Capstone- Week 2 Discussion: Share your articles

Received: 8 January 2022 | Revised: 16 March 2022 | Accepted: 17 March 2022

DOI: 10.1111/jep.13680

OR I G I NA L PA P E R

Shifting interpretations in evidence and guidance in pain and opioids research: A bibliometric analysis of a highly cited case series from 1986

Sina Tafreshi BSc1 | Adam Steiner BA2 | Abhimanyu Sud MD3

1School of Medicine, University of St.

Andrews, St. Andrews, UK

2Faculty of Arts, McGill University, Montreal,

Quebec, Canada

3Department of Family and Community

Medicine, Temerty Faculty of Medicine,

University of Toronto, Toronto, Ontario,

Canada

Correspondence

Abhimanyu Sud, MD, 500 University Avenue,

5th Floor, Toronto, ON M5G 1V7, Canada.

Email: [email protected]

Abstract

Rationale, Aims and Objectives: Portenoy and Foley's 1986 landmark case series

‘Chronic use of opioid analgesics in non‐malignant pain: report of 38 cases’ has been

reproached for opening the floodgates of opioid prescribing for chronic non‐cancer

pain and the attendant harms. This influential article has been cited over 500 times

in the scientific literature over the last four decades. This study seeks to understand

the impact of Portenoy and Foley's article on subsequent discussions and research

about opioids.

Methods: We conducted a multi‐method bibliometric analysis of all citations of this

article from 1986 through 2019 using quantitative relational and qualitative content

analysis to determine how uses and interpretations of this case series and associated

prescribing guidance have changed over time, in relationship to the evolution of the

North American opioid crises.

Results: Using time series analysis, we identified three periods with distinct

interpretations and uses of the index article. In the first ‘exploration’ period

(1986–1996), the index article was well‐received by the scientific community and

motivated further study of the effects of opioids. In the second ‘implementation’

period (1997–2003, coinciding with the release of OxyContin®), this case series was

used as evidence to support widespread prescribing of opioid analgesics, even while

it was recognized that long‐term effects had not yet been evaluated. The third

‘reassessment’ period (2004–2019) focused on how opioid‐related harms had been

overlooked, and in many cases, these harms were directly attributed to this article.

Conclusion: These changes in interpretation demonstrate shifting currents of the use

and mobilization of evidence regarding pain and opioids, and how these currents

both impact and are impacted by clinical practices and major sociohistorical

phenomena such as the opioid crisis. Researchers and clinicians must account for

these shifting dynamics when developing and interpreting scientific knowledge,

including in the form of clinical practice guidelines.

K E YWORD S

analgesics, bibliometrics, drug prescriptions, opioid epidemic, physicians' (MeSH), practice patterns

J Eval Clin Pract. 2022;28:509–519. wileyonlinelibrary.com/journal/jep © 2022 John Wiley & Sons Ltd. | 509

1 | INTRODUCTION

The over‐prescribing of opioids for chronic non‐cancer pain has

played an important role in the development of the contemporary

opioid crises in high‐income North America.1–3 In 2020, there were

6214 and 81,000 opioid‐related deaths in Canada4 and the United

States,5 respectively. The combined costs of opioid use disorder and

opioid overdoses in the United States were estimated to total USD

1.021 trillion in 2017 alone.6 Historically, physicians had considered

opioids to be effective and important treatments primarily for

managing cancer pain. However, by the mid‐1980s, scientific opinion

started to shift towards considering the use of opioids to manage

chronic non‐cancer pain as represented in highly cited literature

which explored the potential benefits of prescribing opioids for

chronic non‐malignant pain.7,8 However, many have pointed specifi-

cally to Portenoy and Foley's 1986 article, ‘Chronic use of opioid

analgesics in non‐malignant pain: report of 38 cases’, as a seminal

article that opened the door to mass opioid prescribing.2,3,9

Portenoy and Foley's retrospective observational case series

of 38 patients fitting the criteria of having ‘non‐malignant pain

syndromes treated with opioid analgesics for at least 6 months’

reported that only two patients—both with histories of substance

abuse or mental illness—experienced management problems. They,

therefore, concluded that opioids could be ‘an alternative therapy

which may be more humane and provide greater benefit at lesser

risk than other approaches’ for the treatment of chronic non‐

cancer pain. Perhaps what has brought more specific attention to

this case series is the authors' connections to opioid manufactur-

ers. It has been widely discussed, and even argued in legal

proceedings, that the authors' financial support from Purdue

Pharma Ltd. helped to amplify the uptake and influence of this case

series and a particular message that opioids are safe for the

management of chronic non‐malignant pain.10,11

While advocating for opioid prescribing to treat chronic pain, the

1986 article also presented opioid prescribing guidance. This included

the tenet that ‘opioid maintenance therapy should be considered only

after all reasonable attempts at pain management have failed’.12 In

many ways, the specifics of this guidance are similar to recent

national guidelines published nearly four decades later in the United

States13 and Canada.14 Even though the processes for producing

these guidelines differed, they arrived at similar conclusions regard-

ing prescribing opioids as a last resort, conducting risk‐benefit

analyses before prescribing, considering the importance of multi‐

disciplinary care, and limiting doses.

Contemporary guidelines were developed as major clinical and

policy responses to mass harms from opioids, while Portenoy and

Foley's guidance, despite its concordance with contemporary

guidance, has been criticized as having initiated these very harms.

This disconnect makes us reconsider the contemporary framing of

the Portenoy and Foley article and raises important questions about

how this paper has been cited, used and interpreted over time.

Our primary objectives for this study were to explore the

following questions:

(1) What was the nature and magnitude of the impact of Portenoy

and Foley's 1986 article on the subsequent scientific literature?

(2) How were Portenoy and Foley's 1986 findings promoted,

rejected and constructed by the medical and scientific communi-

ties over time? What does this tell us about the evolution of pain

and opioid prescribing research and practice?

This will provide insight not only into historical driving forces of

the contemporary opioid crises, but also into historical and

contemporary currents in clinical epidemiology and clinical medicine,

including where the two are supposed to meet—in the realm of

evidence‐based practice.

