Why did GPs Prescribe Rofecoxib? A Qualitative Study of Risk Perception in the Uptake of a New Drug

ABSTRACT

The first Cox-2 selective non-steroidal anti-inflammatory drug (NSAID) in the UK, rofecoxib (Vioxx), was launched in August 1999. However, in September 2004 it was withdrawn from the international market because of concerns about its cardiovascular safety. The objectives of this study were to explore GPs’ perceived risk of a new, innovatory drug (rofecoxib) and how this shaped decisions about prescribing and the processes of new drug adoption. Semi-structured interviews were undertaken with 107 GPs within 6 months of the launch of rofecoxib. Most GPs (63%) prescribed rofecoxib rapidly after its launch. Reasons for prescribing rofecoxib included: a perceived therapeutic advantage or gap in the market particularly in regard to safety, a high level of pre-launch awareness perhaps due to intense direct marketing, hospital prescribing and GPs’ attitudes to risk. There was a general optimism about its value, derived largely from commercial information sources or colleagues. Some GPs were concerned about the long-term safety of rofecoxib but were reassured by their general familiarity with NSAIDs. Specifically, the findings highlight the role of social and contextual factors in GPs’ perception and understanding of risk, and the various strategies they used to manage risk and uncertainty. Thus, the prescribing of rofecoxib can be situated and understood within a socio-cultural theoretical framework that reflects differing beliefs, values and experience in individuals’ constructions of risk.  

KEY WORDS

new drug diffusion / rofecoxib/ prescribing / marketing / risk perception

Introduction

Rofecoxib (Vioxx), was a new drug launched in the UK in August 1999. It belonged to a class of nonsteroidal anti-inflammatory medications (NSAIDs) called COX-2 inhibitors and was the first Cox-2 selective to be licensed. Conventional NSAIDs are associated with adverse effects such as an increase in the risk of gastrointestinal perforations, ulcers, and bleeds. The alleged advantages of COX-2 inhibitors, however, are that they provide the benefits of reducing inflammation but with a reduced risk of stomach ulceration and bleeding.

Like many other new drugs, there was a lack of good clinical data available at the launch of the product to support its effective and safe use in clinical practice. From a commercial perspective however, the manufacturer had to achieve rapid market penetration before a rival (celecoxib) was launched. There was, therefore, an intense marketing effort by the manufacturer: the advertising budget in the US alone for this drug was estimated to have exceeded the world wide advertising budget of Pepsi Cola and Budweiser beer (National Institute for Health Care Management Research and Educational Foundation, 2000). Marketing was also successful in raising awareness of rofecoxib: UK market research reported that 95% of GPs were aware of the drug and that 65% had prescribed it within 12 months of its launch, an extremely high rate (Pitt, 2002).

The first major clinical evidence of rofecoxib’s superiority to existing NSAIDs in reducing the rate of serious gastrointestinal events did not come until November 2000 (Bombardier, et al., 2000), although early concerns were also raised about its cardiovascular risk profile. These concerns lingered, although the manufacturer vigorously defended the safety of its product (Gibson, 2004). But on September 30th 2004, rofecoxib was withdrawn from the international market because of concerns about its cardiovascular safety (Maxwell, & Webb, 2005). There has been much discussion about whether these concerns should have led to the drug’s withdrawal at an earlier stage, and about how far they extend to other drugs of this class. General practitioners (GPs), as the main prescribers of NSAIDs, were the primary target of much of the promotion.

Risk Approaches and Implications for Medical Decision-Making

The concept of risk has become integral to understandings of modern society (Beck, 1992; Giddens, 1991), and has had an increasing emphasis in explaining health behaviour and decision-making. Lupton (1999) identifies medical care and treatment, including drug therapy, as one of six major categories of risk that predominate the concerns of individuals and institutions in Western societies. There are various approaches to examining and analyzing notions of risk in the social sciences, the most dominant being the techno-scientific perspective, the cultural/symbolic perspective, and the social constructionist perspective (for a detailed discussion see Lupton, 1999).  The techno-scientific perspective views risk assessment as a rational, technical approach as measured or estimated from empirical, scientific data (Adams, 1995). It has widespread appeal in industries assessing the risks of new technologies and within the health field (Gabe, 1995).  

Indeed, it follows that the techno-scientific approach may appear logical within the paradigm of evidence-based medicine (EBM) through its implication that the benefits and harms of a treatment can be identified and quantified through appraisal of clinical research findings and scientific measurement and calculation. The EBM approach emphasizes the rational aspects of decision-making and assumes a logical linear progression of information acquisition and appraisal of all relevant drug attributes and possible courses of actions and outcomes. Risks, in keeping with this approach, would be managed according to this knowledge. However, new drug prescribing is an area of high uncertainty. When deciding to use a new drug, a doctor has to weigh up the balance between prescribing a possibly more effective treatment against the potential for unknown, possibly serious side effects. Comprehensive and precise data on the efficacy, safety and long-term effects of new drugs is difficult to obtain, leading some commentators to argue that evidence may not be sufficient to support a new drug’s effective and safe use in clinical practice (Dent & Hawke, 1997; Clarke, et al., 1998; Gale, 2001). For instance, randomized controlled trials (RCTs) of a new drug treatment exclude high risk patient groups, thus limiting the potential for discovering adverse drug reactions in a larger, more heterogeneous population (Ferner, 1996; Rawlins & Jeffries, 1991; Wu & Makuch, 2006). This raises interesting questions about how GPs interpret and manage risk when levels of knowledge may be indeterminate.

