- Email: [email protected]
Prescribing Practices Amid the OxyContin Crisis: Examining the Effect of Print Media Coverage on Opioid Prescribing Among Physicians
The Journal of Pain, Vol 14, No 12 (December), 2013: pp 1686-1693 Available online at www.jpain.org and www.sciencedirect.com
Prescribing Practices Amid the OxyContin Crisis: Examining the Effect of Print Media Coverage on Opioid Prescribing Among Physicians Alexandra Borwein,* George Kephart,* Emma Whelan,y and Mark Asbridge*,y Departments of *Community Health and Epidemiology and ySociology and Social Anthropology, Dalhousie University, Halifax, Nova Scotia, Canada.
Abstract: The pain medication OxyContin (hereafter referred to as oxycodone extended release) has been the subject of sustained, and largely negative, media attention in recent years. We sought to determine whether media coverage of oxycodone extended release in North American newspapers has led to changes in prescribing of the drug in Nova Scotia, Canada. An interrupted time-series design examined the effect of media attention on physicians’ monthly prescribing of opioids. The outcome measures were, for each physician, the monthly proportions of all opioids prescribed and the proportion of strong opioids prescribed that were for oxycodone extended release. The exposure of interest was media attention defined as the number of articles published each month in 27 North American newspapers. Variations in media effects by provider characteristics (specialty, prescribing volume, and region) were assessed. Within-provider changes in the prescribing of oxycodone extended release in Nova Scotia were observed, and they followed changes in media coverage. Oxycodone extended release prescribing rose steadily prior to receiving media attention. Following peak media attention in the United States, the prescribing of oxycodone extended release slowed. Likewise, following peak coverage in Canadian newspapers, the prescribing of oxycodone extended release declined. These patterns were observed across prescriber specialties and by prescriber volume, though the magnitude of change in prescribing varied. Perspective: This study demonstrates that print media reporting of oxycodone extended release in North American newspapers, and its continued portrayal as a social problem, coincided with reductions in the prescribing of oxycodone extended release by physicians in Nova Scotia. ª 2013 by the American Pain Society Key words: Opioids, OxyContin, prescribing practices, media effects.
O
xyContin (Purdue Pharma LP, Stamford, CT) is an extended release form of oxycodone that was approved for use in Canada in 1996. When introduced, oxycodone extended release was considered to be a breakthrough for its ability to provide sustained Received February 26, 2013; Revised August 16, 2013; Accepted August 27, 2013. A.B. was supported by a Canadian Institutes for Health Research master’s award, in addition to a Scotia Scholarship and a Scotia Support Grant from the Nova Scotia Health Research Foundation. The study was supported, in part, by a Health Research Project Grant from the Nova Scotia Health Research Foundation. The authors declare no conflicts of interest. Supplementary data accompanying this article are available online at www.jpain.org/ and www.sciencedirect.com/. Address reprint requests to Mark Asbridge, PhD, Department of Community Health and Epidemiology, Dalhousie University, Centre for Clinical Research, 4th Floor, 5790 University Avenue, Halifax, Nova Scotia, Canada B3H 1V7. E-mail: [email protected] 1526-5900/$36.00 ª 2013 by the American Pain Society http://dx.doi.org/10.1016/j.jpain.2013.08.012
1686
pain relief and because prescribers expected it to be less prone to abuse than other opioids.4 Soon after the introduction of oxycodone extended release, however, it was discovered that when crushed or chewed and inhaled, injected, or swallowed, the oxycodone is released and absorbed rapidly, producing a heroin-like euphoria.7 As a result, within 5 years, the popular press and medical authorities in parts of North America began to report the use of oxycodone extended release as a street drug and a growing social problem.27 This was particularly the case in the Atlantic Coast of Canada and the United States, where many reports surfaced about abuse and addiction, criminal diversion of the drug, as well as oxycodone extended release–related overdoses and deaths.27 In examining these events, a key question is whether the increased, and largely negative, media attention on oxycodone extended release affected prescriber practices. On the one hand, negative media coverage may
Borwein et al propagate fears among prescribers, including concerns about drug diversion, addiction, and crime and the threat of being sanctioned by governing bodies in medicine.3 Alternatively, it has been suggested that users of oxycodone extended release may face stigmatization and reduced access to therapy from wary prescribers.3,27 It is important to understand how media portrayals of oxycodone extended release may have influenced prescribing practices among Canadian physicians. Our objective was to determine whether media reporting of oxycodone extended release in North American newspapers led to changes in the prescribing of the drug in Nova Scotia between 1996 and 2007.
Methods An interrupted time-series design was used to examine the relationship between media attention on physicians’ monthly prescribing of oxycodone extended release. The outcome measures were, for each physician who prescribed opioids, the monthly proportions of all opioids and strong opioids that were for oxycodone extended release. The exposure of interest was media attention defined as the number of articles published per month based on a content analysis of 27 North American newspapers.27 Variations in media effects by provider characteristics (specialty, prescribing volume, and region of practice) were assessed.
Data Source Administrative data on all opioid analgesic prescriptions in Nova Scotia between September 1, 1996 (the first month in which oxycodone extended release was prescribed), and December 31, 2007, were obtained from the Nova Scotia Prescription Monitoring Program, which maintains a database with patient, prescriber, and drugrelated information for all monitored prescription drugs, including opioids, dispensed in Nova Scotia since 1992.8,19 These data do not include prescriptions filled in hospitals or long-term care facilities, or those without a prescriber identification number.
Outcome Variables Two outcome variables were examined. The first outcome variable was the proportion of physicians’ monthly volume of all opioids prescribed that were for oxycodone extended release. The second was the proportion of physicians’ monthly volume of all strong opioids that were for oxycodone extended release. Oxycodone extended release is a strong opioid, and restricting analyses to strong opioids provided a comparison with opioids that are often prescribed for the same or similar conditions. This is particularly relevant in the context of concern over oxycodone extended release, where providers wishing to move away from this drug would be inclined to prescribe another strong opioid. Opioids were classified as strong or weak based on the 2010 Canadian Guideline for the Safe and Effective Use of Opioids for Chronic Non-Cancer Pain.16 Briefly, strong opioids were defined as medications containing sufentanil, fentanyl,
The Journal of Pain
1687
hydromorphone, morphine, and oxycodone. Weak opioids were defined as opioid medications that contained codeine, pethidine (meperidine), or dextropropoxyphene. Outcome variables were highly positively skewed and were therefore logit transformed in all statistical models. To compute prescribing volume across medications with different strengths and normal doses, we employed the Defined Daily Dose (DDD) methodology developed by the World Health Organization.29 The DDD is defined as the ‘‘assumed average maintenance dose per day for a drug used for its main indication in adults.’’29 DDDs for opioids were obtained from the online searchable ATC/ DDD Index for 2012.28 For drugs that contained codeine or propoxyphene in combination with other nonopioid ingredients (eg, acetaminophen), only the opioid portion was considered. Using the strength of the prescription and the total quantity prescribed (the number of tablets, total volume, or number of patches), the total number of DDDs per prescription was calculated and then aggregated by physician and month to compute the outcome variables. The major disadvantage of the DDD methodology is that not all drugs have assigned DDDs. The DDD is determined based on international drug information and may not necessarily reflect prescribing patterns of the associated drug in Canada, as main drug indications and prescribed daily doses may vary between countries. However, in the context of the research presented here, the advantages conferred by allowing comparisons across prescriptions outweighed the disadvantages of the system. In particular, the use of DDDs allowed background opioid prescription rates to be taken into account. We assessed whether the effects of media on prescribing of oxycodone extended release varied by physicians’ overall prescribing volume, specialty, and region of practice. Prescribers were stratified into 4 volume groups, for each of the outcome variables, based on their mean volume of opioids (strong opioids) prescribed across all months. Groups were created using the quartiles of the means for each outcome variable (termed ‘‘Low,’’ ‘‘Low-Medium,’’ ‘‘Medium-High,’’ and ‘‘High’’). Prescriber specialties were grouped into 3 categories (general practitioners, anesthesiologists, and other). Region of practice was defined according to 9 regional District Health Authorities in Nova Scotia.
Explanatory Variable Data on newspaper coverage of oxycodone extended release were drawn from a previous study that examined representations of oxycodone extended release in North American newspapers between 1995 and 2005.27 Searches of newspaper literature databases were conducted and 924 stories published in 27 North American newspapers were examined, with 172 (18.6%) from Canadian newspapers (149 of which were from Nova Scotia; see Supplementary Appendix for a full list). More than 95% of articles identified at least 1 problem related to oxycodone extended release. Analysis of the number of
1688
Media Effects on OxyContin Prescribing
The Journal of Pain
unique stories published each month revealed 2 peak periods of media attention. The first, which began in February 2001 and peaked in July 2001, was largely in U.S. sources. The second, which began in May 2003 and peaked in March 2004, corresponded to attention in Canadian sources. For our statistical models, a media exposure measure was created using linear splines with knots defined according to the onset of the peaks in print media reporting. The linear trends in prescribing are captured as splines with knots created to represent the 3 intervals in the splines: prior to February 2001, February 2001 to April 2003, and May 2003 to December 2007. The use of splines allowed for smoothing of the media data and production of coefficients in the statistical models that represented the change in slope from the preceding interval; that is, the splines allowed for the measurement of the change in prescribing of oxycodone extended release for each time interval compared to the preceding interval.
