print advertisements with drug and company names and logos blacked out. They were asked whether they had seen each advertisement before, and were then read a list of the advertised products and asked if they had prescribed or recommended these in the last month. For 95% of the advertisements the percentage of doctors who prescribed them was greater for those aware of the advertisements than for those not aware of them. However the effect of specialty was not controlled for. That is, doctors may be both more likely to notice and recall, and to prescribe, drugs relevant to their specialty. The study in (1650) appears to be a smaller version or subset of that in (10430).

Matalia (3100) reviews a range of advertising industry-related studies that claim to show the effectiveness of print advertising (15670, 15970, 15650). In the first, family practitioners and internists evaluated advertisements. ‘Prescribing data’ were also collected but it is unclear whether these are self-assessments of willingness to prescribe, or actual prescription data. Matalia claims that as non-prescribers became more familiar with the advertisements their willingness to write trial prescriptions increased. It seems from his earlier description that this study assessed correlations between attitudes and familiarity with ads, so he seems to be extrapolating from data collected at one point in time from a range of people, to trends over time. The account of the second study (15970) is somewhat more convincing but again the methods and analysis are not described well enough for proper evaluation. The study was an experiment where different groups of doctors (who had prescribed similar numbers and value of prescriptions in the last six months) were sent identical journals but with varying numbers of advertisements for a mature cardiovascular drug (i.e. one that had been on the market for some time). Those in the group who received the most advertising increasingly prescribed the drug. After 12 months the manufacturers market share was 4% higher in the high intensity and 2.3% higher in the medium intensity group, than in the lower group. The third study was also a kind of experiment. Companies stopped all promotion for four products from 9 months before the study. Four advertisements were designed for the study and placed in half the copies of eight journals. Doctors were interviewed, and those who had received the ads were more likely to recall the products than those who had not. However, prescribing was not analysed: the outcome variable was simply recall of the products.

CONCLUSION: Exposure to promotion influences prescribing more than some doctors realise.

3.5 Exploring the impact of samples on prescribing

There is little literature on the effect of samples on prescribing. Backer et al. (21620) report an ethnographic study of 18 medical practices. At least four weeks of fieldwork were done in each practice. Samples were used in 19.8% of the 1588 patient encounters observed. This varied widely between practices (range 4% to 39%) and also between doctors within each practice. Reasons given for using samples included to test for efficacy and tolerability, to offer temporary relief or convenience, and/or to reduce costs to patients.

In Morelli and Koenigsberg’s study (1120) samples which were dispensed as new medication for chronic problems were accompanied by a prescription for the same brand 48% of the time. This finding is hard to interpret, but it may suggest that the availability of a sample influences the choice of brand prescribed. This area needs further investigation.

Chew et al. (21590) used three hypothetical case studies and asked their respondents (131 general medicine and family physicians) which drug they would prescribe. They were then given a list of samples available and asked whether they would prescribe their drug of choice, or give a sample of another drug. For a patient with hypertension (and no health insurance) almost all respondents (92%) ideally chose a diuretic or beta-blocker (consistent with practice guidelines). However when samples were available, 27% (35 doctors) said they would dispense a sample. In almost all of these cases the sample was a different class of drug (e.g. ACE inhibitor or calcium channel blocker). Almost all of those who would give a sample (97%) said avoiding cost to the patient was an important or very important reason for their choice. A follow-up scenario in which the patient returns, with their hypertension well controlled on the sample drug, and now with health insurance, was presented. Of the 35 doctors who had said they would dispense a sample, 24 would now write a prescription for the sample drug, to avoid switching the patient. If this reflects real behaviour, it suggests that in some circumstances drug samples may strongly influence prescribing.

CONCLUSION: Samples appear to influence prescribing but more research is required on this issue.

Summary

Doctors’ own assessments of whether promotion affects their prescribing are of limited value in establishing whether this is the case.

The research clearly shows that doctors who report relying more on commercial information, prescribe more heavily, less rationally, and adopt new drugs more quickly. Some researchers have interpreted this finding as showing that ‘relying on drug company information increases prescribing’. This interpretation is not justified by evidence from these studies. These studies cannot show whether doctors would prescribe differently if their level of reliance on promotion were to change. Some doctors may have characteristics (such as attitudes, skills) that lead to both reliance on promotion, and heavy or irrational prescribing.