2 | METHODS

A bibliometric analysis is well‐suited to exploring our research

questions. We used relational content analysis, supported by time

series analysis and qualitative content analysis, to conduct a

scientometric study15 of all scientific citations of Portenoy and

Foley's 1986 article (the index article). The scientometric portion of

our study aimed to quantitatively identify specific trends presented

by the citing sources, allowing us to address our question regarding

the impact of the 1986 article on subsequent scientific literature. We

conducted a relational content analysis to synthesize quantitative and

qualitative data, as well as to explore complex models in the

development of historical scientific and medical thought.

2.1 | Data sources

We used the Web of Science Core Collection (WoS, by Clarivate

Analytics) to collect all studies that cited the index article. WoS is

commonly used for bibliometric analyses,16 and we chose this database

over alternatives such as Scopus and Google Scholar because of the

availability of detailed citation metadata such as research areas, article

type, funding agencies and countries of origin. We conducted a search

in June 2019 and then an update in July 2020 covering the period from

the publication of the index article through to the end of 2019. The

citation information for every article was collected into EndNote (by

Clarivate Analytics), which was used to retrieve full‐text records. The

remaining full‐text records were retrieved manually. We examined all

the retrieved full‐text articles to ensure that the index article was indeed

cited. Metadata collected fromWoS were collected and managed using

Microsoft Excel (by Microsoft).

2.2 | Outcome measures

2.2.1 | Content analysis

After collecting all of the citing sources, we conducted a relational

content analysis17 that explored the relationship between Portenoy

510 | TAFRESHI ET AL.

and Foley's 1986 article and each of the citing works. While similar

studies have focused on whether citations were affirmative or

negative,18 we expanded the scope of our study to construct a more

detailed analysis of the citation patterns in the scientific literature.

We identified five relevant outcomes: sign (affirmational, nega-

tional or neutral reference), accuracy of index article citation,

attribution of the index citation to driving the opioid crisis, agenda‐

setting and article type (Table 1).

2.3 | Data collection and analysis

Three authors (A.S., S.T. and A. Su.) first conducted a calibration

exercise with 50 randomly selected records. We reviewed our

categorizations and identified any areas of disagreement to refine our

coding process. Then two authors (A.S. and S.T.) independently and in

duplicate assessed and categorized each citation against the index

citation using the definition of each outcome. These categorizations

were reconciled and any conflicts were discussed with a third author

(A. Su.). Metadata were also collected directly from WoS. These

included research area, article type, country of origin and funding

sources for each article.

Proportions between categories were compared using Chi‐

squared or Fisher Exact tests. All tests were two sided and p < 0.05

was considered statistically significant.

2.3.1 | Time‐series analysis

Similar to the procedure of a recent analysis,21 we then conducted

time‐series analyses to explore any time‐based trends in the number

of articles published per year for each coded category.

A preliminary visual inspection of all the time‐series plots identified

outstanding trends for further investigation. We used the Chow test to

TABLE 1 Description and definition of each outcome

Variable Description Outcome Definition

Sign Identified whether the citing article affirmatively, negatively or neutrally referred to the index

article

Affirmation The article confirmed, supported, depended on or was influenced by the index article

Negation The article disputed, questioned or negatively evaluated the index article

Neutral The article took an undefined, non‐promotional, non‐combative stance towards the index article

Accuracy Similar to past studies that categorized textual evidence to examine varying perspectives in

areas of interest19

Accurate The concordance of the presented facts with information from the index article

Inaccurate The citation contained any extraneous, contrasting, assumed or misleading information in regard to the index article

Neither The citation did not present enough information from the index article to evaluate its accuracy

Attribution Assessed the perceived impact by the scientific community of the index article on opioid prescribing practices, and thus the opioid crisis

Attributed The citation implied or explicitly linked the publication of the index article to the subsequent development of the opioid‐related harms

Not attributed The citation did not mention the opioid crisis or opioid‐related harms, and did not establish a link between the index article and the opioid crisis

Agenda setting Whether the index article was fundamental to the motivation or design of the study at hand20

Agenda setting The index article was cited in Section 1 or at the beginning of sections (i.e., at the start of the Section 4) and cited as a motivator for subsequent

research

Not agenda setting

The index article was not cited in Section 1, the start of any other section, and not cited as a motivator

Article type Identified the citing article as either empirical or

non‐empirical research

Empirical articles The article was empirical (quantitative or qualitative) that

included data collection, including secondary analyses such as systematic reviews

Non‐empirical articles

All other articles such as commentaries and editorials that did not report new data or analyses

Note: Description of sign, accuracy, attribution, agenda setting and article type along with a definition of each of their respective outcomes. The definitions were used as guides to code each article in the relational content analysis.

TAFRESHI ET AL. | 511

test for the presence of structural breakpoints in time series data.22

Segmented regression of interrupted time series analysis with AR(2)

residuals was used to examine the sudden and gradual change in the

proportions of publications of interest at given time points.23

2.3.2 | Qualitative analysis

Finally, the content and time‐series analyses were supplemented by a

qualitative analysis. We noted a clear time‐based trend in the publication

of citing articles in the anaesthesia field, which had three publication

peaks that stood out from the other research areas. We purposively

sampled all the articles from this field that were published in each of the

three peaks through the study period. These papers were re‐read and

analyzed with a focus on the larger context of the citation of the index

article, as well as the general tone of the article towards opioids and

opioid prescribing for chronic non‐cancer pain. Once each paper was

analyzed, attitudes towards opioid prescribing were summarized for each

peak year. These summaries were then compared to the quantitative

results to add depth and context to our overall findings.

3 | RESULTS

In total, 523 articles were identified throughWoS through the end of

2019, placing it in the 99.8th percentile of all scientific publications in

the WoS categories of anaesthesiology, neurosciences, clinical

neurology, general and internal medicine from 1986. Five hundred

and eleven (97.7%) full‐text records were included in the analysis

(Figure 1). The remaining 12 were either unretrievable (n = 8) or were

erroneously identified by WoS as citing the index article (n = 4). The

reviewers assessed English, French and German articles in the original

language. Three other articles that were published in Polish,

Portuguese and Spanish were reviewed using Google Translate, an

online translation software (by Google).