Possible Confounders A number of confounders were considered in the context of this study, including the possibility of ongoing medical education activities around oxycodone extended release and opioids, the introduction of new products to the market, and the publication of guidelines related to the prescribing of opioids. Investigations into such activities, including consultations with several experts in the fields of pain medicine and medical education, indicated that there were no such activities in the province of Nova Scotia or across Canada that would impact the prescribing of oxycodone extended release in the province during this time.
Statistical Analyses Descriptive analyses examined the time period between September 1996 and December 2007, inclusive. Statistical models to measure effects of media exposure were restricted to the time period between January 2000 and December 2007, because of the small numbers of oxycodone extended release prescriptions prior to 2000. Conditional fixed-effects linear regression models, fit to physician-month data on prescribing, assessed the average within-prescriber change in oxycodone extended release prescribing over time as related to the 3 periods of media exposure. This method treats each physician as his or her own control and adjusts for physician attributes that are fixed over time. Effect modification was examined by including interaction terms between provider attributes (volume, specialty, region of practice) and media periods in the models. All models were weighted by the inverse of the expected sampling variability to address heteroscedasticity resulting from differences in the physicians’ prescribing volume.2,18 For models in which the outcome variable was the total proportion of opioid DDDs that were for oxycodone extended release per prescriber per month, the weight variable was the total number of DDDs prescribed between 2000 and 2007 by a prescriber. In models that
examined the proportion of strong opioid DDDs that were for oxycodone extended release, the total number of strong opioid DDDs per prescriber was used. All analyses were conducted using Stata version 12 (StataCorp, College Station, TX). Approval for this study was obtained from the Dalhousie University research ethics board.
Results Among the 924 newspaper articles published between 1995 and 2005 that discussed oxycodone extended release, 172 (18.6%) were from Canadian newspapers. Of these, 149 articles were from Nova Scotian newspapers, representing 86.6% of Canadian articles and 16.1% of all included articles. The earliest newspaper article was published in March 2000, nearly 4 years after the approval of oxycodone extended release in Canada. The peak in reporting in Canada (in March 2004) occurred nearly 3 years after the evident peak in the American coverage of oxycodone extended release (in July 2001). The peak in American coverage was largely due to a number of reports about increasing thefts and robberies of oxycodone extended release. In Canada, the peak in print media coverage also centered on a series of thefts, as well as reports of overdoses and deaths. More than 95% of articles identified or discussed at least 1 problem related to oxycodone extended release. Further details of the media around oxycodone extended release can be found in the original study by Whelan and colleagues.27 There were 4,212 prescribers included who made 2,803,273 opioid prescriptions to 461,585 unique patients between September 1996 and December 2007 (184,356 prescriber-months). The distribution of prescriber characteristics is shown in Table 1. Between 1996 and 2007, there was a steady trend toward increasing numbers of opioid prescriptions per month (Fig 1). This is due to increasing numbers of prescribers who made opioid prescriptions per month (from 1,225 to 1,475), increasing numbers of patients receiving opioids per month (from 15,750 to 19,000), and increasing numbers of strong opioid prescriptions per month (Fig 1). The percentage of all opioid prescriptions that were for strong opioids also rose steadily, from just under 15% to over 55%. There were 58,482 prescriptions for oxycodone extended release between 1996 and 2007, representing 2.1% of all opioid prescriptions and 6.3% of all strong opioid prescriptions. As Fig 2 indicates, the total number of DDDs prescribed that were for oxycodone extended release rose steadily prior to the American peak in 2001, continued to rise, but at a slower rate, prior to the Canadian peak in 2004, and then declined slightly until 2008. Overall, a total of 2,421,020 DDDs of oxycodone extended release were prescribed during this period (see Fig 3). Similarly, the prescribing of oxycodone extended release, as a proportion of strong opioid DDDs, was associated with the peaks of both American and Canadian media reporting of oxycodone extended release (Fig 4). Before peak media reports on oxycodone extended release (1996–2001), the proportion of strong
Borwein et al Table 1.
The Journal of Pain
1689
Prescriber Characteristics NUMBER OF PRESCRIPTIONS (IN THOUSANDS)
CHARACTERISTIC Specialty General practitioner Anesthesiologist Other Missing Graduation decade 1930–1949 1950–1959 1960–1969 1970–1979 1980–1989 1990–1999 2000–2007 Missing District health authority South Shore Health (population: 58,365) South West Health (population: 60,810) Annapolis Valley Health Authority (population: 81,475) Colchester East Hants Health Authority (population: 69,426) Cumberland Health Authority (population: 32,045) Pictou County Health Authority (population: 46,510) Guysborough Antigonish Strait Health Authority (population: 44,815) Cape Breton District Health Authority (population: 125,375) Capital Health (population: 394,639) Missing Volumey Low Low-Medium Medium-High High
NUMBER OF PRESCRIBERS (N = 4,212)
ALL OPIOIDS (N = 2,803,273)
STRONG OPIOIDS (N = 935,155)
OXYCODONE EXTENDED RELEASE (N = 58,482)
TOTAL OPIOID DDDs PRESCRIBED IN THOUSANDS (N = 63,627,915)
2,092 (49.8) 74 (1.8) 1,838 (43.6) 208 (4.9)
2,316 (82.6) 30 (1.1) 444 (15.8) 14 (.5)
834 (89.2) 17 (1.8) 82 (8.8) 1.6 (.2)
53 (90.4) 3.9 (6.6) 1.7 (3.0) .006 (.01)
58,048 (91.2) 726 (1.1) 4,727 (7.4) 127 (.2)
31 (.7) 129 (3.1) 254 (6.0) 597 (14.2) 922 (21.9) 1,014 (24.1) 829 (19.7) 436 (10.4)
22 (.8) 70 (2.5) 261 (9.3) 857 (30.6) 892 (31.8) 535 (19.1) 109 (3.9) 59 (2.1)
5.7 (.6) 14 (1.5) 81 (8.6) 275 (29.4) 306 (32.7) 205 (21.9) 43 (4.6) 5.8 (.6)
.2 (.3) .4 (.7) 8.1 (13.9) 18 (30.1) 18 (31.1) 12 (20.8) 1.3 (2.4) .04 (.1)
252 (.4) 1,303 (2.0) 6,551 (10.3) 20,532 (32.3) 21,520 (33.8) 11,538 (18.1) 1,407 (2.2) 525 (.8)
155 (3.7) 2.66*
208 (7.4) 3,566*
137 (3.3) 2.25*
94 (3.4) 1,553*
238 (5.7) 2.92*
258 (9.2) 3,169*
166 (3.9) 2.39*
153 (5.5) 2,206*
95 (2.3) 2.96*
62 (6.6) 1,054*
3.1 (5.3) 53*
5,612 (8.8) 96,155*
.8 (1.4) 14*
2,433 (3.8) 40,008*
3.7 (6.3) 45*
6,163 (9.7) 75,638*
59 (6.3) 843*
3.9 (6.7) 56*
4,119 (6.5) 59,336*
64 (2.3) 2,004*
21 (2.3) 658*
1.6 (2.8) 50
1,882 (3.0) 58,691*
116 (2.8) 2.49*
155 (5.5) 3,331*
35 (3.7) 744*
2.8 (4.8) 60*
2,897 (4.6) 62,283*
121 (2.9) 2.70*
114 (4.1) 2,550*
28 (3.0) 630*
2.7 (4.7) 61*
2,598 (4.1) 57,961*
390 (9.3) 3.11*
494 (17.6) 3,938* 158 (16.9) 1,263*
29 (3.1) 475* 101 (10.8) 1,238*
2,638 (62.6) 6.68* 1,245 (44.4) 3,155* 439 (46.9) 1,111* 156 (3.7)
17 (.6)
316 (7.5) 315 (7.4) 316 (7.5) 315 (7.4)
580 (20.6) 512 (18.3) 584 (20.8) 555 (19.8)
4.5 (.5) 158 (16.8) 178 (19.0) 243 (26.0) 244 (26.1)
19 (32.5) 152* 10,240 (16.1) 81,677* 21 (35.4) 52*
27,409 (43.1) 69,455*
.1 (.2)
276 (.4)
1.2 (2.0) 5.1 (8.8) 14.8 (25.3) 37.3 (63.7)
12,741 (20.0) 12,164 (19.1) 15,435 (24.3) 15,295 (24.0)
NOTE. Values are n (%). *Indicates values that are per 1,000 population. Note that these values are NOT given in thousands. yThe volume variable does not include the 2,459 prescribers that had very low prescribing of oxycodone extended release (1 prescription or less per month). Volume data only cover the years 2000 to 2007, the focus of the analysis, whereas the rest of the table covers the years 1996 to 2007.