Despite these limitations in new drug data, it is not evident that even with accurate, comprehensive scientific information there would be agreement between doctors about the quantification and probabilities of risk. This can occur because of the varying ways in which doctors’ perceive risk and respond to uncertainty (Bloor, 1976). Moreover, sociologists have repeatedly drawn attention to the way in which risk is not objective and measurable, but contingent, and better understood as a social construction (e.g. Adams, 1995; Douglas, 1986; Douglas & Wildavsky, 1982; Gabe, 1995; Lupton, 1999; Nelkin, 1989). This approach draws attention to the involvement of complex social and cultural processes in how individuals perceive and mediate risk. Central to this is the recognition of plural rationalities reflecting individual perceptions and meanings in understanding risk in the context of everyday lives.  

Shortly after rofecoxib was launched, a qualitative study of factors influencing GPs’ initial prescribing of a range of new drugs, including rofecoxib, was undertaken (Prosser, et al., 2003). Re-examining the interview data following rofecoxib’s withdrawal from the market, the present investigation sought to gain a deeper understanding of how GPs assessed the risk of this new, innovatory drug, their approach to negotiating and managing risk, and how this shaped their decisions about prescribing. Mindful of Gabe’s (1995) proposition of the need for a sociological approach to medicine and risk, this paper engages with an interpretive method, structured theoretically within an analytic socio-cultural framework of risk. The sociological implications of the study findings are considered in relation to examining the relationship between risk, knowledge and evidence.

Methods

Sampling

GPs in two health authorities were selected purposively by rates of prescribing of new ‘black triangle’ medicines (those designated by the licensing authority as requiring special reporting of all adverse events as their risk profile is as yet unclear) from Prescribing Analysis and Cost (PACT) data.  Initial sampling was based on stratifying practices into tertiles according to their level of this prescribing. These tertiles were defined as high, medium and low prescribing practices.  Purposive sampling was then employed to select a range of high, medium and low prescribing practices of the study drugs, the aim being to identify a comprehensive range of influential factors, and to capture a range of experiences and instances of prescribing amongst GPs. Purposive sampling also ensured that individuals and practices with a range of other characteristics were included, such as sex, number of practice partners and geographic location for example, urban and rural settings.  The practices targeted for interview were selected by firstly dividing the frequency distributions of the total of indexed new drugs prescribed into tertiles.  Sampling was then selected from the central portion and tails of each distribution.  Thus, both the average prescribing practices of new drugs as well as the outliers (ie. the high and low prescribing practices) were included. 

The Critical Incident Technique

GPs were shown a list of black triangle drugs including rofecoxib and asked which if any they had prescribed. This study is based on reported rofecoxib prescribing incidents. The critical incident technique (Flanagan, 1954) was then used to explore GPs’ reasons for prescribing. This is an open-ended retrospective method that facilitates the investigation of significant occurrences, e.g. events, incidents, processes, issues. It is a way of using the individual experience to identify the factors that are recognised as important in defining what led to a particular occurrence. Thus, a critical incident is one that makes a significant contribution, either positively or negatively, to an activity or phenomenon.  Generally, data are collected via a semi-structured interview, which is more flexible than a questionnaire or survey. Through allowing the interviewee flexibility to describe an event and talk about their experiences and views in their own words, the objective is to gain understanding of the incident from the perspective of the individual, taking into account cognitive, affective and behavioural elements (Chell, 1998). The critical incident technique therefore provides a rich and detailed set of data by allowing respondents to determine which factors are most relevant to them for the event being investigated. It was chosen for use in the present context because it provides an opportunity to obtain an in-depth account of actual prescribing events in everyday contexts from those in the best position to make the necessary observations and evaluations. At the same time, it reflects the natural way doctors think without imposing any a priori determination of what will be important.  Thus, the researcher seeks to understand and construct decision-making from the viewpoint of the individual decision-maker.

Interviews

Using a semi-structured interview, GPs were asked to recall the critical factors that had led to the initial prescribing of rofecoxib. The interviews were conducted by cueing and prompting the participant towards a detailed explanation of the events and the decision-making process, through first awareness of a new drug, information sources, factors influencing assessment, the critical and contextual factors leading to its initial prescription and the reasons for prescribing the new drug rather than an alternative.  Respondents were encouraged to reflect in detail on their notions of the authority and legitimacy of new drug information and knowledge.