3.1 | Time‐series analysis

On visual inspection of the time‐based distribution of each of the

outcomes (File S1), we identified possibly significant time‐based trends

only in the sign and attribution outcomes. Piecewise linear regression of

the sign and attribution time series identified two definitive breakpoints

that were consistent across both outcomes (Figure 2).

For affirmational and neutral citations, a breakpoint was identified

in 1996, after which the rate decreased significantly (Figure 2A). Non‐

attributive citations also had a breakpoint in 1996 (Figure 2B).

The time‐series analysis for negational citations identified a

significant breakpoint in 2004 (Figure 2A). Following that year, the

number of negational citations substantially increased. Similarly, the

time‐series analysis for citations that attribute the opioid crisis to

Portenoy's paper had a distinctive breakpoint in 2004 (Figure 2B).

After 2004, many more papers cited Portenoy's paper as a motivating

factor behind the opioid crisis.

F IGURE 1 Number of citing papers published from 1986 to 2019. A time series illustrating the annual number of citations of the index article between 1986 and 2019

512 | TAFRESHI ET AL.

3.2 | Categorical analysis

Given these two breakpoints, and our two endpoints of 1986 and

2019, we categorized the data into three time‐based groups:

1986–1996, 1997–2003 and 2004–2019 (Table 2).

Overall, the largest proportion of citations was neutral (45.4%).

The majority were accurate (78.9%), non‐attributive (78.3%), non‐

agenda‐setting (59.9%) and non‐empirical articles (57.7%). Moreover,

the majority of articles were published in the United States and

Canada (69.7%). Articles were published in a wide range of research

areas, although papers were concentrated within the areas of

neurosciences and neurology (22.6%) anaesthesiology (16.2%), and

general and internal medicine (15.3%).

3.2.1 | Sign

Categorically, there was a substantial and continuous decrease in

affirmational citations over the three periods (50.4%–38.6%–25.4%).

The decrease in affirmation from the first to the second periods was

driven by an increase in neutral articles while the decrease in

affirmation from the second to the third periods was driven by an

increase in negational articles. Agenda‐setting was significantly

higher in negational (41.6%) and neutral (48.3%) articles than

affirmational (28.7%) articles (File S2).

3.2.2 | Accuracy

There was a consistent decrease in accuracy over the three periods

(88.2%–80.0%–73.8%). There was also a consistent increase in

inaccurate citations over these periods (8.4%–13.6%–20.2%).

3.2.3 | Attribution

There were no attributional articles in the 1986–1996 period and

very few (0.7%) in the 1997–2003 period (Figure 2B). There was a

large spike in attributional articles in the 2004–2019 period (43.7%).

Negational articles (54.5%) had the highest proportion of attribution,

followed by neutral (23.3%), and affirmational (1.1%) articles.

3.2.4 | Agenda‐setting

The proportion of agenda‐setting articles remained relatively

constant throughout the period of analysis (38.7%–36.4%–42.9%).

However, over the three time periods, there was a reversal of the

dominant sign of the agenda‐setting citations. In the first period,

agenda‐setting citations were significantly more affirmational (37.0%)

than negational (6.5%), while in the third period, they were

increasingly negational (34.3%) and less affirmational (17.6%)

(File S2).

3.2.5 | Article type

The proportion of empirical articles peaked in the second period

before decreasing (from 47.1% to 32.1%) in the third period.

Affirmational non‐empirical articles (53.6%) were higher in proportion

than affirmational empirical articles (46.0%) in the first period.

However, this difference was reversed in the third period, where

affirmational empirical articles (29.6%) were higher in proportion than

affirmational non‐empirical articles (23.4%), (Digital content S1).

Empirical articles (52.9%) were significantly more agenda‐setting than

non‐empirical articles (31.5%) throughout all three periods.

(A)

(B)

F IGURE 2 Time‐series for citing papers published from 1986 to 2019. Two piecewise linear regressions with smoothed curves of time‐series for sign (A) and attribution (B) outcomes, demonstrating concordant breakpoints in 1996 and 2004.

TAFRESHI ET AL. | 513

T A B L E

2 C at eg

o ri ca l b ib lio

m et ri c d at a ac ro ss

th e th re e ti m e p er io d s

V ar ia b le

O ut co

m e

T o ta l (%

) N um

b er

(% ) P er io d 1 1 9 8 6 –1

9 9 6

N um

b er

(% ) P er io d 2 1 9 9 7 –2

0 0 3

N um

b er

(% ) P er io d 3 2 0 0 4 –2

0 1 9

p‐ va

lu e

Si gn

A ff ir m at io n

1 7 8 (3 4 .8 )

6 0 (5 0 .4 )

5 4 (3 8 .6 )

6 4 (2 5 .4 )

< 0 .0 0 0 1

N eg

at io n

1 0 1 (1 9 .8 )

6 (5 )

1 0 (7 .1 )

8 5 (3 3 .7 )

N eu

tr al

2 3 2 (4 5 .4 )

5 3 (4 4 .5 )

7 6 (5 4 .3 )

1 0 3 (4 0 .9 )

A cc ur ac y

A cc ur at e

4 0 3 (7 8 .9 )

1 0 5 (8 8 .2 )

1 1 2 (8 0 )

1 8 6 (7 3 .8 )

0 .0 2 4 0

In ac cu

ra te

8 0 (1 5 .7 )

1 0 (8 .4 )

1 9 (1 3 .6 )

5 1 (2 0 .2 )

N ei th er

2 8 (5 .5 )

4 (3 .4 )

9 (6 .4 )

1 5 (6 )

A tt ri b ut io n

A tt ri b ut ed

1 1 1 (2 1 .7 )

0 (0 )

1 (0 .7 )