The studies that look at different levels of exposure to promotion (between prescribers or over time) and prescribing provide more convincing evidence that promotion changes behaviour. Further research using this kind of approach would be valuable. Simply replicating the Peay & Peay study in another place, using another drug, would strengthen the evidence considerably: similar findings would add substantial weight to the argument that contact with sales representatives does change prescribing behaviour. In addition, other studies that look at prescribing changes after exposure to promotion would be very useful. Cormack and Howells (3970) and Strickland-Hodge and Jepson (4050) used prescribing data from the Prescription Pricing Authority in the UK. Such data could be utilised further to observe prescribing changes, for example, before and after visits by sales representatives. Other countries where all or most prescriptions are subsidised by the government, such as Australia and New Zealand, have similar data available.

Samples appear to influence prescribing, but this has received little attention and needs further study. Other literature (3850) (990) (1120) has highlighted the widespread misuse of samples by health professionals, sales representatives and others, but ironically less is known about their use for patients.

Marketing literature tends to assume that evidence of behaviour changes is a good outcome: it shows investment in advertising is worthwhile. The public health and medical based literature tends to assume that higher prescribing levels of what is judged to be a suboptimal drug will lead to worse health outcomes. Some of the research suggesting that doctors who rely heavily on promotion prescribe differently does explicitly look at the quality of the prescribing (eg, Haayer’s (2070) use of an expert panel, or the extent of chloramphenicol prescribing in the study by Becker et al. (780)). Such measures of appropriateness need to be utilised more.

3.6 Impact of promotion on overall sales

Some studies have investigated whether promotion affects overall consumption or sales of medicines. There are several ways to do this. Some studies (850, 21370, 3560, 4200, 21830) have simply observed changes that occurred before, during or after promotional activities. These studies are relatively simple and inexpensive but can provide quite convincing evidence if appropriate study periods are chosen and sales or consumption data are available. Other studies have used econometric modelling to investigate the relationship between promotion and sales over time (1880, 3760, 21330). These methods are so complex and sophisticated that it is hard for non-econometricians to judge whether the models are appropriate. Some studies in this area look at levels of promotion and sales of a range of drugs (1540, 4080). These cannot separate effects of promotion on sales from effects of sales on promotion. That is, they ignore the fact that companies may heavily promote their most popular drugs because the robust sales have enabled them to pay for more promotion. In theory it is more likely that the relationship between promotion and sales is not a one-way causation link but a two-way negative feedback loop—more promotion leads to higher sales which leads to more promotion.

Cleary’s small study (850) looked at what happened when the level of promotion varied naturally over time, when a sales representative was away on a sales training course. He examined trends in numbers of new prescriptions for three third-generation antibiotics in one hospital. He found that when the sales representative was away the numbers of new prescriptions for this product dropped. This did not happen to the other products studied, and there was no correlation between the pattern in this hospital or regional or national sales. This study has the advantage of avoiding any effect of sales on promotion: ie the change in level of promotion was not a result of changes in the level of sales. Similarly Dieperink and Drogemuller (21370) report their investigation of the reasons for a dramatic increase in the use of an atypical antipsychotic agent in their Minneapolis hospital. The most plausible explanation for this was a Grand Rounds presentation sponsored by the manufacturer of the product.

A small study reported in a letter to the Lancet by Suresh et al. (13560), suggests that useful drugs may be relatively underutilised if they are not promoted. They describe the under-use of adenosine, an effective first-line treatment for supraventricular tachycardia, until it started to be marketed commercially in 1991. The drug was available and cheap, and there was good evidence of its usefulness, but it was underutilised until an advertising campaign was carried out.

Stern (4200) examined the number of visit to doctors where topical tretinoin was prescribed, and the number of articles in the popular press and medical publications discussing its use. In 1988 a highly publicised study suggest that topical tretinoin improved the appearance of aged skin, and it was prescribed at an increasing number of consultations in the US after this. Most of these prescriptions were probably for the unlabelled unapproved use of tretinoin to treat the effects of aging. Stern’s time series data are sporadic but like Cleary’s and Suresh’s work the paper suggests a link between promotion and overall sales.

Eichner and Maronick (21830) analysed correlations between sales, expenditure on DTCA and patient visits to doctors for four groups of drugs between 1996 and 1998. They concluded that DTCA campaigns had variable success, and that factors other than DTCA (such as product characteristics and other marketing efforts) were important in determining sales levels, but that DTCA did seem to increase patient visits to doctors for advertised conditions.