opioid DDDs prescribed that were for oxycodone extended release generally increased. The trend continued, but at a slower rate after the onset of the American peak in 2001, and then shifted to decline in 2003 following the onset of the Canadian media peak. Changes in proportions were largely driven by changes in the numbers of prescriptions rather than variations in the quantity or strength of oxycodone extended release prescribed per prescription (data not shown), indicating that oxycodone extended release was increasingly replaced with prescriptions for other strong
opioids following the peak in Canadian media reporting. The significance of these shifts in trend was confirmed by the conditional fixed-effects regression results for both outcomes. For brevity, we present results only for the proportion of all strong opioid DDDs that were for oxycodone extended release, as the results were very similar for the proportion of all opioid DDDs that were for oxycodone extended release. For clarity, we present graphs of predicted trends by media period from the regressions, and report P values for the change in trend in the text.
1690
Media Effects on OxyContin Prescribing
The Journal of Pain
Figure 1. Total number of opioid prescriptions per month, by opioid type, 1996 to 2007.
Figure 3. Mean proportion of all opioid DDDs prescribed and strong opioid DDDs prescribed that were for oxycodone extended release, per prescriber per month, 1996 to 2007. The solid gray lines represent the absolute media peaks (with the American peak occurring in July 2001 and the Canadian peak in March 2004). The dotted lines represent the beginning of these peaks and correspond to February 2001 and May 2003, respectively.
For interval 1 (prior to the media peaks), the proportion of DDDs that were for oxycodone extended release increased steadily (P < .0001). Following the American media peak, as measured in interval 2, the prescribing of oxycodone extended release continued to increase, but at a noticeably slower rate than in interval 1 (P = .001). In turn, the proportion of strong opioid DDDs that were for oxycodone extended release declined by 30% (P < .0001) following the Canadian media peak (interval 3). These changes were not consistent across physician characteristics. With regard to prescribing volume (Fig 5), media effects were largely restricted to highervolume, and to a lesser extent, medium-volume prescribers. In terms of physician specialty (Figs 6 and 7), changes in prescribing were noted for general practitioners and, especially, anesthesiologists, who experienced a larger change following the Canadian media peak (P = .04). Finally, there were considerable media effects on prescribing by region of practice, with the
This research is among the first to examine the impact of media reporting of opioids and subsequent changes in prescribing of these drugs, and contributes to a small but growing literature that examines the effect of media and public concern on prescribing practices.1,10,14,22 A few studies have examined whether media attention can affect prescriber behavior. For example, 2 studies of thiazolidinedione10,22 and a study of calcium-channel blockers14 found that prescribing of both drugs declined
Figure 2. Total number of DDDs of oxycodone extended release prescribed per month, 1996 to 2007. The solid gray lines represent the absolute media peaks (with the American peak occurring in July 2001 and the Canadian peak in March 2004). The dotted lines represent the beginning of these peaks and correspond to February 2001 and May 2003, respectively.
Figure 4. Predicted fitted regression lines for the proportion of all opioid DDDs and strong opioid DDDs that were for oxycodone extended release. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. P values: All intervals were significant at the P = .05 level.
largest effects in Cape Breton District Health Authority (P < .0001), where most of the media attention occurred (data not shown).
Discussion
Borwein et al
Figure 5. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, by prescriber volume. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. P values: Low: all intervals not significant at the P = .05 level; Low-Medium: P = .0372 for interval 1, all other intervals nonsignificant; Medium-High: interval 1 and 3 significant at the P = .05 level, interval 2 nonsignificant; High: all intervals significant at the P = .05 level. following sustained media attention on studies that indicated cardiovascular safety concerns related to these drugs. Another example occurred with respect to the use of hormone replacement therapy in menopause following widespread media attention on several large trials (including the Women’s Health Initiative Study, the Million Women Study, and the Heart and Estrogen/ progestin Replacement Study). These large trials found increased risks of cardiovascular events and breast cancer associated with long-term hormone replacement therapy, and several subsequent studies noted substantial
Figure 6. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, by specialty. The dotted lines represent the knots in the splines, which were set at February 2001 and March 2003 to correspond with the beginning of the peaks in media. P values: Anesthesiologists: all intervals significant at the P = .05 level; General practitioners: all intervals significant at the P = .05 level; Other: interval 3 significant at the P = .05 level, intervals 1 and 2 nonsignificant.
The Journal of Pain
1691
Figure 7. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, general practitioners and other specialists. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. decreases in the use of hormone replacement therapy in North America and Europe following the publication of these results.6,11-13 Typically, such studies focus on changes in prescribing of specific drugs following media attention showing safety concerns, rather than broader concerns with diversion, crime, and addiction. In response to substantial and largely negative media attention, we observed a change in oxycodone extended release prescribing practices among Nova Scotia physicians. Media coverage portrayed the medication as a street drug and a social problem. Confirming our finding was that the media reporting did not affect overall opioid prescribing, but rather was restricted to oxycodone extended release. Against a backdrop of rising prescriptions for strong opioids, our findings point to a process of drug ‘‘switching,’’ in that oxycodone extended release prescriptions were increasingly replaced with prescriptions for other strong opioids. Although the total number of DDDs prescribed per month continued to rise, a first peak in media reporting, concentrated in U.S. sources, resulted in a slowing in the rate of oxycodone extended release prescribing. A subsequent Canadian peak in media reporting, concentrated in Nova Scotia, had a larger effect that reversed the trend entirely. Although strong media effects were observed, there was considerable heterogeneity in effects by physician attributes. Most affected were high-volume prescribers: those who wrote either many prescriptions for oxycodone extended release or prescriptions for larger amounts of oxycodone extended release. The potential implications of this are that prescribers who may have felt more comfortable prescribing oxycodone extended release initially were much less likely to do so following peak media attention, possibly restricting access to patients who were previously well managed on oxycodone extended release. We also found stronger effects for anesthesiologists, relative to general practitioners. Anesthesiologists
1692
The Journal of Pain
have been at the forefront of pain management and were the early adopters of oxycodone extended release; they appear to have been equally robust in responding to media concern by reducing their prescribing of oxycodone extended release. Equally possible, changes in prescribing were related to the ever-present threat of sanctioning from the provincial College aimed at ‘‘script-doctors,’’ physicians deemed to be involved in ‘‘inappropriate’’ prescribing and leading to negative outcomes.5,25 It is worth noting that although there were larger decreases among anesthesiologists with regard to oxycodone extended release prescribing, this group of practitioners is responsible for a much smaller proportion of all oxycodone extended release prescriptions compared to general practitioners. Additionally, anesthesiologists likely have a distinct patient population relative to other practitioners and thus differing health conditions for which they prescribe oxycodone extended release. For example, a subset of anesthesiologists work in chronic pain clinics, and it may be that changes in prescribing of oxycodone extended release among anesthesiologists were particularly salient among this patient group. Although every effort was made to address potential confounders through consultation with experts in the fields of pain management and medical education, it was not possible to assess all potential sources of measured and unmeasured confounding. For example, the degree to which prescribers in Nova Scotia were exposed to advertising of oxycodone extended release in academic journals, at professional meetings or conferences, and other marketing strategies that may have been used by drug companies has not been assessed in this study. Although the contribution of advertising to prescribing behavior has been well studied, such studies often show conflicting results. Indeed, a recent systematic review highlights these conflicting results and concluded that many studies have major methodological and design limitations.24 As such, the authors argue that it is difficult to speculate on the role of advertising activities in shaping prescribing practices; this also holds true in assessing advertising’s impact on the prescribing of oxycodone extended release in Nova Scotia. Similarly, another potential confounder is that in May 2007, 3 Purdue Pharma executives pled guilty to charges of misleading the public about the safety of oxycodone extended release and were fined more than $600 million.15 Although this event received considerable media coverage, no short-term changes (data not presented) in oxycodone extended release prescribing were observed in the prescription data following this event. Nonetheless, the magnitude of media effects that we observed in this study in the context of, and in comparison to, formal initiatives directed at changing physician practice (eg, practice guidelines and continuing education) are profound. Historically, the effects of such initiatives are small and often not significant.9,21,26 It is thus striking that an unofficial force, such as the media, was associated with substantial changes in physician prescribing.23
Media Effects on OxyContin Prescribing Interestingly, media attention may have been instrumental in motivating more recent initiatives directed at improving professional practice around pain management in Canada. Four deserve mention. First, debate has increased among national associations and in professional journals about the role of the physician, including pain specialists, in prescribing opioids for pain management.27 Second, Canada has adopted the 2010 Canadian Guideline for the Safe and Effective Use of Opioids for Chronic Non-Cancer Pain, directed at the appropriate prescribing of opioids for pain.16 Third, some provinces (ie, Nova Scotia, Ontario) have adopted mentoring programs to improve knowledge around opioids and pain management among general practitioners. Finally, Purdue Pharma has responded by replacing oxycodone extended release with OxyNeo, a tamper-resistant extended release oxycodone hydrochloride tablet, though generic versions of oxycodone extended release will remain available.17,20 Unfortunately, we were unable to assess changes on patient access or pain patient management. It was not possible to determine whether the changes in oxycodone extended release prescribing resulted in negative or positive outcomes for patients, nor was it possible to assess the motivation behind a prescriber’s choice of opioid prescription. Further, we could not determine which drugs patients were switched to if they were no longer prescribed oxycodone extended release. Rather, our aim was to determine whether media exposure around a particular drug could influence its prescribing rates, and the unanswered questions above are suggested as future research areas. We recognize that opioids are only one of many modalities that exist to treat chronic pain and that the evidence for their effectiveness varies by the type of pain being treated (eg, nociceptive pain, neuropathic pain, soft tissue pain).16 Nevertheless, it is likely that overt concern with addiction and crime has led to reduced access to medications for the treatment of pain among some patients. Moreover, just as prescribers have been exposed to negative media reports, so too have pain patients; these individuals may also have weighed the costs and benefits of oxycodone extended release use in the context of broader social concern. In the end, these findings highlight the dual role of media exposure in shaping medication prescribing practice—as a promoter, through advertising, but also with the potential to devalue a medication, leading to reductions in prescribing, when it is seen as problematic.