1 1 0 (4 3 .7 )

< 0 .0 0 0 1

N o t at tr ib ut ed

4 0 0 (7 8 .3 )

1 1 9 (1 0 0 )

1 3 9 (9 9 .3 )

1 4 2 (5 6 .3 )

A ge

nd a se tt in g

A ge

nd a se tt in g

2 0 5 (4 0 .1 )

4 6 (3 8 .7 )

5 1 (3 6 .4 )

1 0 8 (4 2 .9 )

0 .4 3 0 4

N o t ag

en d a se tt in g

3 0 6 (5 9 .9 )

7 3 (6 1 .3 )

8 9 (6 3 .6 )

1 4 4 (5 7 .1 )

A rt ic le

ty p e

E m p ir ic al

re se ar ch

1 9 7 (3 8 .5 )

5 0 (4 2 )

6 6 (4 7 .1 )

8 1 (3 2 .1 )

0 .0 0 9 4

N o n ‐e m p ir ic al

re se ar ch

3 1 4 (6 1 .5 )

6 9 (5 8 )

7 4 (5 2 .9 )

1 7 1 (6 7 .9 )

R es ea

rc h ar ea

N eu

ro sc ie nc

es an

d ne

ur o lo gy

1 7 9 (2 2 .6 )

4 8 (2 7 .1 )

5 5 (2 4 .2 )

7 6 (1 9 .6 )

0 .0 6 9 9

A na

es th es io lo gy

1 2 8 (1 6 .2 )

2 4 (1 3 .6 )

4 4 (1 9 .4 )

6 0 (1 5 .5 )

G en

er al

an d in te rn al

m ed

ic in e

1 2 1 (1 5 .3 )

2 4 (1 3 .6 )

3 6 (1 5 .9 )

6 1 (1 5 .7 )

P ha

rm ac o lo gy

an d p ha

rm ac y

5 6 (7 .1 )

1 8 (1 0 .2 )

1 8 (7 .9 )

2 0 (5 .2 )

P sy ch

ia tr y

3 9 (4 .9 )

4 (2 .3 )

1 3 (5 .7 )

2 2 (5 .7 )

H ea

lt h ca re

sc ie nc

es an

d se rv ic es

3 6 (4 .6 )

9 (5 .1 )

1 0 (4 .4 )

1 7 (4 .4 )

Su b st an

ce ab

us e

2 1 (2 .7 )

3 (1 .7 )

2 (0 .9 )

1 6 (4 .1 )

P sy ch

o lo gy

1 9 (2 .4 )

6 (3 .4 )

4 (1 .8 )

9 (2 .3 )

Su rg er y

1 4 (1 .8 )

1 (0 .6 )

4 (1 .8 )

9 (2 .3 )

O th er

1 7 9 (2 2 .6 )

4 0 (2 2 .6 )

4 1 (1 8 .1 )

9 8 (2 5 .3 )

C o un

tr y o f o ri gi n

U SA

an d C an

ad a

3 7 1 (6 9 .7 )

7 5 (6 2 )

8 6 (5 8 .9 )

2 1 0 (7 9 .3 )

< 0 .0 0 0 1

R es t o f th e w o rl d

1 6 1 (3 0 .3 )

4 6 (3 8 )

6 0 (4 1 .1 )

5 5 (2 0 .8 )

N ot e:

C at eg

o ri ca l b ib lio

m et ri c d at a fo r si gn

,a cc ur ac y,

at tr ib ut io n,

ag en

d a‐ se tt in g,

ar ti cl e ty p e,

re se ar ch

ar ea

an d co

un tr y o f o ri gi n.

O ve

ra ll d at a an

d d at a b y ea

ch o f th re e p er io d s ar e d is p la ye

d .

514 | TAFRESHI ET AL.

3.2.6 | Research area, country of origin and funding source

The proportion of neurosciences and neurology and pharmacology

and pharmacy articles decreased over time. Substance abuse articles

were uncommon in the first two periods before spiking in the third

period. The other research areas did not show a clear trend over the

three time periods. In all three periods, more papers were published

in the United States and Canada than in the rest of the world. The

proportion of papers published in the United States and Canada

stayed relatively constant in the first two periods before spiking in

the third period. Most citing papers (79%) had no disclosed funding

source. The largest funder of research was the National Institute on

Drug Abuse, which funded 35% of citing articles with known funding

sources.

3.3 | Qualitative analysis

Amongst the anaesthesia papers, we identified three distinctive

peaks that aligned with the middle of each of our three identified

time categories (Figure 3). We qualitatively assessed the six,

seven and eight anaesthesia papers that were published during the

peak years of 1991, 2000 and 2010, respectively.

3.3.1 | Sample 1: 1991

Almost all papers published in 1991 acknowledged the controversy

surrounding opioid prescribing and expressed concern at the lack of

definitive scientific evidence about the long‐term impacts of opioid

use.24–29 Most of these articles were empirical trials (66.7%), half of

which were animal‐based. The authors expressed an interest in trying

to better understand pain and pain prevention. These articles

generally cited the index paper's findings as a motivating reason for

their early‐stage research.

Even though there were concerns about the lack of scientific

evidence in 1991 for opioid prescribing, the majority of authors cited

the index article neutrally or affirmatively. This is exemplified by

Brena, who concludes that ‘[at] the present level of information, a

prudent and unbiased course of action is mandatory. We must limit

and resist the use of opioids in non‐malignant pain at the clinical level

until cross‐validated empirical answers… are available’.25

3.3.2 | Sample 2: 2000

Papers published in 200030–36 generally acknowledged concerns

about opioid prescribing, but viewed them as a secondary

concern after treating pain. For example, Quang‐Cantagrel writes

that the controversy about opioids ‘results from confusion about

pathologies which could benefit from opioids and misconceptions

about addiction and tolerance that may result from long‐term

opioid administration’.34 Similarly, Watson acknowledges that

‘the use of opioids for neuropathic pain, generally when all else

has failed, remains contentious’, but immediately goes on to write

that ‘opioids may be the only avenue of relief for many patients

with severe neuropathic pain’.36

Anaesthesiology articles published in 2000 were more support-

ive of opioids for pain treatment (57.1%) than they were in 1991.