Mackowiak and Gagnon (1880) used econometric modelling to investigate the relationship between promotion and demand for medicines. They looked at diuretics and benzodiazepines from 1977 to 1981, to investigate how overall expenditure on promotion, and individual company promotional expenditure affected demand for a group of drugs (ie overall market size), and how individual company promotional expenditure affected demand for a particular drug (i.e. market share). They used IMS data on the extent of sales representative activity and the extent of journal advertising, and converted these into estimates of expenditure. Advertising agency fees were not included and this seems a significant omission. IMS also provided data on number of new prescriptions for the products studied. Using ARIMA (Auto Regression Integrated Moving Average) modelling, they could find no relationships between promotional expenditure and demand in any of the three areas outlined above. They suggest that this may be due to limitations of the methodology, or it could suggest that companies are spending so much on advertising that they are getting little marginal return for extra dollars spent. Although this study clearly has methodological limitations (such as the choice of 2 drugs whose markets were not very dynamic in the period studied), it seems to make a minimum of unwarranted assumptions.

In another econometric study, Basara (3760) looked at the impact of a campaign of direct to consumer advertisements (DTCA) on sales of a new medicine. She used a quasi-experimental, interrupted time-series research design, comparing the number of new prescriptions before, during and after a real-life DTCA campaign. The paper includes a very long description of the complex analytical method used. Basara concluded that the campaign significantly increased the number of new prescriptions for the product. This effect appeared immediately and then tapered off over the campaign. The number of new prescriptions decreased after the campaign ended. Basara is an employee of Rhone-Poulenc Rorer Pharmaceuticals.

Similarly, using data on expenditure on advertising (from Competitive Media Reporting) and prescribing (from the National Ambulatory Medical Care Survey), Zachry et al. (21330) found positive correlations for some drugs and classes, but not for others. They used quasi-experimental time series techniques. There were significant positive relationships between advertising expenditure and the number of prescriptions written for Zocor and Claritin, but a negative relationship between advertising for acid-peptic disorder medications and prescriptions for Zantac.

Krupka and Vener (1540) compared advertising in the New England Journal of Medicine and the JAMA in 1972, 1977 and 1982, with the number of prescriptions filled for the 15 most advertised drugs in 1972, 73, 77, 78, and 82. They found that about a fifth of the 15 most advertised drugs were also one of the leading 15 drugs in terms of the number of prescriptions filled for the five years analysed. Ten of the fifteen most advertised drugs in 1972 had advanced their ranking in terms of prescription numbers between 1972 and 1973, and two were in the same position. Dajda (4080) plotted the number of advertisements received in three GP practices in Swansea by therapeutic group, and the number of prescriptions written for drugs in these groups. He found a high correlation.

The ideal way to investigate this area would be, as (1880) suggest, to ask manufacturers to experimentally vary promotion over regions and times, monitor the effect of this and publish the results. It is possible that drug companies have done many such studies but not published them. In the absence of such data, the published studies do provide considerable circumstantial evidence for a positive, but not always consistent, association between promotion and overall sales.

CONCLUSION: Increased promotion is associated with increased sales.

3.7 Impact of promotion and industry funding on requests for formulary additions

Two studies in the database address this topic. One relies on self-report and is not very useful. The other is an important and useful study that independently assessed relationships with industry and requests for formulary additions.

Lurie’s study (1100) of faculty at seven university teaching hospitals in the US, and house staff in two of the teaching programmes, found that 20% of faculty and 4% of residents reported recommending additions to hospital formularies at least once in the last year at the suggestion of a sales representative. However the number of staff who recommended additions for other reasons is unknown.

Chren and Landefeld (4440) showed that doctors who requested that drugs be added to a hospital formulary were more likely to have received funding from companies than other doctors. They independently observed requests for formulary additions at a university hospital between Jan 1989 and Oct 1990. The 40 doctors who made these requests, plus 80 randomly selected doctors who acted as controls, were sent a survey asking about their demographic characteristics and their relationships with the pharmaceutical industry, such as acceptance of money to go to educational symposia, speaking at symposia, and receipt of research funding. They found that doctors who interacted with a company were between 9 and 21 times more likely than other doctors to have requested that a drug made by that company be added to the formulary. The relationship between funding from companies and requests for formulary additions was strong and consistent, and independent of many confounding factors. Chren and Landefeld note that they did not establish causality: and that it could be that doctors learn of a new drug, request its addition to the formulary and then interact with the company. However they suggest that this scenario is unlikely because most requested drug represented little or no therapeutic advantage over other drugs already on the formulary.