Interviews were conducted between August 1999 and February 2000, thus up to 7 months after the launch of rofecoxib. Interviews were tape-recorded and transcribed. The experiences and accounts of those studied served as the basis for data analysis, the aim being to work inductively from the data. As outlined above, data examination began tentatively during the fieldwork stage. However, a more detailed and systematic line-by-line analysis of the interview transcripts began once data collection was completed. This process meant that analysis remained grounded in the data and challenged any a priori assumptions in relation to GPs’ decision-making and what constituted knowledge, for example a rational, scientific process and ‘evidence-based’. This was preceded with frequent readings of the transcripts to increase familiarity. Analysis proceeded with two stages: content analysis and thematic analysis. Content analysis produces a relatively systematic and comprehensive summary or overview of the data as a whole, whereas thematic analysis is more distinctive, typically addressing the issue of ‘what is going on and why’ in more analytic depth and detail.  In this study, content analysis was used to address the question of what factors influence new drug uptake and how often these different factors were mentioned. Thematic analysis is used to explore how andeven why new drug uptake occurs within particular situational prescribing contexts. 

Analysis

The initial process followed the methods of the critical incident technique (Flanagan, 1954). Thus, in order to manage the relatively large set of data, the process of content analysis identified the critical influential factors that emerged for each prescribing incident. These were listed and similar reasons coded into initial conceptual categories (e.g. pharmaceutical representative influence). The entire data set was categorized in relation to these concepts and grouped together by means of a coding system. In this way, content analysis provided a useful summary measure of the extent to which influential factors were distributed. At the same time, however, the technique of content analysis overlooks the importance of the process of negotiating meanings and makes little allowance for the contextual aspects of a situation and individual styles of interpretation. It is useful therefore, to integrate content analysis alongside other qualitative data analysis techniques that are able to uncover the dynamics of social processes and allow closer specification of the significance individuals attach to critical influences. 

Secondly therefore, thematic analysis proceeded inductively following a grounded analytical approach (Strauss & Corbin, 1998) to further define the nature of influence and illuminate the underlying processes of decision-making. The data were repeatedly examined until all cited influences were coded in terms of these categories. Conceptual categories were further specified according to their characteristics (e.g. consultant influenced by observation; consultant influence by socialization, etc). The next step was to compare these categories and identify common analytical themes. Categories were then related according to context and interaction. This process effectively reduced the data into core categories specifying the nature of evidential sources and their relationships to various patient and treatment contexts. Three core sub-categories were developed from the interviews: readiness to prescribe; sources of information; managing uncertainty and risk.

Both the initial conceptual categories and core categories were constantly compared with each other and checked against the interpretation in order to revise and refine explanation. The eventual outcome was to provide a theoretical analysis based on the categories and themes arising in the data. Analytic notes and the initial creation of codes and categories were constituted from the interview data early on in the research when analysis and data collection were conducted simultaneously. The themes and questions that emerged in early interviews helped shape subsequent data collection. For instance, data were analysed for the way in which GPs referred to risk in individual prescribing incidents and recounted their understanding of risk in relation to new drug prescribing. The data were organised and conceptually categorised with the aid of the computer software NVIVO (Richards, 1999).

Results

107 GPs (76 male and 31 female) from 54 practices were interviewed, a participation rate of 73% of GPs and 77% of practices contacted. Of the 721 episodes of prescribing of black triangle drugs, rofecoxib was the second most commonly prescribed, after sildenafil (Viagra). Sixty-seven GPs (63% of total) had prescribed rofecoxib. Prescribing incidents based on others’ decisions (i.e. continuation of a GP or hospital colleague initiated prescription) were cited by a further 13 GPs but are not further analysed here. Specific reasons for prescribing rofecoxib are listed in table 1. The findings can be categorised into three broad themes relating to factors influencing the uptake of rofecoxib: Readiness to Prescribe; Sources of Information and Risk and Uncertainty.

Readiness to prescribe

The most common reason for GPs to prescribe rofecoxib was the perception that it was an advance over current NSAIDs, and filled a therapeutic gap:

The thing that’s swung it is it's got significant advantages over its competitors.. (GP4)

There’s a very definite niche, it fills a definite hole. (GP7)

This was largely based on the drug's alleged improved adverse effect profile over current alternatives. GPs saw its value in patients who were currently treated sub-optimally:

There’s a lot of problems with people who can’t tolerate any NSAID because of gastric irritation problems and Vioxx has been launched on the premise that it doesn’t cause these problems… it does actually seem to be a far better bet than its competitors…. it’s a good choice. (GP98)

We’ve got a group of people for whom we’d love an anti-inflammatory that doesn’t shred your
stomach… It’s a very limited group of people generally who have been waiting for it for a long time. (GP73)

However, this may reflect good anticipatory marketing, creating a demand for the drug before its launch:

A patient comes you think, ah, that’s the drug I was thinking about, yes, this person might be appropriate to try it on. (GP27)

I initiated it because there’s a lot of need...again we knew this was coming. I don’t know where I heard about them. I read my magazines but you just see Vioxx on big pages, so I suppose it does come from the advert. (GP24)

Nevertheless, according to the stated approaches of these GPs, prescribing is accepted in circumstances where clearly identifiable benefits and levels of expected utility are perceived to outweigh risks. However, as will become apparent in the following discussion, risk assessment was not readily expressed as an objective, measurable process within the techno-scientific approach that underpins evidence-based medicine.