There was common agreement that opioids can be safe and effective

for short‐term treatment of non‐cancer pain. Yet, there continued to

F IGURE 3 Citing anaesthesiology papers published from 1986 to 2019. A time series illustrating the annual number of citing anaesthesiology papers published between 1986 and 2019. The ‘sign’ of the citations is shown for the 3 years of interest: 1991, 2000 and 2010

TAFRESHI ET AL. | 515

be uncertainty and disagreement about long‐term effectiveness and

risks of opioid treatment for chronic pain. Covington writes that ‘[w]

hat we do not yet know is what the effects will be of years, in fact

decades, of opioid treatment of intractable pain’.31 Only Mindach

took a strong stance against the index article and opioid prescribing in

his commentary, claiming that ‘[n]ot a single convincing clinical study

could be found to prove the claimed absence of addiction in pain

patients’.33 Of note, this article was published in a German, rather

than an American or Canadian, journal.

3.3.3 | Sample 3: 2010

Anaesthesiology papers published in 201037,38–44 expressed increasing

scepticism about the effects of long‐term opioid prescribing and began

documenting opioid‐related harms. Most papers did not necessarily

report that opioids were ineffective for pain relief. Instead, they pointed

out that other negative consequences have been overlooked. A study of

patients who received opioids for chronic non‐cancer pain concluded

‘that the range of possible harms from [chronic opioid therapy] may be

broader and of a different nature than has been described in treatment

guidelines’.44 This is a widely shared sentiment in most of the articles in

this sample. For example, Jay writes that ‘[t]he treatment of the CNCP

patient with only narcotics is problematic and most often leads to failure.

The most appropriate treatment is within an inter‐disciplinary pain

management programme’.39 This represents a significant change in the

tone of anaesthesiology papers from 2000, when opioid prescribing was

generally advocated as a standalone solution for chronic non‐cancer pain.

4 | DISCUSSION

4.1 | Summary of findings

Portenoy and Foley's article has been cited more than 500 times,

placing it amongst the most commonly cited articles from 1986.

These citations have not only been sustained over a period of four

decades, but the annual number of citations has continued to

increase in recent years demonstrating the continued and profound

impact of this article on pain‐ and opioid‐related scholarship and

practice. Besides these gross measures of impact, we identified three

distinct periods in which the index article was interpreted and used

by the scientific and clinical communities. Building on Geels'

sociotechnical analysis of the societal embedding of psychotropic

medications,45 we can frame these three periods more specifically as

waves of exploration, implementation and reassessment (Table 3).

This framing emphasizes the importance of sociohistorical context in

determining how the index article has been interpreted. An under-

standing of this process is important to both practitioners and

researchers as we attempt to interpret scientific evidence to best

inform prescribing practices and research agendas.

Citing articles from the first period (1986–1996) expressed

concern at the lack of scientific evidence for long‐term opioid

prescribing. The articles published in this period were more accurate

and affirmational than the two subsequent periods. Compared to

later interpretations of the index article, these more accurately

represented the actual conclusions and prescribing guidance pro-

posed by Portenoy and Foley. This was a period of exploration with a

focus on increasing the amount of evidence for opioid prescribing for

chronic non‐cancer pain.

However, this period of exploration concluded at our identified

breakpoint in 1997. Just a year before, OxyContin® was approved in

the United States, a drug that would go on to define the early

American opioid crisis. What had been a period of exploration quickly

transitioned into a period of implementation during which sales of

OxyContin® skyrocketed from $48M USD in 1996 to $1.1B in

2000.46 In this second period (1997–2003), the proportion of

empirical articles reached its highest peak of the three periods. At

first glance, this signifies that concerns about the need for more

scientific evidence for opioids from the first period were being

addressed. However, our qualitative analysis suggests that this

empirical research focused mostly on the efficacy of short‐term

opioid use, rather than providing details about long‐term effects,

either negative or positive. This was coupled with a rise in inaccurate

TABLE 3 Summary of findings by period

Characteristic Period of exploration (1986–1996)

Period of implementation (1997–2003)

Period of reassessment (2004–2019)

Dominant sign towards the index article

Affirmation Neutral Negation

Attribution of the index article to the opioid crisis

Not attributed Not attributed Attributed

Concern towards opioid risks Acknowledges the lack of scientific evidence on the long‐term effects of opioids

Little concern for long term effects of opioids

Concern over the negative consequences of opioids

Motivation of research Better understand the benefits

and risks of opioids

Gather evidence in support of opioid

prescribing

Reassess the negative

consequences of opioids

Note: Description of the main characteristics of each time period identified via quantitative and qualitative analyses, including the dominant sign towards index article, attribution of the index article to the opioid crisis, concern towards opioid risks and motivation of research.

516 | TAFRESHI ET AL.

citations demonstrating that interpretations began to stray further

from the substance of the index article. In the first exploration period,

the index article was being accurately cited as impetus to seek more

evidence for opioid prescribing, but by this second implementation

period, it was increasingly being inaccurately cited as evidence in

support of regular opioid prescribing.

In other words, a push for increased opioid prescribing altered

how scientific evidence was interpreted. In their analysis on the

societal embedding of new technologies, Geels et al. write that

‘innovators, entrepreneurs or health advocates try to attract

attention from sponsors and lobby to create a favourable regulatory

environment. Product champions make optimistic, but diffuse

promises about future performance, societal benefits and profits’.45

In this case, the index article was used to make ‘diffuse promises’, by

both the scientific community and the pharmaceutical industry, that

in many cases were inaccurate. This misinterpretation is most obvious

in how the index article's recommendation of interdisciplinary

treatment was largely overlooked. While Portenoy and Foley

originally noted that opioids should be prescribed as part of an

interdisciplinary treatment plan with the ‘concurrent use of ancillary

cognitive/behavioural and physical therapies’, the promised benefits

of opioids quickly overshadowed this recommendation. According to

Bernard et al., ‘what was intended to be multi‐modal care, however,

became unidimensional care in most settings with reliance almost

completely on pharmacologic solutions to pain’.47 The prospect of

opioids as a one‐step solution for chronic pain prompted govern-

mental and non‐governmental organizations to make opioids more

accessible for chronic non‐cancer pain.47 This second period showed

an increasingly lenient and often inaccurate interpretation that was

associated with increasing opioid‐related harms.