This study provides good evidence that association with a pharmaceutical company – such as receiving research funds – leads to requests for formulary additions. This finding is important because formularies determine the drugs available in a hospital and are therefore likely to affect other doctors’ prescribing habits both in the hospital and when doctors in training leave the hospital and set up an independent practice.

CONCLUSION: Funding for doctors from drug companies increases requests for drugs to be added to hospital formularies.

3.8 DTCA and consumers’ decisions

For a good review of published and unpublished evidence about the impact of DTCA see Mintzes et al., Volume II, (21490).

Everett (3200) asked 238 people in Denver, USA to respond to a hypothetical situation in which they had back pain and saw an advertisement for a prescription-only muscle pain reliever. About one-third of the sample said they would ask their doctor for the drug and about 5% said they would change doctor if s/he did not prescribe it. Those who were less educated were more likely to say they would tell the doctor they had seen the advertisement and ask her/him to prescribe the drug. Bell et al. (21560) report on a survey of Sacramento adults’ anticipated responses to a hypothetical situation in which a doctor denies their request for an advertised drug. Nearly half the sample (46%) said they would be disappointed, 25% would attempt to persuade the doctor to prescribe the drug, 24% would seek the prescription elsewhere, and 15% would consider leaving the doctor. Nearly half the sample (47%) said they would not be disappointed, and would take no action. Those who would take action were more likely to rate their doctors’ communication skills as poor, more positive about DTCA and more (unduly) confident about government regulation of DTCA. These studies rely completely on self-report, in response to hypothetical situations. It is very difficult to know if consumers would respond in this way in reality, especially since important contextual factors, such as the doctors’ explanation of why s/he would not prescribe the drug, are excluded.

In a somewhat more realistic study, Perri and Dickson (3350) sent fake advertisements for fictitious prescription medicines through the mail to 200 patients who had scheduled regular appointments with their doctors. They used the advertisements developed by Morris (18240) and (4150), which he had found to produce the highest knowledge and recall scores. The four doctors treating the patients in the study knew about the advertisements, acted as if the drugs were real, and recorded patient behaviour. One hundred and fifty-five patients were observed by doctors. Thirteen made general comments or asked general questions about the medicines, but none made requests for the medicine. Ninety-four patients also completed questionnaires, which showed that those with chronic medical conditions were more receptive to the advertisements and had more favourable attitudes. The four doctors in the study felt that the advertisements had had no negative effect on their relationships with the patients. However this result may have been different if patients had requested or demanded the drugs. The key advantage of this study is that it observes actual patient behaviour in response to the advertisements rather than reported attitudes or behaviour. Using fictitious drugs also means that it is clear that the effect came from the mailed advertisements because there was no other advertising for these medicines.

Three studies used different ways of measuring real responses to real DTC advertisements. In Prevention magazine’s survey of consumers [(15410), 2000-2001 edition] 32% of consumers who had seen a DTC advertisement had talked to their doctor about an advertised medicine. Twenty-six percent of these had asked for a prescription for the advertised medicine. Of these, 71% received a prescription for it, and 10% received a prescription for another medicine. In Bell et al.’s study of Sacramento adults (22040) 19% reported having asked for a prescription, and 35% having asked a doctor for more information, as a result of a DTC advertisement. One difficulty with this kind of study is that it is unclear how much DTCA has brought about this situation. For example, even without DTCA some patients ask their doctors for drugs they have heard about from friends etc, and some of the prescriptions which were reported in the Prevention magazine study might have been written with or without DTCA.

Mintzes et al. (21270) analysed a sample of 1431 visits to primary care physicians in one Canadian and one US city. They found patients requested prescriptions in 12% of visits, and 42% of these requests were for products advertised to consumers. The 50 drugs with the highest US advertising budgets, plus those noted as advertised to the public, in a Canadian medical journal were defined as ‘advertised’. The authors found that patients who requested a prescription were more likely to receive one than those who did not (after controlling for health status, socio-economic status, demographics and doctor characteristics). Doctors were ambivalent about the choice of treatment in 50% of cases where the patient requested an advertised drug versus 12% of the time when no request was made. Although this study suggests links between advertising, consumer demand and suboptimal care, it is by no means conclusive. It is unclear how many of the patient requests were prompted by advertising. Advertised drugs may differ from unadvertised drugs in other ways (e.g. they may be for more common conditions, be newer) and this could make patients more likely to request them.