Acknowledgments The authors thank Denise Pellerin and Kirsten Crabtree at the Nova Scotia Prescription Monitoring Program for their assistance.
Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jpain.2013.08. 012.
Borwein et al
References 1. Avorn J, Chen M, Hartley R: Scientific versus commercial sources of influence on the prescribing behavior of physicians. Am J Med 73:4-8, 1982 2. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR: Introduction to Meta-analysis. Chichester, UK, John Wiley & Sons, 2009 3. Canadian Pain Society: Media coverage of OxyContin is one-sided: A statement from the Canadian Pain Society [press release]. Available at: http://www.canadianpainsociety.ca/ news/PressReleaseCPSAugust202004.pdf. Accessed August 19, 2004 4. Cicero TJ, Inciardi JA, Munoz A: Trends in abuse of OxyContin and other opioid analgesics in the United States: 2002-2004. J Pain 6:662-667, 2005 5. Dhalla IA, Mamdani MM, Gomes T, Juurlink DN: Clustering of opioid prescribing and opioid-related mortality among family physicians in Ontario. Can Fam Physician 57: e92-e96, 2011 6. Faber A, Bouvy ML, Loskamp L, van de Berg PB, Egberts TCG, de Jong-van den Berg LTW: Dramatic change in prescribing of hormone replacement therapy in the Netherlands after publication of the million Women Study: A follow-up study. Br J Clin Pharmacol 60:641-647, 2005 7. Fact Sheet: Misuse and abuse of oxycodone-based prescription drugs [Internet]. Health Canada. Available at: http://www.hc-sc.gc.ca/hc-ps/pubs/precurs/oxycodone/fs-fi/ index-eng.php. Accessed December 2, 2012
The Journal of Pain
1693
interventions and adverse news about calcium-channel blockers on first-line prescribing of antihypertensive drugs to elderly people in British Columbia. Lancet 352:943-948, 1998 15. Meier B: In guilty plea, OxyContin maker to pay $600 million. New York Times. Available at: http://www.nytimes. com/2007/05/10/business/11drug-web.html?_r=1&pagewan ted=print&. Accessed May 10, 2007 16. National Opioid Use Guideline Group (NOUGG): Canadian guideline for safe and effective use of opioids for chronic non-cancer pain. 2010. Available at: http:// nationalpaincentre.mcmaster.ca/opioid/ 17. National Post: Canada won’t block generic OxyContin: Health minister. November 19, 2012 18. Neter J, Wasserman W: Applied Linear Statistical Models; Regression, Analysis of Variance, and Experimental Designs. Homewood, IL, R.D. Irwin, 1974 19. Nova Scotia Prescription Monitoring Program: Nova Scotia Prescription Monitoring Program annual report: 2010-2011. 2011. Available at: http://www.nspmp.ca/ library/00000188-library.pdf 20. Purdue Pharma Std: Product Monograph: OxyNeo. Pickering, Ontario, Canada, Purdue Pharma, 2011 21. Sbarbaro JA: Can we influence prescribing patterns? Clin Infect Dis 33:S240-S244, 2001 22. Shah BR, Juurlink DN, Austin PC, Mamdani MM: New use of rosiglitazone decreased following publication of a metaanalysis suggesting harm. Diabet Med 25:871-874, 2008
8. Fisher JE, Zhang Y, Sketris I, Johnston G, Burge F: The effect of an educational intervention on meperidine use in Nova Scotia, Canada: A time series analysis. Pharmacoepidemiol Drug Saf 21:177-183, 2012
23. Sketris IS, Langille Ingram EM, Lummis HL: Strategic opportunities for effective optimal prescribing and medication management. Can J Clin Pharmacol 16:e103-e125, 2009
9. FitzGerald JM, Boulet LP, McIvor RA, Zimmerman S, Chapman KR: Asthma control in Canada remains suboptimal: The Reality of Asthma Control (TRAC) study. Can Respir J 13:253-259, 2006
24. Spurling GK, Mansfield PR, Montgomery BD, Lexchin J, Doust J, Othman N, Vitry AI: Information from pharmaceutical companies and the quality, quantity, and cost of physicians’ prescribing: A systematic review. PLoS Med 7: e1000352, 2010
10. George J, Hannah S 5th, Lang CC: Thiazolidinediones and the influence of media adverse reporting on prescribing attitudes in practice (TZD-IMPACT) study. Cardiovasc Ther 27:83-88, 2009
25. The Chronicle Herald. Doctor’s suspension worries groups. May 3, 2006
11. Haas JS, Kaplan CP, Gerstenberger EP, Kerlikowske K: Changes in the use of postmenopausal hormone replacement therapy after the publication of clinical trial results. Ann Intern Med 140:184-188, 2004 12. Hersch AL, Stefanick ML, Stafford RS: National use of postmenopausal hormone therapy: National trends and response to recent evidence. J Am Med Assoc 291:47-53, 2004 13. Lawton B, Rose S, McLeod D, Dowell A: Changes in the use of hormone replacement therapy after the report from the Women’s Health Initiative: Cross sectional survey of users. BMJ 327:845-846, 2003 14. Maclure M, Dormuth C, Naumann T, McCormack J, Rangno R, Whiteside C, Wright JM: Influences of educational
26. Victor TW, Alvarez NA, Gould E: Opioid prescribing practices in chronic pain management: Guidelines do not sufficiently influence clinical practice. J Pain 10:1051-1057, 2009 27. Whelan E, Asbridge M, Haydt S: Representations of OxyContin in North American newspapers and medical journals. Pain Res Manag 16:252-258, 2011 28. WHO ATC/DDD index 2012. Oslo, Norway, WHO Collaborating Centre for Drug Statistics Methodology. 2011. Available at: http://www.whocc.no/atc_ddd_index/. Accessed December 2, 2012 29. WHO Collaborating Centre for Drug Statistics Methodology: Guidelines for ATC classification and DDD assignment 2012. 15th ed. Oslo, Norway. 2011. Available at: http://www. whocc.no/filearchive/publications/2012_guidelines_with_ front_pa.pdf
1693.e1
Media Effects on OxyContin Prescribing
The Journal of Pain
Supplementary Appendix.
Newspapers Included
in the Sample
CANADIAN NEWSPAPERS
AMERICAN NEWSPAPERS
West The Vancouver Sun Midwest The Winnipeg Free Press Central The Globe and Mail (Toronto) The Montreal Gazette East Coast The Chronicle Herald (Halifax) The Cape Breton Post
West The Denver Post The San Francisco Chronicle The Seattle Times The Houston Chronicle Midwest The Chicago Sun-Times The Cleveland Plain Dealer The Columbus Dispatch The Omaha World Herald The St Louis Post-Dispatch The Minnesota Star Tribune Northeast The Boston Globe The Boston Herald The Buffalo News The New York Daily News The Pittsburgh Post-Gazette The Wall Street Journal Southeast The Atlanta Journal The Tampa Tribune The Washington Post National USA Today
Prescribing Practices Amid the OxyContin Crisis: Examining the Effect of Print Media Coverage on Opioid Prescribing Among Physicians Alexandra Borwein,* George Kephart,* Emma Whelan,y and Mark Asbridge*,y Departments of *Community Health and Epidemiology and ySociology and Social Anthropology, Dalhousie University, Halifax, Nova Scotia, Canada.
