Interactions with the pharmaceutical industry and the practice, knowledge and beliefs of medical oncologists and clinical haematologists: a systematic review

Abstract

Background

No previous review has assessed the extent and effect of industry interactions on medical oncologists and haematologists specifically.

Methods

A systematic review investigated interactions with the pharmaceutical industry and how these might affect the clinical practice, knowledge and beliefs of cancer physicians. MEDLINE, Embase, PsycINFO and Web of Science Core Collection databases were searched from inception to February 2021.

Results

Twenty-nine cross-sectional and two cohort studies met the inclusion criteria. These were classified into three categories of investigation: (1) extent of exposure to industry for cancer physicians as whole (n = 11); (2) financial ties among influential cancer physicians specifically (n = 11) and (3) associations between industry exposure and prescribing (n = 9). Cancer physicians frequently receive payments from or maintain financial ties with industry, at a prevalence of up to 63% in the United States (US) and 70.6% in Japan. Among influential clinicians, 86% of US and 78% of Japanese oncology guidelines authors receive payments. Payments were associated with either a neutral or negative influence on the quality of prescribing practice. Limited evidence suggests oncologists believe education by industry could lead to unconscious bias.

Conclusions

There is substantial evidence of frequent relationships between cancer physicians and the pharmaceutical industry in a range of high-income countries. More research is needed on clinical implications for patients and better management of these relationships.

Registration

PROSPERO identification number CRD42020143353.

Introduction

Almost a fifth of the global medication market will be anticancer drugs by 2024, more than four times the nearest competing therapeutic area [1]. Sales representatives from the pharmaceutical industry routinely approach medical oncologists and haematologists, the prescribers of these medications, who are described together here as ‘cancer physicians’. These interactions intend to affect prescribing practice and to maximise sales, which may have negative consequences for patient care.

Previous reviews have investigated the effect of these interactions on physicians in general. In 2000, Wazana found that physicians’ attitudes towards interactions with industry representatives were mainly positive and that most studies showed an association between exposure to industry interactions and behaviours favouring promoted drugs [2]. Lotfi et al. showed more variable attitudes towards these interactions in low-middle-income countries, albeit based on a limited available body of evidence [3]. Regarding prescribing practice per se, Wazana’s review showed consistent evidence for preferential and more costly prescribing following interaction with the pharmaceutical industry. Several subsequent systematic reviews supported these findings, demonstrating a general association between industry-provided information and payments and higher prescribing costs and frequency and lower prescribing quality [4–7].

To our knowledge, following literature and systematic review register search, no previous review has investigated the extent and effect of pharmaceutical industry interactions on the knowledge, beliefs or clinical practice of cancer physicians specifically. A review by Tibau et al. showed in 2015 that reported rates of financial conflicts of interest for authors of clinical practice guidelines of anticancer drugs had increased over time, suggesting these interactions among practice-influencing clinicians are widespread and may lead to potential bias [8]. We performed a systematic review to investigate the extent of interactions with the pharmaceutical industry their effect on the clinical practice, knowledge and beliefs of cancer physicians.

Methods

Protocol and registration

This review was pre-registered on the International Prospective Register of Systematic Reviews (PROSPERO) with the identification number CRD42020143353, with a limited protocol available online [9]. The full protocol is available on request to the corresponding author and includes additional details about pre-specified methods. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria in designing and reporting this study (see Supplementary Appendix S1) [10].

Eligibility criteria

The target population was defined as practising medical oncologists and haematologists globally, including residents training in these specialties specifically. The investigated intervention was any interaction with the pharmaceutical industry. We aimed to identify any study that assessed either an association between interactions and behaviour or a prevalence of these interactions. This was kept purposefully broad to maximise the number of included studies.

The interactions could be either financial or non-financial, as long as they involved some form of direct contact with the pharmaceutical industry or its sales representatives. The relevant comparator was either a lower level or absence of these interactions. All interventional and observational studies with quantitative results were included.