Sources of Information

GPs showed scant systematic or comprehensive search for scientific research evidence. Information was acquired opportunistically from commercial sources or the observation of consultant prescribing and judgement.In only 16 instances did GPs report having gained information from non-industry literature, with two claiming to have read about it in the Drugs and Therapeutics Bulletin and the rest in articles in non-peer reviewed journals (i.e. those free to GPs and financed by advertising, such as Pulse).

Many GPs spoke of an initial creeping or background awareness of rofecoxib, largely through advertising and the GP press.

it may have been articles in things like GP or Pulse, those kind of ones. I couldn’t say for definite, it may even be adverts in the journals as you’re flicking through the pages. (GP90)

There was a lot of promotion on this one, it’s everywhere in the magazines. (GP64)

Four GPs acknowledged that they had prescribed rofecoxib solely on the basis of promotional literature:

It was just the adverts in the rags that made me look at it. I haven’t heard about it from anywhere else. (GP7)

For 26 prescribers, the company representative was the sole information source:

I’ve probably had about 4 or 5 patients that I have put on that. I met with the representative and I basically was quite impressed with the studies and the evidence that was put forward. (GP65)

I only actually saw the rep last week and I’ve used it twice since then (GP49)

… and often seen as personally highly credible:

The other reason for prescribing [rofecoxib] is the drug rep. He’s probably one of the best that I know as a salesman. He’s very persistent and very sociable and so he can sell anybody anything, but I do take him with a pinch of salt, unless it’s a good drug. (GP64)

Vioxx I’ve used that quite a bit. We’ve got a rep who I’ve known for fifteen years and I trust him. He’s a proper pharmacist, he’s not just any old rep and I value his judgement, so it means if he tells me it’s a good drug I would go along with that. (GP82)

The indirect influence of a hospital consultant was important in 30 incidents, in two ways: either by experiential knowledge, based on observation of the effects of rofecoxib prescribed at a consultant’s request; or simply, observed consultant prescribing. GPs considered consultant use as validation of the drug’s value and so provided an authoritative standard for GPs’ own practice:

One of the local rheumatologists has been using it quite a bit when I had barely heard of it, so that more or less tipped the balance in me deciding to give it a try. (GP26)

I didn’t actually prescribe [rofecoxib] until some patients had come out from hospital on it. I was just a bit skeptical as to how good it was going to be, so seeing the effect of the drug in the patient, rather than just the fact that the hospital were prescribing it. (GP96)

Furthermore, rofecoxib presented an alternative before referral, when GPs anticipated the action of specific hospital doctors:

My influence really has been our local consultant rheumatologists who’ve gone potty on the stuff. I’m not a great one for trying things new, just because they’re new, but now this has taken off it does seem to be quite useful…You’ve seen how well Vioxx suits the patient, the next patient who comes in with a similar sort of problem, you’re thinking, ‘I don’t have to refer this patient to rheumatology, I’m going to try this one on Vioxx. (GP74)

Patient requests for rofecoxib were influential in only a few incidents, in some cases inducing GPs’ awareness of it:

I actually first heard about this when a patient brought a newspaper cutting in. (GP47)

Risk and uncertainty

GPs’ perception of risk plays a fundamental role in their decision-making. GPs were aware of the uncertainty around the early use of a new drug like rofecoxib, and many were ambivalent about its safety and outcomes due to a lack of long-term, scientific research evidence:

I'm still a bit wary about giving something like Vioxx until it's got a bit more data behind it.

However, while gaps in GPs’ knowledge can create uncertainty, the extent to which this is understood as ‘risk’ is not equal for all prescribing incidents and for all GPs, nor is risk perception restricted to the accumulation of scientific research information and an objective assessment of calculable probabilities:

This is quite a good idea in principle I think. I’m not sure that there are enough studies to make me happy with it. I’m happy to initiate it. The worry is that it hasn’t been around that long yet, so again I don’t know what it does to people in the long term. (GP77)

Indeed, the perception of risk, and the decision to prescribe rofecoxib, incorporates a high element of subjectivity and is open to individual interpretation. Risk and its negotiation are constructed around not only the nature and breadth of GPs’ underpinning knowledge relating to rofecoxib, but also around individual risk preference and the perceived level of uncertainty and risk a GP is willing to accept in a given situation. As such, GPs arrive at their own notion of risk drawn from a set of influences that include personal beliefs, past clinical experience, social and cultural factors, social and professional relationships, concepts of trust and credibility and clinical contexts. What might be considered risky in one clinical situation or patient context may not be considered so in another. Furthermore, despite the indeterminate nature of new drug knowledge, risk and uncertainty are not a constant feature of prescribing.  On the contrary, there were many prescribing incidents in which risk was not an issue, or in which risk was controlled and de-sensitised through various rationalities and practical reasoning that simplified decision-making. While there is considerable variation in GPs’ understanding and response to risk, the data also point to certain patterns in the organization of how GPs conceive, negotiate and control risk.