Abstract: The pain medication OxyContin (hereafter referred to as oxycodone extended release) has been the subject of sustained, and largely negative, media attention in recent years. We sought to determine whether media coverage of oxycodone extended release in North American newspapers has led to changes in prescribing of the drug in Nova Scotia, Canada. An interrupted time-series design examined the effect of media attention on physicians’ monthly prescribing of opioids. The outcome measures were, for each physician, the monthly proportions of all opioids prescribed and the proportion of strong opioids prescribed that were for oxycodone extended release. The exposure of interest was media attention defined as the number of articles published each month in 27 North American newspapers. Variations in media effects by provider characteristics (specialty, prescribing volume, and region) were assessed. Within-provider changes in the prescribing of oxycodone extended release in Nova Scotia were observed, and they followed changes in media coverage. Oxycodone extended release prescribing rose steadily prior to receiving media attention. Following peak media attention in the United States, the prescribing of oxycodone extended release slowed. Likewise, following peak coverage in Canadian newspapers, the prescribing of oxycodone extended release declined. These patterns were observed across prescriber specialties and by prescriber volume, though the magnitude of change in prescribing varied. Perspective: This study demonstrates that print media reporting of oxycodone extended release in North American newspapers, and its continued portrayal as a social problem, coincided with reductions in the prescribing of oxycodone extended release by physicians in Nova Scotia. ª 2013 by the American Pain Society Key words: Opioids, OxyContin, prescribing practices, media effects.
O
xyContin (Purdue Pharma LP, Stamford, CT) is an extended release form of oxycodone that was approved for use in Canada in 1996. When introduced, oxycodone extended release was considered to be a breakthrough for its ability to provide sustained Received February 26, 2013; Revised August 16, 2013; Accepted August 27, 2013. A.B. was supported by a Canadian Institutes for Health Research master’s award, in addition to a Scotia Scholarship and a Scotia Support Grant from the Nova Scotia Health Research Foundation. The study was supported, in part, by a Health Research Project Grant from the Nova Scotia Health Research Foundation. The authors declare no conflicts of interest. Supplementary data accompanying this article are available online at www.jpain.org/ and www.sciencedirect.com/. Address reprint requests to Mark Asbridge, PhD, Department of Community Health and Epidemiology, Dalhousie University, Centre for Clinical Research, 4th Floor, 5790 University Avenue, Halifax, Nova Scotia, Canada B3H 1V7. E-mail: [email protected] 1526-5900/$36.00 ª 2013 by the American Pain Society http://dx.doi.org/10.1016/j.jpain.2013.08.012
1686
pain relief and because prescribers expected it to be less prone to abuse than other opioids.4 Soon after the introduction of oxycodone extended release, however, it was discovered that when crushed or chewed and inhaled, injected, or swallowed, the oxycodone is released and absorbed rapidly, producing a heroin-like euphoria.7 As a result, within 5 years, the popular press and medical authorities in parts of North America began to report the use of oxycodone extended release as a street drug and a growing social problem.27 This was particularly the case in the Atlantic Coast of Canada and the United States, where many reports surfaced about abuse and addiction, criminal diversion of the drug, as well as oxycodone extended release–related overdoses and deaths.27 In examining these events, a key question is whether the increased, and largely negative, media attention on oxycodone extended release affected prescriber practices. On the one hand, negative media coverage may
Borwein et al propagate fears among prescribers, including concerns about drug diversion, addiction, and crime and the threat of being sanctioned by governing bodies in medicine.3 Alternatively, it has been suggested that users of oxycodone extended release may face stigmatization and reduced access to therapy from wary prescribers.3,27 It is important to understand how media portrayals of oxycodone extended release may have influenced prescribing practices among Canadian physicians. Our objective was to determine whether media reporting of oxycodone extended release in North American newspapers led to changes in the prescribing of the drug in Nova Scotia between 1996 and 2007.
Methods An interrupted time-series design was used to examine the relationship between media attention on physicians’ monthly prescribing of oxycodone extended release. The outcome measures were, for each physician who prescribed opioids, the monthly proportions of all opioids and strong opioids that were for oxycodone extended release. The exposure of interest was media attention defined as the number of articles published per month based on a content analysis of 27 North American newspapers.27 Variations in media effects by provider characteristics (specialty, prescribing volume, and region of practice) were assessed.
Data Source Administrative data on all opioid analgesic prescriptions in Nova Scotia between September 1, 1996 (the first month in which oxycodone extended release was prescribed), and December 31, 2007, were obtained from the Nova Scotia Prescription Monitoring Program, which maintains a database with patient, prescriber, and drugrelated information for all monitored prescription drugs, including opioids, dispensed in Nova Scotia since 1992.8,19 These data do not include prescriptions filled in hospitals or long-term care facilities, or those without a prescriber identification number.
Outcome Variables Two outcome variables were examined. The first outcome variable was the proportion of physicians’ monthly volume of all opioids prescribed that were for oxycodone extended release. The second was the proportion of physicians’ monthly volume of all strong opioids that were for oxycodone extended release. Oxycodone extended release is a strong opioid, and restricting analyses to strong opioids provided a comparison with opioids that are often prescribed for the same or similar conditions. This is particularly relevant in the context of concern over oxycodone extended release, where providers wishing to move away from this drug would be inclined to prescribe another strong opioid. Opioids were classified as strong or weak based on the 2010 Canadian Guideline for the Safe and Effective Use of Opioids for Chronic Non-Cancer Pain.16 Briefly, strong opioids were defined as medications containing sufentanil, fentanyl,
The Journal of Pain
1687
hydromorphone, morphine, and oxycodone. Weak opioids were defined as opioid medications that contained codeine, pethidine (meperidine), or dextropropoxyphene. Outcome variables were highly positively skewed and were therefore logit transformed in all statistical models. To compute prescribing volume across medications with different strengths and normal doses, we employed the Defined Daily Dose (DDD) methodology developed by the World Health Organization.29 The DDD is defined as the ‘‘assumed average maintenance dose per day for a drug used for its main indication in adults.’’29 DDDs for opioids were obtained from the online searchable ATC/ DDD Index for 2012.28 For drugs that contained codeine or propoxyphene in combination with other nonopioid ingredients (eg, acetaminophen), only the opioid portion was considered. Using the strength of the prescription and the total quantity prescribed (the number of tablets, total volume, or number of patches), the total number of DDDs per prescription was calculated and then aggregated by physician and month to compute the outcome variables. The major disadvantage of the DDD methodology is that not all drugs have assigned DDDs. The DDD is determined based on international drug information and may not necessarily reflect prescribing patterns of the associated drug in Canada, as main drug indications and prescribed daily doses may vary between countries. However, in the context of the research presented here, the advantages conferred by allowing comparisons across prescriptions outweighed the disadvantages of the system. In particular, the use of DDDs allowed background opioid prescription rates to be taken into account. We assessed whether the effects of media on prescribing of oxycodone extended release varied by physicians’ overall prescribing volume, specialty, and region of practice. Prescribers were stratified into 4 volume groups, for each of the outcome variables, based on their mean volume of opioids (strong opioids) prescribed across all months. Groups were created using the quartiles of the means for each outcome variable (termed ‘‘Low,’’ ‘‘Low-Medium,’’ ‘‘Medium-High,’’ and ‘‘High’’). Prescriber specialties were grouped into 3 categories (general practitioners, anesthesiologists, and other). Region of practice was defined according to 9 regional District Health Authorities in Nova Scotia.
Explanatory Variable Data on newspaper coverage of oxycodone extended release were drawn from a previous study that examined representations of oxycodone extended release in North American newspapers between 1995 and 2005.27 Searches of newspaper literature databases were conducted and 924 stories published in 27 North American newspapers were examined, with 172 (18.6%) from Canadian newspapers (149 of which were from Nova Scotia; see Supplementary Appendix for a full list). More than 95% of articles identified at least 1 problem related to oxycodone extended release. Analysis of the number of
1688
Media Effects on OxyContin Prescribing
The Journal of Pain
unique stories published each month revealed 2 peak periods of media attention. The first, which began in February 2001 and peaked in July 2001, was largely in U.S. sources. The second, which began in May 2003 and peaked in March 2004, corresponded to attention in Canadian sources. For our statistical models, a media exposure measure was created using linear splines with knots defined according to the onset of the peaks in print media reporting. The linear trends in prescribing are captured as splines with knots created to represent the 3 intervals in the splines: prior to February 2001, February 2001 to April 2003, and May 2003 to December 2007. The use of splines allowed for smoothing of the media data and production of coefficients in the statistical models that represented the change in slope from the preceding interval; that is, the splines allowed for the measurement of the change in prescribing of oxycodone extended release for each time interval compared to the preceding interval.