Where studies investigated the effects of interactions, the primary outcomes were any examples of affected clinical practice (such as prescribing behaviour), knowledge or beliefs following interaction with the pharmaceutical industry. These could be either objectively assessed or self-reported.

The knowledge referred to differences in the level of cancer physicians’ knowledge about specific aspects of patient care associated with different levels of exposure to pharmaceutical industry interactions. Beliefs referred to self-reported attitudes around interactions with the industry, including the perceived benefits or harms of these interactions.

We excluded editorials, perspectives, letters to the editor, case series, case reports and qualitative studies such as interviews, semi-structured interviews and focus group analyses. We also excluded both narrative and systematic reviews. Our included studies were limited to those in English, French or Italian. Studies investigating medical students, interns and pre-vocational resident medical officers were excluded. There were no geographic or time limits, nor any setting (i.e. clinical versus non-clinical) restrictions.

Information sources and search strategy

To obtain relevant articles, we performed a systematic search using the MEDLINE, Embase and PsycINFO databases via the Ovid interface, in addition to the Web of Science Core Collection from their inception to September 2019, with the initial searches carried out on October 9, 2019. An updated search of all databases was performed on February 19, 2021. Additional citations were sought through Google Scholar and via a pre-planned forward citation search of included studies.

The search strategies were designed using a combination of keywords and medical subheading (MeSH) terms, tailored to each database. Each strategy was reviewed by two specialist medical librarians following the Peer Review of Electronic Search Strategies (PRESS) guidelines [11], and are reported in detail in the Supplementary Appendix S2.

The criteria used in the search were purposefully broad, so as to minimise the risk of omitting any relevant articles. For example, the initial searches included all articles with physicians as participants, rather than limiting these to cancer physicians specifically. When studies were identified that were not found in the initial search, we performed an additional search using an initially omitted MeSH term and keyword combination (‘exp “Conflict of Interest”/ and (conflict of interest or conflicts of interest).tw and exp Oncologists/’) in the Ovid-based databases to ensure no further citations were missed. These terms were additionally included in the updated search.

Study selection

A single reviewer (AP) screened all citations during the initial title and abstract screen to identify articles considered potentially suitable for inclusion. For the full-text screen of these, all papers were independently screened by five reviewers, working in pairs.

Prior to the full-text screen, we performed a pilot screen of five articles by all reviewers for calibration. When a disagreement between two reviewers occurred during the formal full-text screen, a third reviewer independently adjudicated the final decision. We calculated Cohen’s Kappa statistic to estimate inter-reviewer reliability for the decision to include a paper.

Data collection process and extracted items

We extracted data from the included studies using the standardised data extraction headings for systematic reviews of aetiology and risk provided by the Joanna Briggs Institute (JBI) Reviewer’s Manual [12]. Data were initially extracted by a single reviewer (AP) and confirmed by a second (BM).

In line with the JBI recommendations, relevant data from each study included baseline details about the study, its methodology and characteristics, dependent variables (outcomes), the data analysis methods used and the study results.

Quality assessment

To assess the quality of individual studies, we used the critical appraisal tools provided by the JBI Reviewer’s Manual. These assessments were performed independently by two reviewers (AP and BM) and investigated the studies as a whole rather than focusing on specific outcomes. Disagreements in the quality assessment were resolved by discussion, with no studies requiring third-reviewer adjudication, although this had been planned if necessary. Authors of any included studies were not involved in the selection, data extraction and quality assessment of these studies [13–16]. An external reviewer (PD—see 'Acknowledgements') was engaged for three studies to minimise bias in the assessments, as all other potential reviewer pairings involved either an author of or close professional relationships with the authors of these studies [14–16].