Most obviously, most GPs felt that the benefits of rofecoxib outweighed the risks in particular patients:

I’m a bit sceptical about new non-steroidals. I know there’s a difference in the Cox 1, Cox 2 inhibitors but, you know, in terms of adverse effects… I had somebody with rheumatoid, who was really quite bad and in a lot of discomfort with it, despite being on methotrexate and opioids and everything, and couldn’t tolerate Arthrotec, so I thought I’ll give it a try. (GP68)

… and, more importantly, that the risk of adverse effects was less than with other NSAIDs:

I don’t often prescribe new drugs but, I mean, really something like Vioxx, one has such problems with gastric problems that, you know, when somebody [the rep] says,‘this will not give gastric problems,’ you really do have to listen. (GP21)

For the most part, GPs considered an alternative treatment for individual patients based on their knowledge of the patient’s previous medication history, or for those patients whom they considered at high risk of developing serious gastrointestinal (GI) adverse events. Decision-making thus involved making complex evaluative judgements, weighing the potential risks and benefits of prescribing a new drug with alternative courses of action in individual patients. These instances emphasise the situated rationality of GPs’ decision-making. Against this background, rofecoxib was seen as a rational alternative for many patients:

There’s a lot of problems with people who can’t tolerate any NSAID because of gastric irritation problems and Vioxx has been launched on the premise that it doesn’t cause these problems… it does actually seem to be a far better bet than its competitors…. it’s a good choice. (GP98)

It’s particularly very elderly patients who have previously had long-standing arthritis that has not been treated adequately…it’s something to try and help them have a quality of life. (GP29)

Thus, GPs tend to prescribe rofecoxib for those patients perceived to be at highest risk of not being managed effectively with other treatment. GPs’ accounts conveyed the increase in risk acceptability of rofecoxib with an increase in their perception of its potential benefit. Here, risk-taking was viewed optimistically and rofecoxib distinguished as a means of challenging limited therapeutic options and providing opportunity for health benefit, or avoiding the risks associated with other medicines.

Furthermore, it seems that risk may be more tolerated when choice is constrained. In many prescribing incidents, judgements about risk therefore reflected not only characteristics of the risk itself, but also its contingency in relation to other potential risks if the drug was withheld. In a number of circumstances, the rationalisation of prescribing rested on the notion that doing something was better than doing nothing. While an orientation to matching the patient to a particular drug underpinned these decisions, a new drug was prescribed as a default action when preferred alternatives had been exhausted. In other words, GPs chose to take a chance rather than to accept certain losses. At this early stage in its market life, rofecoxib was restricted to particular cases, rather than used as a first choice NSAID:

It’s not quite the last resort but it’s not a routine, there’s got to be a good indication so you tend to have worked your away through the options already. (GP23)

Other strategies to reduce risk were to limit duration of prescription, and paradoxically in the light of the main reason to use it to reduce risk, restrict use to relatively healthy patients:

I wouldn’t prescribe a four-month course, but I'd perhaps give it to them for a week to see how they got on. (GP71)

Vioxx, I didn’t have much problem there because most of the patients were otherwise healthy. It’s when you’re dealing with a sick patient and it’s a totally new class of drug, then I would tend to think, wait until I’ve seen it being used or recommended by hospital colleagues. (GP87)

I tend to be a little cautious, but I suppose it depends what it’s for. If it’s something like Vioxx where it’s a relatively sort of benign condition OK, but I’d be much more cautious with new blood pressure tablets, or diabetic tablets or something like that. (GP91)

Previous experience of prescribing NSAIDs without serious adverse consequences…

Another anti-inflammatory maybe it’s a slightly better side effect profile, but it’s just another
anti-inflammatory, we’re confident with using them, give it a go. (GP36)

In contrast, however, in a very small number of incidents, there appeared to be a lack of congruence between the situational context, as described above, and GPs’ reasons for prescribing.  This might be explained by the notion that individual interest and attitudes towards rofecoxib affect risk perception, the interpretation of evidence and the timing of new drug initiation. It would appear that thresholds for the management of uncertainty differ between individual GPs. For instance, some GPs preferred to proceed cautiously, adopting a 'play safe, wait and see' policy, while others were more willing to apply risks and prescribe in the absence of solid or legitimate evidence, in order to offer innovation to patients. In short, risk-taking was essentially accepted and viewed as an indivisible part of driving medical progress and providing opportunity for improved therapeutic benefit:

It’s risky but you have to have risk built into the system to make progress, you can’t stop every conceivable risk. (GP31)

Some of them you just think ‘right well, I’ll give them a go!’  I’ll initiate those in my practice to see how they get on.  Vioxx because it’s the only drug in it’s class (GP4)

Acknowledging uncertainty, GPs saw their initial prescribing of rofecoxib as experimental, a process in which the benefits and risks were tested through personal experience:

The first time I used it was in the mother of a local GP who came to me and said my daughter said I should I try this. I said, oh great, I was looking for a guinea pig to try it on… (GP43)

Crucially, it was the outcomes of interventions such as this that served to establish GPs’ notions of efficacy and which encouraged or discouraged further use:

I’ve not been terribly impressed with the results that I’ve had so far. Obviously if something is effective in one patient, I’m more likely to try it in another. If I have a lukewarm response or they have a reaction to it, then I’m reluctant to start somebody else on it. (GP41)