Between 2004 and 2019, opioid‐related harms increasingly

permeated North American society. Moreover, Portenoy and Foley's

article began to be more widely recognized in the media and popular

culture as a catalyst for these rising harms. For example, the 2003

book Pain Killer documents the use of the index article by Purdue

Pharma Ltd. to aggressively advertise the safe use of opioids. More

recently, this has been echoed again in the TV series ‘Dopesick’

wherein Portenoy's connections to Purdue Pharma Ltd. were

portrayed as part of a larger opioid marketing strategy, and the

index article was portrayed as playing a major role in the

development of the crisis. Whereas the previous periods explored

and implemented opioids as a solution for chronic pain, this latter

period has been one of concern and reassessment. The number of

papers that attributed the opioid crisis to the index article grew

exponentially. Negative citations also grew significantly, making this

by far the most negational period. As opioids, and the role of industry

in promoting their use, drew more ire from the scientific community,

so too did the index article. This tracks with Geels' model that the

‘peak of inflated expectations’ is often followed by a trough of

disillusionment in which ‘there are some successes with the new

technology, but more failures’.45 This process is exemplified by yet

another rise in inaccurate citations in the third period. However,

these now grew alongside attributive and negational citations.

Whereas earlier inaccurate citations were affirmational and encour-

aged liberal opioid prescribing practices, these later inaccurate

citations were negational and misinterpreted the index article as

having been a carte blanche for opioid prescribing that incited the

crisis.

4.2 | Implications

The timeline periods identified in this study accord strongly with

other studies which have demonstrated three phases of scientific

communication relating to opioid dose reduction and multi‐

disciplinary care21 and even mass media coverage of the opioid

crisis.48 The former study specifically identified a cross‐correlation

with a 2‐year lag between opioid overdose deaths in the United

States and scientific communication about opioids, clearly identi-

fying a two‐way, push‐pull relationship between clinical science

and the opioid crisis, a major sociohistorical phenomenon. These

studies have collectively identified the 2004 moment as a major

inflection point in perceptions and activity relating to opioid‐

related harms, but at the same time have identified that this

inflection is tied directly to shifting norms and forces building over

the preceding two decades.

Besides communicating outcomes of a series of cases related to

opioid prescribing, Foley and Portenoy's article also included

substantive clinical guidance for opioid prescribing for non‐cancer

pain. By contemporary standards of evidence‐based medicine, this

guidance can certainly be criticized regarding both the quality of the

underlying evidence but also on the opaque and likely idiosyncratic

process through which this guidance was developed. Yet, in the

specifics of its content, this guidance very much accords with

contemporary clinical practice guidelines for opioid prescribing, most

of which were developed as responses to overprescribing and its

attendant harms.13,14 We see here the dual possibilities outlined by

Timmermans and Berg49 of similar clinical guidance being used to

stake out new professional territory and expand medical autonomy in

the period of expansion in the mid‐1980s versus holding physicians

accountable for their practices during our contemporary period of

reassessment (2004–2019). Rather than attending to guidance

content alone or even the processes of developing guidance,50 this

concordance draws our attention back to the importance of

sociohistorical context in determining how clinical guidance is

interpreted and utilized within health systems, and thus influencing

the ultimate impacts of this guidance. As the challenges with the

implementation of contemporary guidelines have emphasized,51,52

understanding and accounting for this contextual effect is a major

challenge for future clinical practice guidance in the complex area of

opioid prescribing. Indeed, a plurality of scientific approaches and

perspectives are needed to provide appropriate, context‐sensitive

guidance for clinicians and policy‐makers.53–55 These are particularly

relevant in this therapeutic area of opioid prescribing where

processes for developing revised clinical practice guidelines are

currently underway.56

TAFRESHI ET AL. | 517

4.3 | Strengths

To our knowledge, this is the first systematic study of the effects of

Portenoy and Foley's seminal 1986 article. By analyzing 511 full‐text

citing articles, this study paints a detailed picture of the substantial and

ongoing scientific impact of the index article. We analyzed all identified

and retrievable articles regardless of country of origin or language of

publication allowing us to establish a global, rather than only North

American, understanding of the index article's impact. Finally, this study

uniquely combined bibliometrics, time‐series analyses, relational content

analysis and qualitative analysis. These multiple analyses demonstrate

strong concordance and thus strengthen our confidence in the findings,

and provide a deeper and broader understanding of the evolution of

opioid‐related research between 1986 and 2019.

4.4 | Limitations

The index article is only one of several seminal articles discussing opioid

treatment for chronic non‐cancer pain that preceded the opioid crisis.

Therefore, our study does not capture the entire picture of interpreta-

tions and reinterpretations of scientific knowledge concurrently with the

dynamics of the opioid crisis. While other bibliometric analyses of

comparable articles have demonstrated similar time‐based trends,18 an

examination of a set of seminal articles could provide a more

comprehensive view, for which this study could provide an initial

conceptual basis. Furthermore, citing articles were only collected from

WoS. While this database is commonly used for bibliometric analyses, a

similar study using other databases such as Scopus or Google Scholar may

have yielded different outcomes and thus alternative conclusions.

The coding of the citing articles would be subject to the biases

and idiosyncratic interpretations of the reviewers. We aimed to

mitigate these effects by using clear coding schema and coding

independently and in duplicate. However, there is still the possibility

of the coding procedure introducing unknown biases to this study.

Finally, our study did not explore the specific reasons behind the

shifts in interpretation. Other approaches to analyzing this corpus,

such as discourse analysis are merited and would be better suited to

further investigate why these changes in interpretations occurred.