Summary measures and synthesis

We undertook a descriptive analysis of the included studies, presenting their characteristics, settings and populations. Given the heterogeneous and observational nature of all the identified studies, we were unable to produce any summary statistics of effect. Instead, we categorised the studies by their focus of investigation and discussed the results using a qualitative synthesis approach.

When articles included cancer physicians as a subgroup, we reported results for these participants only. If relevant, we additionally reported on comparisons between cancer physicians and other physician groups.

Publication and sponsorship bias

We prospectively planned to look for both publication bias and sponsorship bias in the included studies as a whole. However, publication bias could not be assessed due to the lack of suitable studies for a meta-analysis required to perform inverted funnel plots of results against sample size. We collected data on funding sources and author conflicts of interest (if reported) for each included study.

Results

Study selection

The search flow is displayed in Fig. 1, including reasons for exclusion at the full-text review stage. Of the 5150 unique articles identified through our searches, 31 reports met our inclusion criteria for the final qualitative analysis. The kappa statistic for the full-text screening was 0.730, considered substantial for inter-reviewer reliability.

Fig. 1.

PRISMA flow diagram of search and screening processes.

Study characteristics

The characteristics of each study are described in Table 1. All identified studies were observational, with no assessments of planned interventions. All the studies were conducted retrospectively, and a majority (n = 29) were cross-sectional studies with analysis carried out over a single time period. The remaining two were retrospective cohort studies.

Table 1.

Characteristics of included studies.