Discussion

This study illustrates the factors that affect the uptake of an innovative new drug: a perceived therapeutic advantage or gap in the market particularly in regard to safety, a high level of pre-launch awareness perhaps due to intense direct marketing, hospital prescribing and GPs’ attitudes to risk. At the same time, GPs did not articulate risks in quantitative, measurable terms, and their accounts suggest that a techno-scientific approach does not satisfactorily account for the ways in which GPs conceptualize and manage prescribing risks. Rather, the prescribing of rofecoxib can be situated and understood within a socio-cultural theoretical framework that reflects differing beliefs, values and experience in individuals’ constructions of risk. The findings highlight the role of specific social and clinical contexts in which risk is interpreted and characterized, and the various strategies GPs use to manage risk and uncertainty. This is contrary to the assumptions reflected within the evidence-based medicine paradigm, which purports that clinical practice should be based on rigorous scientific enquiry and evaluation. The key dimensions underlying GPs’ risk perceptions and their relationship to knowledge are considered below.

Prescribing was invariably subject to the specific circumstances of individual patients and their experiences with previous treatment. Initial prescribing of rofecoxib was clearly not based on extensive published clinical evidence but on “a strategy of desire”(Scott & Ferner, 1994), and optimism about clinical benefits and lack of harm as yet unproven (Dowden, 2003). More importantly ‘tacit’ knowledge derived from socially mediated forms of evidence and personal experience reduced uncertainty and formed the conceptual and practical base for much prescribing. The GPs’ accounts here, therefore, reaffirm the distinction between the formal rationality of science and what has been termed ‘the art of medicine’ based on clinical judgement and methods of practical reasoning. This is characteristic of much clinical practice (Armstrong, 2002; Gabbay & leMay, 2004; Greer, 1988).

GPs’ perspectives represent the competing elements to risk-taking.  In effect, approaches to risk are frequently contested, with many prescribing incidents illustrative of the tension between constraint and opportunity, with decision-making dependent upon interpretation of the risk within the context to which it is applied. To a large extent, the interpretation of risk is inseparable from, and framed around, the everyday reality of specific clinical and patient contexts. In this sense, orientations to risk can be understood as rational and contingent. Concern over a deteriorating clinical situation, the severity of illness symptoms and the lack of, or failure of, treatment alternatives were seen to expose patients to specific risks and to influence the ways in which GPs viewed prescribing rofecoxib as being a risk or not. Thus, while the prescribing of a drug may still be perceived to contain some element of risk, uncertainty could be de-sensitised when there was pressure to make people well or when specific circumstances restrained choice. GPs’ accounts of weighing up options, and the rationalities and practical reasoning engaged with in decision-making, provide evidence of the clinical reflexivity shown in responses to risk.

What is more, the research data shows that risk is a contained feature of new drug prescribing in that there were many prescribing contexts in which risk was not a cause for doubt for GPs. This was observed in relation to GPs’ own risk-taking preferences or because, to some extent at least, risk is regulated and assessed through accumulated clinical experience and professional knowledge sources. In relation to the latter, an important dimension of risk relates to subjectively based perceptions of trust and credibility embedded in social interaction and professional relationships. In negotiating uncertainty and risk, GPs demonstrated their investment of trust in sources of information they felt were reliable and credible, and in whose judgement they felt was safe.

Likewise, information itself does not have a straightforward, rational impact on risk perception, but is evaluated through filters of trust. In this vein, the plausibility of others’ prescribing was a specific target for trust, a common source being the hospital consultant. The essence of this is that evaluation of risk is a product of the interpretation of the information source and depends not simply on what is being communicated, but who communicates it.  GPs frequently constructed risk perception on their knowledge and empirical observations of hospital prescribing and the clinical behaviour of others around them. These were frequently taken as expressions of assurance that GPs relied upon.  This is reminiscent of Luhmann’s (1979) description of trust as the “blending of knowledge and ignorance.” In ascribing this perspective to clinical practice, trust is a way of negotiating uncertainty and complexity in the absence of scientific assurances. It is such findings that draw attention to the ambiguous nature of both knowledge and risk. This is also seen in the way that the relative benefits of rofecoxib were frequently interpreted on the basis of individual beliefs and attitudes and various implicit norms and systems of judgement, such as perceived information credibility; situated observation and clinical contexts; and embedded knowledge developed and internalized through prior prescribing of similar drugs and direct personal ‘trialling’.

Evidence and risk are thus interpreted along a number of dimensions, which are predominantly informal and social, rather than scientific or technical, in character. This should perhaps not surprise us, and it may be argued that within the context of new drug prescribing, employing scientific rationality as a basis for certainty counteracts the possibility of improved therapeutic benefit. Since measurable objective risks and the safety of a new drug cannot be entirely specified and guaranteed at its launch, the need for reflective practice and clinical interpretation is inevitable if doctors are not to become paralysed by uncertainty. Equally of course, risk evaluations based on contextualised informal knowledge and reflective experience may not lead to "optimal" drug choice because of the tendency for selective interpretation of information.