Further research is needed to monitor the impact of DTC advertising, particularly on overall consumption of advertised drugs, non-advertised drugs, and non-drug treatments for health problems.

CONCLUSION: DTCA is associated with increased requests from patients for drugs, and some evidence suggests that when doctors respond positively to these requests they are ambivalent about the product they are prescribing.

3.9 Impact of sponsorship on content of CME courses

Bowman (1740) analysed the content of two continuing medical education sessions on calcium channel blockers, funded by different companies, and taught by faculty members. In one of the courses the funding company’s drug was mentioned many more times than other drugs. In both courses the clinical effect ascribed to the funding company’s drug were more positive. There were few comparative statements made, but most favoured the funding company’s drug. This bias was in spite of university policies being instituted between the courses that required the institution rather than the company to control the course content. Bowman and Pearle (1730) then examined self-reported changes in prescribing patterns related to three company-funded CME courses. The method they used is not very satisfactory. They attempted to ask course participants before, and six months after each course, about their prescribing of the group of drugs covered in the course. For two courses there was no matching of responses from individuals pre and post the course were not matched, and the response rates were not high. Bowman and Pearle conclude that in all three courses the sponsoring company’s drug had the greatest increase in absolute terms. However some increases occurred in prescribing of other company’s drugs. This study is limited by its reliance on self-report instead of prescribing data. Participants may have wanted to please the authors by saying they prescribe more of the drug that was presented as the best at the course, if the authors were also the course organisers (this is unclear in the papers).

CONCLUSION: Sponsorship may affect the content of CME. More research is needed to examine this.

3.10 Impact of industry funding on research

a. Extent of industry funding

Many authors document considerable reliance on industry funding for medical research. This also appears to be increasing over time. Massie and Rothenburg (5270) surveyed authors of papers on the medical treatment of angina. Sixty-nine percent of the studies were funded by the pharmaceutical industry and 45% of the authors would have done the study without industry funding. Anderson et al. (5930) found that, over time, it has become increasingly common for clinical trials of second line agents for rheumatoid arthritis to be supported by the pharmaceutical industry. In 1945-1969 they found no published studies that were fully funded by the industry, but by 1980-89, 61% of studies were. Kunin (4410) found 39% of his respondents (members of the Infectious Diseases Society of America) had obtained research funds from pharmaceutical companies. This sum accounted for 34% of the funding reported. About half the researchers felt that they needed industry support for their work. Dorman et al. (21770) found that the proportion of published trials on acute stroke which were apparently supported by the industry increased substantially from 1955 to 1995, from 0 to 54%. They also concluded that descriptions of the nature and extent of industry involvement were poor, and consequently 54% could underestimate the real level of industry involvement.

Blumenthal and colleagues have documented extensive relationships between industry and university life sciences faculty in the US. They have also shown that such links lead to more secrecy around research findings. In 1986 they surveyed faculty members involved in biotechnology in 40 major universities in the US (1760). Those who had received industry funding for their work were found to publish more, patent more, participate in administration and professional activities more, and to earn more. They were also four times as likely to report that their work had resulted in trade secrets, and four times as likely to say that commercial considerations had influenced their choice of research projects, as biotech faculty members without industrial support. Most faculty, with or without industry support, agreed that relationships between industry and universities led to a risk of shifting too much emphasis toward applied research.

Similarly, Krimsky et al. (1960) describe extensive relationships between academic scientists and biotechnology companies. They developed a database of 889 US and Canadian biotechnology companies, and 832 scientists with whom they had formal ties. They found that at least 37% of members of the National Academy of Sciences (a group that provides advice to Congress and other government bodies) had formal ties with biotechnology companies.