First author	Study design	Country	Timeframe	Participants	Exposure/interaction	Outcomes	Funding/COI declared
Category 1: Studies investigating the prevalence of industry exposure among cancer physicians in general
Behdarvand et al. [13]	Cross-sectional (for subgroup data)	Australia	Oct 2011–Sep 2015	- Attendees of sponsored NOAC-related events (n = 635 events for haematologist-specific events subgroup)	- Industry-sponsored NOAC-related events for haematologists	- Attendance rates of events - Expenditure on events in Australian Dollars.	Funding: none COI: none
Chimonas et al. [23]	Cross-sectional	United States (Vermont)	Jul 2002–Jun 2006	- Top 100 recipients of payments from pharmaceutical industry (n = 3 oncologists and n = 1 haematologist)	- Non-research industry payments	- Value of payments from pharmaceutical industry to subgroups of interest	Funding: public COI: none
DeCensi et al. [18]	Cross-sectional (survey)	Italy	Mar–Apr 2017	- Italian medical oncologists (n = 321) (response rate 13%)	- Broad interactions with industry	- Rates of conflicts of interest with the pharmaceutical industry - Perception of conflicts as an outstanding issue	Funding: public COI: none
Fabbri et al. [14]	Cross-sectional	Australia	Oct 2011– Sep 2015	- All pharmaceutical industry-sponsored events during the exposure period, (n = 22,987 events for oncologists and n = 8200 events for haematologists)	- Industry-sponsored events	-Direct assessment of event number by clinical area of focus and professional status of attendees.	Funding: public COI: public
Inoue et al. [25]	Cross-sectional	United States	2015–2017	- All physicians who received industry payments between 2015 and 2017 (n = 9369 in haematologists/oncologists subgroup)	- General and research industry payments - Ownership interests	- Frequency and value of payments from industry (general and research), and ownership interests.	Funding: public COI: pharma
Lee et al. [19]	Cross-sectional (survey)	Australia	Mar–Jun 2015	- Medical oncologists and medical oncology trainees (n = 157 [n = 120 oncologists and n = 37 trainees]) (response rate 24%)	- Industry-sponsored continuing professional development (CPD)	- Funding sources of CPD - Frequency of attending industry-sponsored CPD -Knowledge of institutional policies around industry-sponsored CPD -Attitudes towards industry-sponsored CPD.	Funding: none COI: none
Marshall et al. [22]	Cross-sectional	United States	2014	- All physicians licensed to practice in the United States in 2014 (n = 883,438 in total, n = 15,494 in medical oncology subgroup).	- General and research industry payments	- Number of physician type receiving payments measured as percentage of total number of physician type - Median values of payments measured in 2014 US dollars.	Funding: public COI: commercial (not pharma).
Ozaki et al. [20]	Cross-sectional	Japan	2016	- All Japan Society of Medical Oncology (JSMO) certified oncology specialists on April 1, 2016 (n = 1080)	- Non-research industry payments	- Proportion and value of payments to oncology specialists.	Funding: indirect pharma COI: pharma
Pokorny et al. [15]	Cross-sectional	Australia	Nov 2018 –Apr 2019	- Cancer physicians who received industry payments for registration fees, travel costs and fees for service during the exposure period (n = 236 medical oncologists and 189 haematologists)	- Non-research industry payments	- Proportion of physicians who received payments. - Value of payments in Australian dollars. - Comparison to other specialties.	Funding: public COI: none
Robertson et al. [21]	Cross-sectional	Australia	Jul–Dec 2007	- Attendees at events sponsored by industry, as disclosed by industry (n = 3377 events total, n = 514 oncology events subgroup)	- Industry-sponsored events	- Proportion of events attended by oncologists - Average cost per head spent on hospitality (for oncology subgroup).	Funding: none COI: public
Tao et al. [24]	Cross-sectional	United States	2014	- Haematologist-oncologists who were active on Twitter in Aug 2016 (n = 634)	- General and research industry payments	- Proportion of participants who received payments - Descriptive statistics of payment values	Funding: Not stated COI: professional (not pharma)
Category 2: Studies investigating financial ties among influential cancer physicians specifically (authors of clinical trials and guidelines)
Cherla et al. [36]	Cross-sectional	United States	Jan 2014–Jun 2016	- Authors of clinical literature during the study period from five pre-specified specialty groups 500 randomly chosen articles (100 articles (n = 737 authors), for haematology subgroup)	- COI with industry, including general and research payments	- Discordance between self-disclosed COI and industry disclosures on Open Payments website during the same period - Median rate of payments to full disclosure and incomplete disclosure groups.	Funding: public COI: none