Possible Confounders A number of confounders were considered in the context of this study, including the possibility of ongoing medical education activities around oxycodone extended release and opioids, the introduction of new products to the market, and the publication of guidelines related to the prescribing of opioids. Investigations into such activities, including consultations with several experts in the fields of pain medicine and medical education, indicated that there were no such activities in the province of Nova Scotia or across Canada that would impact the prescribing of oxycodone extended release in the province during this time.
Statistical Analyses Descriptive analyses examined the time period between September 1996 and December 2007, inclusive. Statistical models to measure effects of media exposure were restricted to the time period between January 2000 and December 2007, because of the small numbers of oxycodone extended release prescriptions prior to 2000. Conditional fixed-effects linear regression models, fit to physician-month data on prescribing, assessed the average within-prescriber change in oxycodone extended release prescribing over time as related to the 3 periods of media exposure. This method treats each physician as his or her own control and adjusts for physician attributes that are fixed over time. Effect modification was examined by including interaction terms between provider attributes (volume, specialty, region of practice) and media periods in the models. All models were weighted by the inverse of the expected sampling variability to address heteroscedasticity resulting from differences in the physicians’ prescribing volume.2,18 For models in which the outcome variable was the total proportion of opioid DDDs that were for oxycodone extended release per prescriber per month, the weight variable was the total number of DDDs prescribed between 2000 and 2007 by a prescriber. In models that
examined the proportion of strong opioid DDDs that were for oxycodone extended release, the total number of strong opioid DDDs per prescriber was used. All analyses were conducted using Stata version 12 (StataCorp, College Station, TX). Approval for this study was obtained from the Dalhousie University research ethics board.
Results Among the 924 newspaper articles published between 1995 and 2005 that discussed oxycodone extended release, 172 (18.6%) were from Canadian newspapers. Of these, 149 articles were from Nova Scotian newspapers, representing 86.6% of Canadian articles and 16.1% of all included articles. The earliest newspaper article was published in March 2000, nearly 4 years after the approval of oxycodone extended release in Canada. The peak in reporting in Canada (in March 2004) occurred nearly 3 years after the evident peak in the American coverage of oxycodone extended release (in July 2001). The peak in American coverage was largely due to a number of reports about increasing thefts and robberies of oxycodone extended release. In Canada, the peak in print media coverage also centered on a series of thefts, as well as reports of overdoses and deaths. More than 95% of articles identified or discussed at least 1 problem related to oxycodone extended release. Further details of the media around oxycodone extended release can be found in the original study by Whelan and colleagues.27 There were 4,212 prescribers included who made 2,803,273 opioid prescriptions to 461,585 unique patients between September 1996 and December 2007 (184,356 prescriber-months). The distribution of prescriber characteristics is shown in Table 1. Between 1996 and 2007, there was a steady trend toward increasing numbers of opioid prescriptions per month (Fig 1). This is due to increasing numbers of prescribers who made opioid prescriptions per month (from 1,225 to 1,475), increasing numbers of patients receiving opioids per month (from 15,750 to 19,000), and increasing numbers of strong opioid prescriptions per month (Fig 1). The percentage of all opioid prescriptions that were for strong opioids also rose steadily, from just under 15% to over 55%. There were 58,482 prescriptions for oxycodone extended release between 1996 and 2007, representing 2.1% of all opioid prescriptions and 6.3% of all strong opioid prescriptions. As Fig 2 indicates, the total number of DDDs prescribed that were for oxycodone extended release rose steadily prior to the American peak in 2001, continued to rise, but at a slower rate, prior to the Canadian peak in 2004, and then declined slightly until 2008. Overall, a total of 2,421,020 DDDs of oxycodone extended release were prescribed during this period (see Fig 3). Similarly, the prescribing of oxycodone extended release, as a proportion of strong opioid DDDs, was associated with the peaks of both American and Canadian media reporting of oxycodone extended release (Fig 4). Before peak media reports on oxycodone extended release (1996–2001), the proportion of strong
Borwein et al Table 1.
The Journal of Pain
1689
Prescriber Characteristics NUMBER OF PRESCRIPTIONS (IN THOUSANDS)
CHARACTERISTIC Specialty General practitioner Anesthesiologist Other Missing Graduation decade 1930–1949 1950–1959 1960–1969 1970–1979 1980–1989 1990–1999 2000–2007 Missing District health authority South Shore Health (population: 58,365) South West Health (population: 60,810) Annapolis Valley Health Authority (population: 81,475) Colchester East Hants Health Authority (population: 69,426) Cumberland Health Authority (population: 32,045) Pictou County Health Authority (population: 46,510) Guysborough Antigonish Strait Health Authority (population: 44,815) Cape Breton District Health Authority (population: 125,375) Capital Health (population: 394,639) Missing Volumey Low Low-Medium Medium-High High
NUMBER OF PRESCRIBERS (N = 4,212)
ALL OPIOIDS (N = 2,803,273)
STRONG OPIOIDS (N = 935,155)
OXYCODONE EXTENDED RELEASE (N = 58,482)
TOTAL OPIOID DDDs PRESCRIBED IN THOUSANDS (N = 63,627,915)
2,092 (49.8) 74 (1.8) 1,838 (43.6) 208 (4.9)
2,316 (82.6) 30 (1.1) 444 (15.8) 14 (.5)
834 (89.2) 17 (1.8) 82 (8.8) 1.6 (.2)
53 (90.4) 3.9 (6.6) 1.7 (3.0) .006 (.01)
58,048 (91.2) 726 (1.1) 4,727 (7.4) 127 (.2)
31 (.7) 129 (3.1) 254 (6.0) 597 (14.2) 922 (21.9) 1,014 (24.1) 829 (19.7) 436 (10.4)
22 (.8) 70 (2.5) 261 (9.3) 857 (30.6) 892 (31.8) 535 (19.1) 109 (3.9) 59 (2.1)
5.7 (.6) 14 (1.5) 81 (8.6) 275 (29.4) 306 (32.7) 205 (21.9) 43 (4.6) 5.8 (.6)
.2 (.3) .4 (.7) 8.1 (13.9) 18 (30.1) 18 (31.1) 12 (20.8) 1.3 (2.4) .04 (.1)
252 (.4) 1,303 (2.0) 6,551 (10.3) 20,532 (32.3) 21,520 (33.8) 11,538 (18.1) 1,407 (2.2) 525 (.8)
155 (3.7) 2.66*
208 (7.4) 3,566*
137 (3.3) 2.25*
94 (3.4) 1,553*
238 (5.7) 2.92*
258 (9.2) 3,169*
166 (3.9) 2.39*
153 (5.5) 2,206*
95 (2.3) 2.96*
62 (6.6) 1,054*
3.1 (5.3) 53*
5,612 (8.8) 96,155*
.8 (1.4) 14*
2,433 (3.8) 40,008*
3.7 (6.3) 45*
6,163 (9.7) 75,638*
59 (6.3) 843*
3.9 (6.7) 56*
4,119 (6.5) 59,336*
64 (2.3) 2,004*
21 (2.3) 658*
1.6 (2.8) 50
1,882 (3.0) 58,691*
116 (2.8) 2.49*
155 (5.5) 3,331*
35 (3.7) 744*
2.8 (4.8) 60*
2,897 (4.6) 62,283*
121 (2.9) 2.70*
114 (4.1) 2,550*
28 (3.0) 630*
2.7 (4.7) 61*
2,598 (4.1) 57,961*
390 (9.3) 3.11*
494 (17.6) 3,938* 158 (16.9) 1,263*
29 (3.1) 475* 101 (10.8) 1,238*
2,638 (62.6) 6.68* 1,245 (44.4) 3,155* 439 (46.9) 1,111* 156 (3.7)
17 (.6)
316 (7.5) 315 (7.4) 316 (7.5) 315 (7.4)
580 (20.6) 512 (18.3) 584 (20.8) 555 (19.8)
4.5 (.5) 158 (16.8) 178 (19.0) 243 (26.0) 244 (26.1)
19 (32.5) 152* 10,240 (16.1) 81,677* 21 (35.4) 52*
27,409 (43.1) 69,455*
.1 (.2)
276 (.4)
1.2 (2.0) 5.1 (8.8) 14.8 (25.3) 37.3 (63.7)
12,741 (20.0) 12,164 (19.1) 15,435 (24.3) 15,295 (24.0)
NOTE. Values are n (%). *Indicates values that are per 1,000 population. Note that these values are NOT given in thousands. yThe volume variable does not include the 2,459 prescribers that had very low prescribing of oxycodone extended release (1 prescription or less per month). Volume data only cover the years 2000 to 2007, the focus of the analysis, whereas the rest of the table covers the years 1996 to 2007.
opioid DDDs prescribed that were for oxycodone extended release generally increased. The trend continued, but at a slower rate after the onset of the American peak in 2001, and then shifted to decline in 2003 following the onset of the Canadian media peak. Changes in proportions were largely driven by changes in the numbers of prescriptions rather than variations in the quantity or strength of oxycodone extended release prescribed per prescription (data not shown), indicating that oxycodone extended release was increasingly replaced with prescriptions for other strong
opioids following the peak in Canadian media reporting. The significance of these shifts in trend was confirmed by the conditional fixed-effects regression results for both outcomes. For brevity, we present results only for the proportion of all strong opioid DDDs that were for oxycodone extended release, as the results were very similar for the proportion of all opioid DDDs that were for oxycodone extended release. For clarity, we present graphs of predicted trends by media period from the regressions, and report P values for the change in trend in the text.