In addition, these findings substantiate previous work on the diffusion of innovation in clinical practice in that they highlight the concept of a product’s relative advantage, experiential testing and the role of social influence and local practice in decision-making (Coleman, et al., 1957; 1966; Greer, 1988; Fitzgerald, et al., 2002; Rogers, 1995). However, this study also reveals the importance of marketing, generally absent from diffusion models. Marketing not only raised awareness, but also influenced individual decision-making. Two recent North American studies have also shown this in relation to the coxibs: the first from Canada (Klein, et al., in press) showed in addition that availability of free samples to doctors was important to allow them to trial a drug in an individual patient, where patient co-payments were a disincentive. The second (Alexander, et al., in press) documents the peaks of intensive promotional activities initially to gain a place in the market, and later apparently intended to counter growing concerns about the risks of these drugs.

The reasons for GPs’ application of information from representatives have been explored elsewhere (Prosser & Walley, 2003). Some GPs acknowledged their dependence on commercially driven information and gave what could only be a normative response, suggesting that the respected Drugs and Therapeutics Bulletin had been their source of information: in fact this journal published nothing on rofecoxib until after our data collection, in November 2000 (Anon, 2000).  That 75% of GPs in this sample had prescribed rofecoxib within six months of its launch illustrates the speed with which use of a drug can disseminate. However the use of rofecoxib stabilised over the next two years and it achieved 12% of the total market by volume (about 200,000 of the 1.7 million prescriptions in England for all NSAIDs dispensed in September 2004, and about half of all Cox 2 inhibitor prescriptions). Other countries reported more dramatic market penetration by Cox-2 inhibitors – over 50% of the market by number of NSAID prescriptions within one month of approval for reimbursement in Australia (Kerr, et al., 2003), and 40% at peak in the United States (Villaneuva , 2003). This relatively low use in England might be related to a number of factors including: confusion over how real the gastrointestinal benefits of Cox-2 inhibitors were; advice from the National Institute of Clinical Excellence (National Institute for Clinical Excellence, 2001), suggesting that these drugs should be restricted to high risk patients; uncertainty in cardiovascular disease; and the relative high cost of the drug and budgetary pressures on GPs.

A limitation of this study is that we relied on GPs’ subjective recall of prescribing events. Their disclosure of contributory factors may perhaps be prejudiced by normative responses, as seen in some cases. Nevertheless, this is somewhat overcome by the validity of the critical incident technique which uses specific factual prescribing contexts, an interview structure that is probing and interactive, and because the interviews followed closely behind the actual prescribing events. This study is based on interviews conducted some years ago and it is not clear whether views about new drug uptake and its hazards have changed since, in particular since the high profile withdrawal of rofecoxib and new evidence undermining the perceived, but unproven, benefits of other established therapies (Minelli, et al., 2004).  At the same time, there has been a growing awareness of, and unease about, the role of marketing with themed issues of the British Medical Journal and the development of organisations such as No Free Lunch (http://www.nofreelunch-uk.org) and Healthy Skepticism (http://www.healthyskepticism.org/index.htm), dedicated to promoting better understanding of marketing.

The original study was funded under the Prescribing Research Initiative, Department of Health.  This study had no specific funding.

Contributors: HP and TW designed the study. HP carried out interviews and the data analysis.TW contributed to data analysis. Both HP and TW wrote the paper. Karen Clayson transcribed the interviews. TW is guarantor.

Table 1 – Reasons for prescribing rofecoxib in 67 GP initiated prescribing episodes (more than one reason cited in many cases)

Readiness to prescribe - Relative advantage

Better side-effect profile - 35
Efficacy - 8
Existing therapy sub-optimal - 47

Evidential sources

Pharmaceutical industry - 35
Rep - 30
Adverts/mailings - 6
Hospital Doctor – 30

References

Adams, J. (1995) Risk. London: UCL Press

Alexander, G.C., Majumdar, S.R., & Stafford, R.S. Promotion and prescribing of coxibs during accumulating evidence of risk. PLoS (in press)

Anon. (2000). Are rofecoxib and celecoxib safer NSAIDs? Drugs and Therapeutics Bulletin, 38, 81-86

Armstrong, D. (2002) Clinical autonomy, individual and collective: The problem of changing doctors’ behaviour. Social Science and Medicine, 55(10), 1771–1777.

Beck, U. (1992) Risk Society, London: Sage

Bloor, M. (1976) Bishop Berkeley and the adenotonsillectomy enigma: An exploration of variation in the social construction of medical disposals. Sociology, 10: 43-61.

Bombardier, C., Laine, L., Reicin, A., Shapiro, D., Burgos-Vargas, P.H. R., Davis, B., Day, R., Bosi Ferraz, M., Hawkey, C.J., Hochberg, M.C.,  Kvien, T.K., & Schnitzer, T.J. (2000) Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. New England Journal of Medicine, 343, 1520-1528.

Chell, E., (1998) The critical incident technique in Qualitative Methods and Analysis in Organisational Research: A practical Guide. G. Symon & C Cassells (Eds.) Thousand Oaks, CA; Sage.

Clarke, W., Fisher, M., & Misson, J. (1998) Data available before products are launched are poor.  Letter to the British Medical Journal, 316, 228.