In 1994 the Blumenthal group surveyed both faculty about relationships with industry, and industry about their relationships with universities. In their work ‘industry’ refers not only to the pharmaceutical industry, but to other life-science related sectors, such as agriculture. In (4210) the group describe the survey of agricultural, chemical and pharmaceutical companies in the US about their links with academic institutions. More than 90% of the 210 companies they surveyed had some relationship with academia. The most common was the use of faculty members as consultants. From these results the authors estimated that these industries as a whole supported 6000 research projects at a cost of $1.5 billion. Of the respondents, 34% of companies reported disputes with their academic partners over intellectual property, 82% sometimes required academics to keep information secret until a patent application was filed, and 47% occasionally required secrecy longer. In (4270, 4220, 4280) the Blumenthal group reported results of their survey of over 2000 faculty members from the 50 US universities that received the most NIH funding in 1993. In (4270) the group found that nearly 20% of the scientists reported having delayed publication of their results for more than six months for a commercial reason, and 8.9% had refused to share results with other university scientists. Those who had received industry funding were more likely to report having delayed publication for commercial reasons, but those who relied more on industry (i.e. obtained a higher proportion of their funding from industry) were less likely to report having refused to share results. In (4220) the group reported that 28% of the faculty surveyed had received research funding from industry, that the receipt of industry support was more common in clinical than non-clinical departments, and that industry had supplied 8.9% of all research funds (excluding overheads). Those who had received industry funding for their research published more, participated more in administrative activities and were more commercially active than others. But those receiving more than two-thirds of their research support from industry were less academically productive than those receiving lower levels of support.

In 1998, Campbell and colleagues from the Blumenthal group (4280) described the extent of research related gifts given to academic life scientists by companies. Twenty-four percent of respondents had been given biomaterials, 15% discretionary funds, 11% equipment, and 11% travel to professional meetings. Thirty-two percent of respondents thought that the donors expected to review reports and articles before publication, as a consequence of the gift.

Choudhry et al. (21440) investigated relationships between authors of clinical practice guidelines and the pharmaceutical industry. They looked at 44 guidelines endorsed by European or North American professional societies. Eighty seven percent of the authors they contacted had some relationship with industry, 59% had relationships with companies whose products were considered or included in guidelines, and almost all of these pre-dated the guidelines. In 42 of the 44 guidelines no declarations were made about potential conflicts of interest.

CONCLUSION: The percentage of research studies funded by drug companies has increased over the past 50 years. Funding of research by drug companies is associated with influence over choice of topic, secrecy, delayed publication for commercial reasons and conflict of interest problems for authors of guidelines.

b. The effect of industry funding on published results

Several studies explore this area. Some (2540, 21680, 21730, 21800, 5550, 21910) suggest different mechanisms by which the published evidence on drugs is likely to over-estimate their benefits. One of these (2540) also suggests that company funded research is less likely to be published than non-company funded research. Many studies (1390, 21760, 21670, 21700, 14870, 2820, 1150, 30, 21790, 21810, 4520, 22060, 21940, 21750), show that funded studies are more likely to present positive results about the study drug. One study (21950) examines the effect of company research sponsorship on what kind of research is done.


c. Is there an association between funding source and publication status?

Several studies have looked at whether the source of funding for studies affects their publication status. If drug company funded studies with negative results are less likely to be published this could lead to an over-estimate of treatment effects or risk-factor associations in published work, and in meta-analyses that rely only on published work.

Easterbrook et al. (2540) attempted to follow up studies that had been approved by the Oxford Regional Ethics Committee. They found that drug company sponsored clinical trials were significantly less likely to be published or presented than unfunded studies. Stern and Simes (21680) successfully followed up 520 studies out of 748 submitted to a Sydney hospital ethics committee in 1979-1988. In their study pharmaceutical industry funding was not a statistically significant predictor of time to publication. Ioannidis (21730) looked at Phase 2 and Phase 3 trials related to HIV treatments and did not find that the source of funding affected the time it took the results to appear in peer reviewed literature. Dickersin et al. (21800) followed up studies approved by institutional review boards at two centres. The publication rate at one of the two was considerably higher for studies funded by the National Institutes of Health than for drug industry-funded studies but there was no evidence that the tendency to publish important results differed.

d. Multiple publication

Huston and Moher (5550) report that trials can be published in different forms, with different authors, which can make it seem that there is more evidence favouring a treatment than there actually is. They recount trying to untangle the genealogy of a risperidone trial funded by its manufacturer. Similarly, Johansen and Gotzsche (21910) describe the difficulties they encountered trying to carry out a meta-analysis on trials of anti-fungal agents. It was often unclear whether data in multi-centre trials were also published separately, and when contacted, many authors did not respond, or said they no longer had access to the data because it was with the manufacturing company or their previous employer. Tramer et al. (27870) reviewed 84 trials investigating the use of ondansetron for postoperative emesis to quantify the impact of duplicate data on estimates of efficacy. They found that 17% of published full reports and 28% of patient data were duplicated, concluding that trials reporting greater treatment effect were significantly more likely to be duplicated and that inclusion of duplicated data in meta-analysis led to a 23% overestimation of ondansetron's antiemetic efficacy.
CONCLUSION: The evidence that trials sponsored by a drug company are less likely to be published is contradictory. Some major company-funded trials have been published in multiple papers that make them appear to be separate studies, and this can distort the findings of systematic reviews or meta-analyses.


e. Is there an association between industry funding and published results?