1690
Media Effects on OxyContin Prescribing
The Journal of Pain
Figure 1. Total number of opioid prescriptions per month, by opioid type, 1996 to 2007.
Figure 3. Mean proportion of all opioid DDDs prescribed and strong opioid DDDs prescribed that were for oxycodone extended release, per prescriber per month, 1996 to 2007. The solid gray lines represent the absolute media peaks (with the American peak occurring in July 2001 and the Canadian peak in March 2004). The dotted lines represent the beginning of these peaks and correspond to February 2001 and May 2003, respectively.
For interval 1 (prior to the media peaks), the proportion of DDDs that were for oxycodone extended release increased steadily (P < .0001). Following the American media peak, as measured in interval 2, the prescribing of oxycodone extended release continued to increase, but at a noticeably slower rate than in interval 1 (P = .001). In turn, the proportion of strong opioid DDDs that were for oxycodone extended release declined by 30% (P < .0001) following the Canadian media peak (interval 3). These changes were not consistent across physician characteristics. With regard to prescribing volume (Fig 5), media effects were largely restricted to highervolume, and to a lesser extent, medium-volume prescribers. In terms of physician specialty (Figs 6 and 7), changes in prescribing were noted for general practitioners and, especially, anesthesiologists, who experienced a larger change following the Canadian media peak (P = .04). Finally, there were considerable media effects on prescribing by region of practice, with the
This research is among the first to examine the impact of media reporting of opioids and subsequent changes in prescribing of these drugs, and contributes to a small but growing literature that examines the effect of media and public concern on prescribing practices.1,10,14,22 A few studies have examined whether media attention can affect prescriber behavior. For example, 2 studies of thiazolidinedione10,22 and a study of calcium-channel blockers14 found that prescribing of both drugs declined
Figure 2. Total number of DDDs of oxycodone extended release prescribed per month, 1996 to 2007. The solid gray lines represent the absolute media peaks (with the American peak occurring in July 2001 and the Canadian peak in March 2004). The dotted lines represent the beginning of these peaks and correspond to February 2001 and May 2003, respectively.
Figure 4. Predicted fitted regression lines for the proportion of all opioid DDDs and strong opioid DDDs that were for oxycodone extended release. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. P values: All intervals were significant at the P = .05 level.
largest effects in Cape Breton District Health Authority (P < .0001), where most of the media attention occurred (data not shown).
Discussion
Borwein et al
Figure 5. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, by prescriber volume. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. P values: Low: all intervals not significant at the P = .05 level; Low-Medium: P = .0372 for interval 1, all other intervals nonsignificant; Medium-High: interval 1 and 3 significant at the P = .05 level, interval 2 nonsignificant; High: all intervals significant at the P = .05 level. following sustained media attention on studies that indicated cardiovascular safety concerns related to these drugs. Another example occurred with respect to the use of hormone replacement therapy in menopause following widespread media attention on several large trials (including the Women’s Health Initiative Study, the Million Women Study, and the Heart and Estrogen/ progestin Replacement Study). These large trials found increased risks of cardiovascular events and breast cancer associated with long-term hormone replacement therapy, and several subsequent studies noted substantial
Figure 6. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, by specialty. The dotted lines represent the knots in the splines, which were set at February 2001 and March 2003 to correspond with the beginning of the peaks in media. P values: Anesthesiologists: all intervals significant at the P = .05 level; General practitioners: all intervals significant at the P = .05 level; Other: interval 3 significant at the P = .05 level, intervals 1 and 2 nonsignificant.
The Journal of Pain
1691
Figure 7. Predicted fitted regression lines for the proportion of strong opioid DDDs that were for oxycodone extended release, general practitioners and other specialists. The dotted lines represent the knots in the splines, which were set at February 2001 and May 2003 to correspond with the beginning of the peaks in media. decreases in the use of hormone replacement therapy in North America and Europe following the publication of these results.6,11-13 Typically, such studies focus on changes in prescribing of specific drugs following media attention showing safety concerns, rather than broader concerns with diversion, crime, and addiction. In response to substantial and largely negative media attention, we observed a change in oxycodone extended release prescribing practices among Nova Scotia physicians. Media coverage portrayed the medication as a street drug and a social problem. Confirming our finding was that the media reporting did not affect overall opioid prescribing, but rather was restricted to oxycodone extended release. Against a backdrop of rising prescriptions for strong opioids, our findings point to a process of drug ‘‘switching,’’ in that oxycodone extended release prescriptions were increasingly replaced with prescriptions for other strong opioids. Although the total number of DDDs prescribed per month continued to rise, a first peak in media reporting, concentrated in U.S. sources, resulted in a slowing in the rate of oxycodone extended release prescribing. A subsequent Canadian peak in media reporting, concentrated in Nova Scotia, had a larger effect that reversed the trend entirely. Although strong media effects were observed, there was considerable heterogeneity in effects by physician attributes. Most affected were high-volume prescribers: those who wrote either many prescriptions for oxycodone extended release or prescriptions for larger amounts of oxycodone extended release. The potential implications of this are that prescribers who may have felt more comfortable prescribing oxycodone extended release initially were much less likely to do so following peak media attention, possibly restricting access to patients who were previously well managed on oxycodone extended release. We also found stronger effects for anesthesiologists, relative to general practitioners. Anesthesiologists
1692
The Journal of Pain
have been at the forefront of pain management and were the early adopters of oxycodone extended release; they appear to have been equally robust in responding to media concern by reducing their prescribing of oxycodone extended release. Equally possible, changes in prescribing were related to the ever-present threat of sanctioning from the provincial College aimed at ‘‘script-doctors,’’ physicians deemed to be involved in ‘‘inappropriate’’ prescribing and leading to negative outcomes.5,25 It is worth noting that although there were larger decreases among anesthesiologists with regard to oxycodone extended release prescribing, this group of practitioners is responsible for a much smaller proportion of all oxycodone extended release prescriptions compared to general practitioners. Additionally, anesthesiologists likely have a distinct patient population relative to other practitioners and thus differing health conditions for which they prescribe oxycodone extended release. For example, a subset of anesthesiologists work in chronic pain clinics, and it may be that changes in prescribing of oxycodone extended release among anesthesiologists were particularly salient among this patient group. Although every effort was made to address potential confounders through consultation with experts in the fields of pain management and medical education, it was not possible to assess all potential sources of measured and unmeasured confounding. For example, the degree to which prescribers in Nova Scotia were exposed to advertising of oxycodone extended release in academic journals, at professional meetings or conferences, and other marketing strategies that may have been used by drug companies has not been assessed in this study. Although the contribution of advertising to prescribing behavior has been well studied, such studies often show conflicting results. Indeed, a recent systematic review highlights these conflicting results and concluded that many studies have major methodological and design limitations.24 As such, the authors argue that it is difficult to speculate on the role of advertising activities in shaping prescribing practices; this also holds true in assessing advertising’s impact on the prescribing of oxycodone extended release in Nova Scotia. Similarly, another potential confounder is that in May 2007, 3 Purdue Pharma executives pled guilty to charges of misleading the public about the safety of oxycodone extended release and were fined more than $600 million.15 Although this event received considerable media coverage, no short-term changes (data not presented) in oxycodone extended release prescribing were observed in the prescription data following this event. Nonetheless, the magnitude of media effects that we observed in this study in the context of, and in comparison to, formal initiatives directed at changing physician practice (eg, practice guidelines and continuing education) are profound. Historically, the effects of such initiatives are small and often not significant.9,21,26 It is thus striking that an unofficial force, such as the media, was associated with substantial changes in physician prescribing.23
Media Effects on OxyContin Prescribing Interestingly, media attention may have been instrumental in motivating more recent initiatives directed at improving professional practice around pain management in Canada. Four deserve mention. First, debate has increased among national associations and in professional journals about the role of the physician, including pain specialists, in prescribing opioids for pain management.27 Second, Canada has adopted the 2010 Canadian Guideline for the Safe and Effective Use of Opioids for Chronic Non-Cancer Pain, directed at the appropriate prescribing of opioids for pain.16 Third, some provinces (ie, Nova Scotia, Ontario) have adopted mentoring programs to improve knowledge around opioids and pain management among general practitioners. Finally, Purdue Pharma has responded by replacing oxycodone extended release with OxyNeo, a tamper-resistant extended release oxycodone hydrochloride tablet, though generic versions of oxycodone extended release will remain available.17,20 Unfortunately, we were unable to assess changes on patient access or pain patient management. It was not possible to determine whether the changes in oxycodone extended release prescribing resulted in negative or positive outcomes for patients, nor was it possible to assess the motivation behind a prescriber’s choice of opioid prescription. Further, we could not determine which drugs patients were switched to if they were no longer prescribed oxycodone extended release. Rather, our aim was to determine whether media exposure around a particular drug could influence its prescribing rates, and the unanswered questions above are suggested as future research areas. We recognize that opioids are only one of many modalities that exist to treat chronic pain and that the evidence for their effectiveness varies by the type of pain being treated (eg, nociceptive pain, neuropathic pain, soft tissue pain).16 Nevertheless, it is likely that overt concern with addiction and crime has led to reduced access to medications for the treatment of pain among some patients. Moreover, just as prescribers have been exposed to negative media reports, so too have pain patients; these individuals may also have weighed the costs and benefits of oxycodone extended release use in the context of broader social concern. In the end, these findings highlight the dual role of media exposure in shaping medication prescribing practice—as a promoter, through advertising, but also with the potential to devalue a medication, leading to reductions in prescribing, when it is seen as problematic.