Jagsi et al. [28]	Cross-sectional	United States (assessed journals), Not defined (study authors)	2006	- Authors of oncology-related studies in eight high-impact medical journals (n = 1534 oncology studies)	- COI with industry, including general and research payments	- Percent of studies with COI - Percent of funding sources for trials - Comparative outcomes of trials by the presence of conflicts of interest.	Funding: not stated COI: none
Haque et al. [35]	Cross-sectional	United States (location of editors)	2013–2018	- Oncology physician editors of 26 medical oncology journals (n = 433)	- Non-research industry payments	- Frequency and extent of payments to US oncologist journal editors - Correlation of extent of payments with journal impact factor.	Funding: not stated COI: none
Harada et al. [37]	Cross-sectional	Japan	2016–2017	- Authors of Japanese haematology clinical practice guidelines of two professional organisations from 2015 to 2018 (n = 74)	- Non-research industry payments (limited to speaking, writing or consulting fees)	- Frequency and extent of payments made to clinicians by members of the Japanese Pharmaceutical Manufacturers Association	Funding: pharma COI: pharma
Lexchin et al. [29]	Cross-sectional	Canada	Oct 2016–Feb 2019	- Clinicians making submissions to pan-Canadian Oncology Drug Review (pCODR) for funding of oncology drugs. (n = 261 submissions, n = 125 individual clinicians).	- FCOI with industry, including general and research payments	- Primary outcome: number of submissions with FCOI. - Secondary outcome: frequency of clinicians agreeing or disagreeing with pCODR recommendation, and how this associates with their FCOI.	Funding: none COI: mixed (public)
Liu et al. [30]	Cross-sectional	United States (assessed journals and editors)	2014	- Editors of 52 high-impact US medical journals (n = 67 editors for oncology subgroup who are eligible for payments (i.e. cancer physicians), consisting of two analysed journals (J Clin Oncol (n = 8) and JNCI (n = 59)))	- General and research industry payments	-Level of general and research payments made to editors during study period (2014 US Dollars) (Open Payments database).	Funding: public COI: none
Mitchell et al. [31]	Cross-sectional	United States	2014	- Physician members of National Comprehensive Cancer Network (NCCN) guideline committees for four guidelines (lung, breast, prostate and colorectal cancer) at the end of 2014 (n = 125 individual authors).	- General and research industry payments	-Proportion of authors with FCOI (Open Payments database) and the average amount of payments received by authors from industry (2014 US Dollars)	Funding: not stated COI: none
Moynihan et al. [16]	Cross-sectional	United States	2017–2019	- Leaders (eg board members) of professional medical associations (exact number not stated, but n = ~37 oncologists)	- General and research industry payments	- Frequency and extent of industry payments among leaders	Funding: public COI: public
Riechelmann et al. [32]	Cross-sectional	United States (assessed journals), Not defined (study authors)	Jan 2005–Jan 2006	- Authors of clinical trials and editorials of anticancer and supportive care drugs published in the J Clin Oncol (n = 332 studies (n = 289 clinical trials and n = 43 editorials). - First authors, senior authors and other authors analysed separately.	- COI with industry, including general and research payments	- Proportion of authors with self-declared COI and the nature of these COI. - Likelihood of COI based on geographic location and funding source of study.	Funding: not stated COI: none
Saito et al. [33]	Cross-sectional	Japan	Jan 2016–Sep 2017	- Authors of six prominent oncology clinical practice guidelines in Japan (gastric, colorectal, hepatocellular, lung, pancreatic and breast cancers) (n = 326).	- Non-research industry payments	- Proportion of authors with industry payments Payment values (disclosed by industry members of the Japan Pharmaceutical Manufacturers Association (JPMA)) in US dollars (converted from Japanese Yen).	Funding: indirect pharma COI: pharma
Wayant et al. [34]	Cross-sectional	United States	Jan 2016–Aug 2017	- Oncologist authors of clinical trials of cancer drugs receiving FDA approval. (n = 344)	- General, research and associated research industry payments	- Descriptive statistics of payments in US dollars	Funding: not stated COI: none
Category 3: Studies investigating associations between industry exposure and prescribing
Bandari et al. [27]	Cross-sectional	United States	2012 (prescription claims) and 2013 (period of Open Payments disclosures)	- Prescribers of degarelix and denosumab (n = 7 for degarelix and n = 1336 for denosumab in the oncologists subgroup).	- Non-specific payments received by prescribers from industry (Open Payments database)	- Total Medicare reimbursement from specified prescription codes, as a surrogate measure of prescribing information.	Funding: none COI: none