Stelfox et al. (1390) examined links between financial relationships with pharmaceutical manufacturers and doctors’ published positions on calcium channel antagonists. They found seventy articles about calcium channel antagonists, and classified them, and their authors as either ‘supportive’, ‘neutral’ or ‘critical’ about the use of these medicines. They then contacted the authors and asked them about their financial relationships with manufacturers of calcium channel antagonists and/or competing products. They found that authors who supported the use of calcium channel antagonists were more likely than others to have financial relationships with manufacturers of these products. Unexpectedly they found that authors who criticised the used of calcium channel antagonists were less likely than other authors to have financial relationships with manufacturers of competing products. The timing of the authors’ position on calcium channel antagonists and their financial relationships were not explored. It is possible that authors supportive of calcium channel antagonists were sought out by companies, rather than company sponsorship leading to more positive positions on calcium channel antagonists. The authors note that only 2 of the 70 articles disclosed potential conflicts of interest.

In their meta-analysis of third generation oral contraceptives and the risk of venous thromboembolism, Kemmeren et al. (21760) found odds ratios of 1.3 (1.0-1.7) in industry funded studies, and 2.3 (1.7-3.2) in other studies. On the same topic, in a letter to the editor of the BMJ, Vandenbroucke et al. (21670) report that of nine unsponsored studies eight found relative risks of 1.5 to 4.0, while four sponsored studies found relative risks of 0.8 to 1.5. Similarly, Mandelkern wrote to the Journal of Clinical Psychiatry (21700) that in 1997 all sixteen industry supported studies in the journal were favourable to the manufacturer’s drug, while all six unsupported studies were not favourable to the study drug.

Wahlbeck and Adams, in a letter to the editor of the BMJ, outline their findings that industry-funded trials on clozapine reported more positive results than non-industry sponsored trials (14870). Similarly, Cho and Bero found that articles that acknowledged support from the industry were more likely to present results which favoured the drug of interest (2820). Rochon et al. (1150) analysed 56 published trials of NSAIDs that they defined as ‘manufacturer associated’. This included studies where the manufacturer had only provided study drugs. They found that the manufacturer’s drug was always reported as comparable to (71%) or superior to (29%) the comparison drug. This was usually justified by the results. In 22 trials one drug was claimed to be less toxic, and in 19 of these this was the manufacturer-associated drug. This claim was justified by a test of statistical significance only 54% of the time. Liebeskind et al. (30), in a poster presentation, outlined a study of controlled clinical trials in acute ischaemic stroke from 1957 to 1997. They suggest there is underreporting of trials showing adverse effects of experimental agents, and that the time from the start of enrolment to publication is longer for trials with negative outcomes than positive outcomes, and that this difference is greater for trials with corporate sponsorship.

Azimi and Welch (21790) looked at cost-effectiveness analyses in journals most visible to clinicians and found that funding source significantly affected the authors’ conclusions about whether therapies requiring additional expenditure were justified, regardless of the quantitative conclusions of the study. Nine out of ten articles which acknowledged industry funding supported additional expenditure, while 15 of the 34 with no industry funding did. Authors of articles supported by the industry supported the use of new technologies at higher costs than other authors.

Friedberg et al. (21810) looked at studies on the cost or cost-effectiveness of new oncology drugs. They found that industry sponsored studies were less likely to report unfavourable qualitative conclusions than studies funded by non-profit organisations. Eighty-nine percent of the studies they looked at used a retrospective design, which Friedberg et al. say allows the sponsor to look at the results of clinical trials and fund economic studies based on those most likely to give favourable economic results. Knox et al. (21710) used the same trials as Friedberg et al., and concluded that industry-funded studies provided less information about the generalisability of their findings. They tended to highlight specific settings where the drug was most likely to be cost-effective.