5 | CONCLUSION

This systematic analysis of 511 articles citing Portenoy and Foley's

1986 article demonstrates the evolution of its interpretation and

related impact on pain and opioid scholarship. A time‐series analysis

identified three distinct periods of interpretation of the index article

which we labelled as periods of exploration, implementation and

reassessment. These periods of interpretation align well with

inflection points identified by other studies and with major socio-

historical phenomena related to pain management and opioid

prescribing. This illustrates both the changing nature of scientific

interpretation in pain medicine and research, and the importance of

sociohistorical context to this interpretation. Practitioners and

researchers should be attuned to this shifting nature of interpretation

to better develop, critically evaluate and apply scientific knowledge.

ACKNOWLEDGEMENTS

The authors would like to thank Rahim Moineddin for supporting

statistical analysis and Darren Cheng for supporting figure and

manuscript formatting.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the

corresponding author upon reasonable request.

ORCID

Sina Tafreshi http://orcid.org/0000-0003-3222-7018

Adam Steiner http://orcid.org/0000-0002-2457-3132

Abhimanyu Sud http://orcid.org/0000-0001-8354-6153

REFERENCES

1. Makary MA, Overton HN, Wang P. Overprescribing is major contributor to opioid crisis. BMJ. 2017;359:j4792.

2. Kolodny A, Courtwright DT, Hwang CS, et al. The prescription opioid

and heroin crisis: a public health approach to an epidemic of addiction. Annu Rev Public Health. 2015;36:559‐574.

3. Rummans TA, Burton MC, Dawson NL. How good intentions contributed to bad outcomes: the opioid crisis. Mayo Clin Proc.

2018;93(3):344‐350. 4. Opioid‐ and Stimulant‐related Harms in Canada. Accessed August

30, 2021. https://health-infobase.canada.ca/substance-related- harms/opioids-stimulants/

5. Overdose deaths accelerating during COVID‐19. Accessed August 30,

2021. https://www.cdc.gov/media/releases/2020/p1218-overdose- deaths-covid-19.html

6. Luo F, Li M, Florence C. State‐level economic costs of opioid use disorder and fatal opioid overdose—United States, 2017. MMWR

Morb Mortal Wkly Rep. 2021;70(15):541‐546. 7. France RD, Urban BJ, Keefe FJ. Long‐term use of narcotic analgesics

in chronic pain. Soc Sci Med. 1984;19(12):1379‐1382. doi:10.1016/ 0277-9536(84)90027-3

8. Taub A. Opioid analgesics in the treatment of chronic intractable pain of non‐neoplastic origin. Pain. 1984;18:S204.

9. Meldrum M. Opioids' long shadow. AMA J Ethics.2020;22(1): E729‐E734.

10. Meldrum ML. The ongoing opioid prescription epidemic: historical context. Am J Public Health, 106(8):1365‐1366. https://www.ncbi.

nlm.nih.gov/pmc/articles/pmc4940677/ 11. State of Minnesota v. Purdue Pharma L.P. (District Court). https://www.

ag.state.mn.us/opioids/docs/Purdue_FirstAmendedComplaint.pdf 12. Portenoy RK, Foley KM. Chronic use of opioid analgesics in non‐

malignant pain: report of 38 cases. Pain. 1986;25(2):171‐186. 13. Dowell D, Haegerich TM, Chou R. CDC guideline for prescribing

opioids for chronic pain—United States, 2016. MMWR Recomm Rep. 2016;65(1):1‐49.

14. Busse JW, Craigie S, Juurlink DN, et al. Guideline for opioid therapy and chronic noncancer pain. CMAJ. 2017;189(18):E659‐E666.

15. Leydesdorff L, Milojević S. Scientometrics. arXiv [csDL]. http://arxiv. org/abs/1208.4566

518 | TAFRESHI ET AL.

16. Birkle C, Pendlebury DA, Schnell J, Adams J. Web of science as a data source for research on scientific and scholarly activity. Quant Sci Stud.2020;1(1):363‐376.

17. Wilson V. Research methods: content analysis. Evid Based Libr Inf

Pract. 2016;11(1(S)):41. 18. Leung PTM, Macdonald EM, Stanbrook MB, Dhalla IA, Juurlink DN.

A 1980 letter on the risk of opioid addiction. N Engl J Med. 2017; 376(22):2194‐2195.

19. Stahl NA, Brozo WG, Simpson ML. Developing college vocabulary: a

content analysis of instructional materials. Read Res Instr. 1987; 26(3):203‐221. doi:10.1080/19388078709557910

20. Thelwall M. Are classic references cited first? An analysis of citation order within article sections. Scientometrics. 2019;120(2):723‐731.

21. Sud A, Zlahtic E, Cheng D, Moinnedin R, Upshur R. Time series‐ based bibliometric analysis of a systematic review of multidisciplin- ary care for opioid dose reduction: exploring the origins of the North American opioid crisis. Scientometrics, 126(11):8935‐8955.

22. Chow GC. Tests of equality between sets of coefficients in two linear regressions. Econometrica. 1960;28(3):591‐605.

23. Wagner AK, Soumerai SB, Zhang F, Ross‐Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther. 2002;27(4):299‐309.

24. Attal N, Chen YL, Kayser V, Guilbaud G. Behavioural evidence that

systemic morphine may modulate a phasic pain‐related behaviour in a rat model of peripheral mononeuropathy. Pain. 1991;47(1):65‐70.

25. Brena SF, Sanders SH. Opioids in nonmalignant pain: questions in search of answers. Clin J Pain. 1991;7(4):342‐345.

26. Dubner R. A call for more science, not more rhetoric, regarding

opioids and neuropathic pain. Pain. 1991;47(1):1‐2. 27. Jazat F, Guilbaud G. The “tonic” pain‐related behaviour seen in

mononeuropathic rats is modulated by morphine and naloxone. Pain. 1991;44(1):97‐102.