Bandari et al. [38]	Cross-sectional	United States	2013 (prescription claims) and 2014 (period of Open Payments disclosures)	- Prescribers of abiraterone and enzalutamide (n = 1715 for abiraterone and n = 680 for enzalutamide in the oncologists subgroup)	- Non-specific payments received by prescribers from industry (Open Payments database)	- Prescription counts of enzalutamide and abiraterone, as recorded in Medicare Part D public use file.	Funding: none COI: none
Eisenberg et al. [40]	Cross-sectional	United States	2013–2016	- Prescribers of opioids at 85 US academic medical centres (n = ~2360 oncologists)	- Introduction of marketing restriction policies around gifts and meals, speaking or consulting, visits by sales representatives, and disclosure requirement	- Percentage difference in days prescribing opioids between first year of policy introduction and later years	Funding: not stated COI: not stated
Hadland et al. [42]	Retrospective cohort. Research letter	United States	2014 (year of payments), 2015 (year of opioid prescribing)	- All opioid prescribers in the US with Medicare Part D data (n = 8053 for haematology/oncology subgroup).	- Absolute number of non-research opioid-related payments received from industry per individual (≥1 payment compared to no payment) (Open Payments database)	- Mean number of opioid prescriptions for individuals in each group (i.e. no payments vs ≥1 payment).	Funding: mixed (public) COI: none
Hollander et al. [44]	Cross-sectional	United States	2014–2016	- Physicians who prescribed opioids more than 11 times in 2015 and 2016 (n = 17,323 in haematologists/oncologists subgroup)	- Quartile of value of opioid-related gifts (including meals, travel, accommodation) in a year ($0, >$0-<$20, ≥$20-<$100, and > $100)	- Odds of prescribing by quartile of gifts value, compared to non-recipients.	Funding: not stated COI: none
Mitchell et al. [41]	Cross-sectional	United States	2013–2015 (period of time for payments received), 2015 (period of prescriptions).	- Prescribers of “orally administered cancer drugs for four cancers: prostate (abiraterone, enzalutamide), renal cell (axitinib, everolimus, pazopanib, sorafenib, sunitinib), lung (afatinib, erlotinib), and chronic myeloid leukaemia (CML; dasatinib, imatinib, nilotinib).” N = 2766 total for consistent prescribers from 2013 to 2015 (n = 1483 for prostate, n = 674 for renal, n = 966 for lung and n = 367 for CML).	- Non-research industry payments (specific to manufacturers of the studied drugs)	- Relative prescribing of drugs made by companies making payments to the prescriber compared to other drugs for that cancer.	Funding: not stated COI: pharma (spousal)
Mitchell et al. [43]	Cross-sectional Research letter	United States	2013 (year of payments), 2014 (year of prescriptions)	- Prescribers of on-patent drugs used to treat metastatic renal cell cancer (mRCC) (sorafenib, sunitnib, pazopanib) and chronic myeloid leukaemia (CML) (dasatinib, imatinib, nilotinib), listed as provider type ‘oncologist’ (n = 354 for mRCC group and n = 2,225 for CML group).	- General and research payments from industry received during 2013 from specific drugs’ manufacturers (dichotomous, yes or no) (Open Payments database)	- Prescriptions made in 2014 for drugs of interest (Medicare Part D).	Funding: mixed (public) COI: none
Perlis et al. [26]	Cross-sectional	United States	2013 (full year for prescriptions, Jul–Dec for industry payments)	- All prescribers in the US with Medicare Part D data in 2013 (n = 10,985 for haematology/oncology subgroup).	- Non-research industry payments per individual during the period July–December 2013.	- Cost of prescriptions per beneficiary.	Funding: none COI: commercial (not pharma)
Zezza et al. [39]	Retrospective cohort	United States	2013–2015 (payments and prescriptions, depending on cohort group)	- Prescribers of opioids during exposure period. - Cohort 1: Payments in 2014 and 2015, but not 2013, with matched comparison prescribers who did not receive payments in any year (n = 310 (received payments) and n = 5299 (no payments) for oncology subgroup). - Cohort 2: Payments in 2015, but not 2013 and 2014, with matched comparison prescribers who did not receive payments in any year (n = 461 (received payments) and n = 2081 (no payments) for oncology subgroup).	- Any opioid-related payment from industry in any given year. Categorised as dichotomous (yes or no), as well as total amount in four subgroups (less than 33rd centile, 33rd–66th centile, 66th to 95th centile, greater than 95th centile) for Cohort 1 (Open Payments database)	- Opioid prescribing described in three ways: 1. Mean opioid expenditures (prescriptions filled) during study period, in 2015 US dollars (Medicare Part D). 2. Number of daily doses filled 3. Mean daily dose expenditure in 2015 US dollars.	Funding: public COI: none