Davidson (4520) analysed all trials with concurrent or cross-over control groups published in 1984 in five major medical journals. He classified them according to whether they favoured a new therapy or intervention, or favoured traditional management. He also recorded whether the authors acknowledged industry funding or not. Provision of the study drugs or placebos was not counted as industry support. Davidson found a statistically significant association between industry support and whether studies favoured new therapies. Only four industry supported studies favoured traditional therapies. In two of these the manufacturer who supported the study did not make either drug, and in one they made both drugs. Davidson speculates on the mechanism for the relationship between funding and published results. He suggests that industry funding may allow researchers to include large sample sizes which increases their ability to detect statistically significant differences and therefore to publish in a major journal. He also suggests companies may select drugs for study that have already been shown to be efficacious, that they may discontinue studies if the results are appearing to be negative and that they may pressure investigators not to submit negative results.

Jadad et al. (22060) looked at meta-analyses and systematic reviews of treatments for asthma. The six industry-funded reviews were of low quality and the conclusions of all but one favoured the intervention associated with the sponsoring company. The one exception examined the effect of Vitamin C, which was not a new proprietary compound.

Djulbegovic et al. (21940) (21750) argue that randomised controlled trials should only be done if there is substantial uncertainty about which treatment is best. Therefore over time, roughly half of trials should favour standard treatments, and half should favour experimental treatments. However, when looking at 126 published randomised trials on one disease, those supported by commercial organisations mostly (74%) supported new treatments over standard treatments. This was not due to low quality of commercially-supported trials. Kjaergard et al. (21720) also found industry-funded trials to be higher quality than some others (those not receiving external funding).

Freemantle et al. (21740) found that sponsored trials were likely to show greater efficacy of the sponsors’ drugs, but this result was not statistically significant.

CONCLUSION: Drug company funded research is more likely to show results favourable to the product being studied than research funded from other sources. There is an association between the opinions of investigators about products and their source of funding but causality has not been established.

3.10 Does funding affect the research agenda?

Tallon, Chard and Dieppe (21950), report that of 930 controlled trials of treatment for osteoarthritis of the knee, 59% were drug trials, and 26% trials of surgery. They suggest that many of these address questions of little relevance to current management of the disease. They used focus groups and a postal survey to investigate the views and priorities of research consumers. They found that people use a range of treatments and want research in all of these. Dieppe et al. (14890) present a subset of these data. In Chard, Tallon and Dieppe (21780) they show that research on oral drugs produced positive results more often than research on other interventions, and that commercially funded studies were more likely to show positive results than non commercially-funded studies.

CONCLUSION: Drug company funding of research influences the topics studied.

3.11 Do authors reveal funding sources?

Wilkes and Kravitz (21900) surveyed 221 North American medical journal editors and found that only 26% required authors to reveal their funding sources. Sacristan et al. (21690) report that in a large number of pharmacoeconomic studies funding sources are not specified. Moynihan et al. (21930) found that ties between researchers and industry were often omitted from media reports about drugs.

CONCLUSION: Funding from drug companies is often not disclosed.

Directions for future research

There are major gaps and weaknesses in the evidence. An important gap is the lack of evidence about public health outcomes of behaviour changes: does promotion lead to appropriate levels of utilisation of medicines? The evidence that shows conclusively that doctors who rely more on promotion are poorer prescribers suggests that it does not. However because these results could be due to other underlying doctor characteristics, this argument is somewhat weak. More work is needed to establish causal relationships between promotion and prescribing of drugs which have little or no place in rational prescribing, or which have serious adverse consequences when over-prescribed, such as antibiotics.

Other gaps include the lack of evidence from developing countries. All of the studies presented in this review are from developed countries. It is very difficult to untangle the effect of promotion from other inadequacies in systems of medicines distribution in developing countries. In addition there is less funding available for sophisticated studies. However the Cleary study (850) shows how a small, low budget project can provide quite convincing evidence.

Weaknesses include lack of clarity about what studies can and cannot prove. Some researchers do not seem appropriately sceptical about self-report data, and many infer causality from data which simply show associations. There is also some laxity in use of concepts, and in describing previous research. For example, in their survey Lurie et al. (1100) asked about ‘changes in practice’ but in their discussion they discuss ‘changes in prescribing habits’. These are not equivalent. A change in practice could be one small change and may not be related to prescribing, while a change in habit suggests an ongoing and substantial change. Inaccurate descriptions of previous studies are sometimes found in literature reviews at the beginning of articles, particularly inaccurate claims about the conclusions that can be drawn from these studies. Future research needs greater methodological rigour in order to yield more definitive answers to the questions being posed.

 

See Index | Date Entered : Monday, June 02, 2003