Acknowledgments The authors thank Denise Pellerin and Kirsten Crabtree at the Nova Scotia Prescription Monitoring Program for their assistance.
Supplementary Data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jpain.2013.08. 012.
Borwein et al
References 1. Avorn J, Chen M, Hartley R: Scientific versus commercial sources of influence on the prescribing behavior of physicians. Am J Med 73:4-8, 1982 2. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR: Introduction to Meta-analysis. Chichester, UK, John Wiley & Sons, 2009 3. Canadian Pain Society: Media coverage of OxyContin is one-sided: A statement from the Canadian Pain Society [press release]. Available at: http://www.canadianpainsociety.ca/ news/PressReleaseCPSAugust202004.pdf. Accessed August 19, 2004 4. Cicero TJ, Inciardi JA, Munoz A: Trends in abuse of OxyContin and other opioid analgesics in the United States: 2002-2004. J Pain 6:662-667, 2005 5. Dhalla IA, Mamdani MM, Gomes T, Juurlink DN: Clustering of opioid prescribing and opioid-related mortality among family physicians in Ontario. Can Fam Physician 57: e92-e96, 2011 6. Faber A, Bouvy ML, Loskamp L, van de Berg PB, Egberts TCG, de Jong-van den Berg LTW: Dramatic change in prescribing of hormone replacement therapy in the Netherlands after publication of the million Women Study: A follow-up study. Br J Clin Pharmacol 60:641-647, 2005 7. Fact Sheet: Misuse and abuse of oxycodone-based prescription drugs [Internet]. Health Canada. Available at: http://www.hc-sc.gc.ca/hc-ps/pubs/precurs/oxycodone/fs-fi/ index-eng.php. Accessed December 2, 2012
The Journal of Pain
1693
interventions and adverse news about calcium-channel blockers on first-line prescribing of antihypertensive drugs to elderly people in British Columbia. Lancet 352:943-948, 1998 15. Meier B: In guilty plea, OxyContin maker to pay $600 million. New York Times. Available at: http://www.nytimes. com/2007/05/10/business/11drug-web.html?_r=1&pagewan ted=print&. Accessed May 10, 2007 16. National Opioid Use Guideline Group (NOUGG): Canadian guideline for safe and effective use of opioids for chronic non-cancer pain. 2010. Available at: http:// nationalpaincentre.mcmaster.ca/opioid/ 17. National Post: Canada won’t block generic OxyContin: Health minister. November 19, 2012 18. Neter J, Wasserman W: Applied Linear Statistical Models; Regression, Analysis of Variance, and Experimental Designs. Homewood, IL, R.D. Irwin, 1974 19. Nova Scotia Prescription Monitoring Program: Nova Scotia Prescription Monitoring Program annual report: 2010-2011. 2011. Available at: http://www.nspmp.ca/ library/00000188-library.pdf 20. Purdue Pharma Std: Product Monograph: OxyNeo. Pickering, Ontario, Canada, Purdue Pharma, 2011 21. Sbarbaro JA: Can we influence prescribing patterns? Clin Infect Dis 33:S240-S244, 2001 22. Shah BR, Juurlink DN, Austin PC, Mamdani MM: New use of rosiglitazone decreased following publication of a metaanalysis suggesting harm. Diabet Med 25:871-874, 2008
8. Fisher JE, Zhang Y, Sketris I, Johnston G, Burge F: The effect of an educational intervention on meperidine use in Nova Scotia, Canada: A time series analysis. Pharmacoepidemiol Drug Saf 21:177-183, 2012
23. Sketris IS, Langille Ingram EM, Lummis HL: Strategic opportunities for effective optimal prescribing and medication management. Can J Clin Pharmacol 16:e103-e125, 2009
9. FitzGerald JM, Boulet LP, McIvor RA, Zimmerman S, Chapman KR: Asthma control in Canada remains suboptimal: The Reality of Asthma Control (TRAC) study. Can Respir J 13:253-259, 2006
24. Spurling GK, Mansfield PR, Montgomery BD, Lexchin J, Doust J, Othman N, Vitry AI: Information from pharmaceutical companies and the quality, quantity, and cost of physicians’ prescribing: A systematic review. PLoS Med 7: e1000352, 2010
10. George J, Hannah S 5th, Lang CC: Thiazolidinediones and the influence of media adverse reporting on prescribing attitudes in practice (TZD-IMPACT) study. Cardiovasc Ther 27:83-88, 2009
25. The Chronicle Herald. Doctor’s suspension worries groups. May 3, 2006
11. Haas JS, Kaplan CP, Gerstenberger EP, Kerlikowske K: Changes in the use of postmenopausal hormone replacement therapy after the publication of clinical trial results. Ann Intern Med 140:184-188, 2004 12. Hersch AL, Stefanick ML, Stafford RS: National use of postmenopausal hormone therapy: National trends and response to recent evidence. J Am Med Assoc 291:47-53, 2004 13. Lawton B, Rose S, McLeod D, Dowell A: Changes in the use of hormone replacement therapy after the report from the Women’s Health Initiative: Cross sectional survey of users. BMJ 327:845-846, 2003 14. Maclure M, Dormuth C, Naumann T, McCormack J, Rangno R, Whiteside C, Wright JM: Influences of educational
26. Victor TW, Alvarez NA, Gould E: Opioid prescribing practices in chronic pain management: Guidelines do not sufficiently influence clinical practice. J Pain 10:1051-1057, 2009 27. Whelan E, Asbridge M, Haydt S: Representations of OxyContin in North American newspapers and medical journals. Pain Res Manag 16:252-258, 2011 28. WHO ATC/DDD index 2012. Oslo, Norway, WHO Collaborating Centre for Drug Statistics Methodology. 2011. Available at: http://www.whocc.no/atc_ddd_index/. Accessed December 2, 2012 29. WHO Collaborating Centre for Drug Statistics Methodology: Guidelines for ATC classification and DDD assignment 2012. 15th ed. Oslo, Norway. 2011. Available at: http://www. whocc.no/filearchive/publications/2012_guidelines_with_ front_pa.pdf
1693.e1
Media Effects on OxyContin Prescribing
The Journal of Pain
Supplementary Appendix.
Newspapers Included
in the Sample
CANADIAN NEWSPAPERS
AMERICAN NEWSPAPERS
West The Vancouver Sun Midwest The Winnipeg Free Press Central The Globe and Mail (Toronto) The Montreal Gazette East Coast The Chronicle Herald (Halifax) The Cape Breton Post
West The Denver Post The San Francisco Chronicle The Seattle Times The Houston Chronicle Midwest The Chicago Sun-Times The Cleveland Plain Dealer The Columbus Dispatch The Omaha World Herald The St Louis Post-Dispatch The Minnesota Star Tribune Northeast The Boston Globe The Boston Herald The Buffalo News The New York Daily News The Pittsburgh Post-Gazette The Wall Street Journal Southeast The Atlanta Journal The Tampa Tribune The Washington Post National USA Today