NOAC novel oral anticoagulants, COI conflicts of interest, FCOI financial conflicts of interest, pharma pharmaceutical industry, J Clin Oncol Journal of Clinical Oncology, JNCI Journal of the National Cancer Institute.

Among the 31 study reports, we identified three broad categories of analysis: (1) exposure assessments, or investigations of the prevalence of exposure to the pharmaceutical industry for cancer physicians in general, predominantly through receipt of payments or attendance at events, as well as attitudes and beliefs around such exposure (n = 11); (2) financial ties among influential physicians specifically (trial and guideline authors), or investigations of potential bias in decision-making, predominantly through the conduct of clinical trials and clinical guidelines (n = 11) and (3) prescribing outcome studies, or investigations of associations between industry exposure and prescribing (n = 9).

Quality assessment of studies

The quality assessments for each study are presented in Supplementary Appendix S3 and S4, showing results across each critical appraisal domain using the JBI Reviewer’s Manual using McGuinness’s robvis program [17]. The most frequent areas of concern for the quality appraisal were the non-identification of and control for confounders. The studies that conducted surveys were also limited by low response rates and use of non-validated survey instruments. Concerns around the interpretation of specific studies are discussed in detail below.

Results of individual studies

The baseline characteristics of each study is described in Table 1. Summative descriptions of studies within each identified category are described below. Most studies (21 out of 31 [68%]) were based in the US, followed by Australia (5), Japan (3), Italy (1) and Canada (1). All studies were published from 2007 onwards, with a majority (n = 27) published from 2016 onwards.

Category 1: Investigations of exposure to the industry among cancer physicians in general

As shown in Table 2, 11 studies directly analysed the frequency and types of exposure to the pharmaceutical industry, five with medical oncologist participants [18–22], one with haematologist participants [13] and five with both [14, 15, 23–25]. Three studies assessed industry payments made to all clinicians within the specialist subgroup [20, 22, 23]. These were widespread, with Marshall et al. and Ozaki et al. showing that 63% of US and 70.6% of Japanese medical oncologists received general payments in 2014 and 2016, respectively [20, 22]. In Australia, over a 6-month period between 2018 and 2019, 32% of medical oncologists and 31% of haematologists received non-research payments [15]. Among US clinicians active on Twitter, 72.4% received general payments in 2014 [24]. Importantly, Inoue et al. showed that between 2015 and 2017, 80% of all payments made to US haematologists and oncologists were for non-research purposes [25].

Table 2.

Prevalence of industry exposure among cancer physicians in general: key results.