Coleman, J., Katz, E., & Menzel, H. (1957) The diffusion of an innovation among physicians. Sociometry, 20(4), 253-270.

Coleman, J., Katz, E., & Menzel, H. (1966) Medical innovation: A diffusion study.  Bobbs-Merrill.

Dent, T.H., & Hawke, S. (1997) Too soon to market. British Medical Journal, 315:1248-9

Douglas, M. (1986) Risk Acceptability According to the Social Sciences, London: Routledge.

Douglas, M., & Wildavsky, A. (1982) Risk and Culture, Berkeley: University of California Press.

Dowden, J. (2003) Coax, COX and cola.  Medical Journal of Australia, 179, 397-8.

Ferner, R.E. (1996) Newly licensed drugs. British Medical Journal, 313, 1157-1158.

Fitzgerald, L., Ferlie, E., Wood, M. & Hawkins, C. (2002) Interlocking interactions, the diffusion of innovations in health care. Human Relations, 55(12); 1429-1449.

Flanagan JC. (1954) The critical incident technique. Psychological Bulletin , 51: 327-358.

Gabbay, J., & le May, A. (2004) Evidence-based guidelines or collectively constructed “mindlines?” Ethnographic study of knowledge management in primary care. British Medical Journal, 329, 1013-1027.

Gabe, J. (1995) Health, medicine and risk: the need for a sociological approach. In J. Gabe (Ed.) Medicine, Health and Risk, Blackwell Publishers Ltd: Oxford.

Gale, E.A. (2001) Lessons from glitazones: a story of drug development. Lancet, 357, 1870-75.

Gibson, L. (2004) Spanish drug editor wins case brought by Merck, Sharp & Dohme, British Medical Journal, 328, 307.

Giddens, A. (1991) The Consequences of Modernity, Cambridge: Polity Press.

Greer AL. (1988) The state of the art versus the state of the science: the diffusion of new medical technologies into practice. International Journal of Technology Assessment in Health Care, 4: 5-26.

Kerr, S., Mant, A., Horn, F., McGeechan, K. & Sayer, G. (2003) Lessons from early large-scale adoption of celecoxib and rofecoxib by Australian general practitioners, Medical Journal of Australia, 179, 403-407.

Klein, D., McDonald, A., Drumond, N., & Cave, A. (In press) A qualitative approach to determine factors influencing coxib prescribing for musculoskeletal disorders in Alberta Canada. Family Practice.

Luhmann, N. (1979) Trust, Chichester: John Wiley.

Lupton, D. (1999) Risk, London: Routeledge.

Maxwell, S.R.J., & Webb, D.J. (2005) COX-2 selective inhibitors—important lessons learned. Lancet, 365, 449-451.

Minelli, C., Abrams, K.R., Sutton, A.J., & Cooper, N.J. (2004) Benefits and harms associated with hormone replacement therapy: clinical decision analysis. British Medical Journal, 328,371.

National Institute for Health Care Management Research and Educational Foundation. (2000) Prescription drugs and mass media advertising. Washington: NIHCM Foundation.

National Institute for Clinical Excellence. (2001) Guidance on the use of cyclo-oxygenase (COX) II selective inhibitors, celecoxib, rofecoxib, meloxicam and etodolac for osteoarthritis and rheumatoid arthritis. London: NICE, 2001.

Nelkin, D. (1989) Communicating technological risk: The social construction of risk perception. Annual Review of Public Health, 10, 95-113.

Pitt, M. (January 2002) MSD: Vioxx/Rofecoxib - achieve a 95% level of awareness
http://www.allaboutmedicalsales.com/pharmareview/january_02.html Accessed 10th Feb 2006.

Prosser, H., Almond, S., & Walley, T. (2003) Influences on GPs’ decision to prescribe new drugs – The importance of ‘who says what’. Family Practice, 20, (1) 61-68.

Prosser, H., & Walley, T. (2003) Understanding why GPs see pharmaceutical representatives: a qualitative interview study. British Journal of General Practice, 53(4): 305-11.

Rawlins, M.D., & Jeffreys, D.B. (1991) Study of United Kingdom product licence applications containing new active substances, 1987-9.  British Medical Journal, 302, 223-5.

Richards L. (1999) Using NVivo in qualitative research. Bundoora, Victoria, Australia: Qualitative Solutions and Research.

Rogers, E.M. (1995) Diffusion of Innovations.  New York: The Free Press.

Scott, D.K., & Ferner, R.E.  (1994) 'The strategy of desire' and rational prescribing. British Journal of Clinical Pharmacology, 37, 217-9.

Strauss, A., & Corbin, J. (1998) Basics of qualitative research: Techniques and procedures for developing grounded theory. London: Sage Publications.

Villaneuva, P., Peiro, S., Librero, J., & Pereiro, I. (2003) Accuracy of pharmaceutical advertisements in medical journals. Lancet, 361, 27-32.

Wu, Y., & Makuch, R. (2006) Detecting rare adverse events in post-marketing studies: sample size considerations. Drug Information Journal, 40, 89-98.