28. Kupers RC, Konings H, Adriaensen H, Gybels JM. Morphine

differentially affects the sensory and affective pain ratings in neurogenic and idiopathic forms of pain. Pain. 1991;47(1):5‐12.

29. Yamamoto T, Yaksh TL. Spinal pharmacology of thermal hyperes- thesia induced by incomplete ligation of sciatic nerve. I. Opioid and nonopioid receptors. Anesthesiology. 1991;75(5):817‐826.

30. Baron R. Neuropathic pain. The long path from mechanisms to mechanism‐based treatment. Anaesthesist. 2000;49(5):373‐386.

31. Covington EC. Opiophobia, opiophilia, opioagnosia. Pain Med. 2000; 1(3):217‐223.

32. Fairbanks CA, Oanh Nguyen H, Grocholski BM, Wilcox GL. Moxonidine, a selective imidazoline–α2‐adrenergic receptor agonist, produces spinal synergistic antihyperalgesia with morphine in nerve‐ injured mice. Anesthesiology. 2000;93(3):765‐773. doi:10.1097/ 00000542-200009000-00026

33. Mindach M. No addiction in pain patients? Questions of a reading physician. Schmerz. 2000;14(3):186‐191.

34. Quang‐Cantagrel ND, Wallace MS, Magnuson SK. Opioid substitu- tion to improve the effectiveness of chronic noncancer pain control: a chart review. Anesth Analg. 2000;90(4):933.

35. Taylor WF, Finkel AG, Robertson KR, et al. Methadone in the treatment of chronic nonmalignant pain: a 2‐year follow‐up. Pain Med. 2000;1(3):254‐259.

36. Watson CP. The treatment of neuropathic pain: antidepressants and opioids. Clin J Pain. 2000;16(suppl 2):S49‐S55.

37. Laxmaiah Manchikanti M, Bert Fellows M, Hary Ailinani M. Therapeutic use, abuse, and nonmedical use of opioids: a ten‐year perspective. Pain Physician. 2010;13:401‐435.

38. Bailey JA, Hurley RW, Gold MS. Crossroads of pain and addiction.

Pain Med. 2010;11(12):1803‐1818. 39. Jay GW. Opioid medications and correct medical usage. In: Jay GW,

ed. Clinician's guide to chronic headache and facial pain. CRC Press; 2010:235‐258.

40. Kidner CL, Gatchel RJ, Mayer TG. MMPI disability profile is associated with degree of opioid use in chronic work‐related musculoskeletal disorders. Clin J Pain. 2010;26(1):9‐15.

41. Koeppe J, Armon C, Lyda K, Nielsen C, Johnson S. Ongoing pain

despite aggressive opioid pain management among persons with HIV. Clin J Pain. 2010;26(3):190‐198.

42. Rhodin A, Stridsberg M, Gordh T. Opioid endocrinopathy: a clinical problem in patients with chronic pain and long‐term oral opioid treatment. Clin J Pain. 2010;26(5):374‐380.

43. Slater ME, De Lima J, Campbell K, Lane L, Collins J. Opioids for the management of severe chronic nonmalignant pain in children: a retrospective 1‐year practice survey in a children's hospital. Pain Med. 2010;11(2):207‐214.

44. Sullivan MD, Von Korff M, Banta‐Green C, Merrill JO, Saunders K.

Problems and concerns of patients receiving chronic opioid therapy for chronic non‐cancer pain. Pain. 2010;149(2):345‐353.

45. Geels FW, Pieters T, Snelders S. Cultural enthusiasm, resistance and the societal embedding of new technologies: psychotropic drugs in the 20th century. Technol Anal Strateg Manag. 2007;19(2):145‐165.

46. Van Zee A. The promotion and marketing of oxycontin: commercial triumph, public health tragedy. Am J Public Health. 2009;99(2): 221‐227.

47. Bernard SA, Chelminski PR, Ives TJ, Ranapurwala SI. Management of

pain in the United States—A brief history and implications for the opioid epidemic. Health Serv Insights. 2018;11:1178632918819440.

48. Webster F, Rice K, Sud A. A critical content analysis of media reporting on opioids: the social construction of an epidemic. Soc Sci

Med. 2020;244:112642.

49. Timmermans S, Berg M, Berg PAM. The gold standard: the challenge

of evidence‐based medicine and standardization in health care. Temple University Press; 2003.

50. Busse JW, Juurlink D, Guyatt GH. Addressing the limitations of the CDC guideline for prescribing opioids for chronic noncancer pain.

CMAJ. 2016;188(17‐18):1210‐1211. 51. Dowell D, Haegerich T, Chou R. No shortcuts to safer opioid

prescribing. N Engl J Med. 2019;380(24):2285‐2287. 52. Kroenke K, Alford DP, Argoff C, et al. Challenges with implementing

the centers for disease control and prevention opioid guideline: a

consensus panel report. Pain Med. 2019;20(4):724‐735. 53. Pawson R, Greenhalgh T, Harvey G, Walshe K. Realist review—a new

method of systematic review designed for complex policy interven- tions. J Health Serv Res Policy. 2005;10(suppl 1):21‐34.

54. Upshur REG. Do clinical guidelines still make sense? No. Ann Fam

Med. 2014;12(3):202‐203. 55. Sud A, Buchman D, Furlan A, Selby P, Spithoff S, Upshur R. Chronic pain

and opioid prescribing: three ways for navigating complexity at the clinical‐population health interface. Am J Public Health. 2022;112:56.

56. Process for updating the opioid prescribing guideline. Accessed December 24, 2021. https://www.cdc.gov/opioids/guideline- update/index.html

SUPPORTING INFORMATION

Additional supporting information can be found online in the

Supporting Information section at the end of this article.

How to cite this article: Tafreshi S, Steiner A, Sud A. Shifting

interpretations in evidence and guidance in pain and opioids

research: a bibliometric analysis of a highly cited case series

from 1986. J Eval Clin Pract. 2022;28:509‐519.

doi:10.1111/jep.13680

TAFRESHI ET AL. | 519

Copyright of Journal of Evaluation in Clinical Practice is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.