First author	Extent of exposure	Authors’ conclusions	Summary finding
Behdarvand et al. [13]	-23% of Australian NOAC events from 2011 to 2015 were for haematologists (635/2797). - $A10,578,745 spent on all NOAC events during the study period.	Significant sponsoring of NOAC-related events for promotional reasons	Industry payments to cancer physicians and sponsorship of events for cancer physicians are widespread internationally.
Chimonas et al. [23]	Mean payment of US$19,714.67 to oncologists (3/43 total in Vermont) and US$10,121.71 haematologist (1/16 in Vermont) from 2002–2006 (top 100 recipients only)	More effort needed to make industry relationships transparent
Fabbri et al. [14]	-22,987 oncology-related events in Australia from 2011 to 2015 (19.7% of total—the highest of any specialty) - 8200 haematology-related events (7.0% of total—the fifth highest of any specialty)	Widespread industry sponsorship of events for health professionals.
Inoue et al. [25]	-9369 US haematologists or oncologists received industry payments between 2015 and 2017 -80% (7513) were general (non-research) payments -Mean general payment: $5,695 (95% CI 5120–6270) (3rd highest of all internal med spec).	The pharmaceutical industry targets specific specialities to influence prescribing
Ozaki et al. [20]	-763 (70.6%) of Japanese oncologists received at least one payment in 2016 -7325 payments made in total, valued at ¥585,453,314. -Median payment ¥120,016 (IQR 0–¥449,378), median payment count per specialist 2 (IQR 0–7). -Majority value of payments (¥467,802,690, 79.9%) for speaking.	Regulations are needed for managing conflicts of interest among Japanese oncologists.
Robertson et al. [21]	-514 events attended by Australian oncology specialists (15.2% of total) from July to December 2009 -Average cost per head of A$71.53. The second-highest number of events for any specialty (behind psychiatry, 606 events) and the second-highest average cost of any specialty (behind endocrinology, A$71.77 per head).	More transparency needed for industry relationships with Australian health professionals.
Marshall et al. [22]	-9765 US medical oncologists received general payments (63%) and 738 received research payments (4.8%) in 2014. -For general payments: - Median of 11 payments (IQR 3–43) for medical oncologists, compared to 6 (IQR 2–21) for non-oncology specialties (P < 0.001). - Median value of payments $632 (IQR 136–2500), compared to $193 (IQR 57–723) for non-oncology specialties. -Medical oncologists OR for receiving general payments was 1.09–1.75 compared to non-oncologists (overall P < 0.001).	Financial relationships between oncologists and industry are common.
Pokorny et al. [15]	-31.7% of medical oncologists and 30.9% of haematologists received payments (1st and 2nd highest of any specialty group) - Median payments (IQR): - Medical oncologists: AU$2131.26 ($4808.41) (2nd highest of specialty groups) - Haematologists: AU$1519.95 ($2709.37) - Mean payments (95% CI): - Medical oncologists: AU$4353.19 ($3636.17–$5070.21) (highest of specialty groups) - Haematologists: AU$3,807.22 ($3087.13–$4527.31)	The pharmaceutical industry targets Australian cancer physicians, frequently paying for advisory services and subsidising travel.
Tao et al. [24]	- 504 (79.5%) of haematologist-oncologists on Twitter received any payments. - 459 (72.4%) received general payments. - Median $1644.77 (IQR $129.57-$13,744.48) - 307 (48.4%) received research payments. - Median $11,064.21 (IQR 0-$175,164.46)	Haematologist–oncologists on Twitter frequently maintain FCOI with industry.
DeCensi et al. [18]	- 68% of Italian oncologists believe the majority of colleagues have COI with industry - 82% feel most education is sponsored by industry - 75% feel the disproportionate focus on marketing by industry - 60% agree with direct payments for patients on clinical trials (79% feel this should be stated in consent form).	COI is a major issue for Italian oncologists and better policies are needed.	In surveys, oncologists report conflicts of interest with the pharmaceutical industry are frequent and commonly poorly managed.
Lee et al. [19]	- 16% of all CPD for Australian oncologists sponsored by industry. - 13% of oncologists and 27% of trainees report direct industry sponsorship to attend CPD. - 60% of participants report being unaware of any conflicts of interest policy. - 81% received no formal training about conflicts of interest when interacting with industry. - Most participants felt a separation between them and industry was adequate (P < 0.01) and that industry CPD was helpful and education (P < 0.01), while also agreeing that unconscious bias towards a company’s drug could result from their support for CPD (P = 0.05).	There is a lack of understanding or existence of institutional policies for external funding of CPD.

NOAC novel oral anticoagulants, COI conflicts of interest, CPD continuing professional development, CI confidence interval, IQR interquartile range.

Two additional studies reported prevalence findings for payments, though the main findings of these papers related to associations with prescribing. Nonetheless, these showed that cancer physicians in the US receive payments at a higher prevalence than any other specialists [26, 27]. Similar findings were made for payments to Australian cancer physicians [15]. In addition, Behdarvand et al., Fabbri et al. and Robertson et al. all found evidence of oncology- and haematology-related industry-sponsored events in Australia occurring more frequently than or near the highest frequency of any subspecialty group [13, 14, 21].

Attitudes around and prevalence of continuing medical education provided by the industry were assessed by DeCensi et al. in Italy in 2017 and Lee et al. in Australia in 2015 [18, 19]. While limited by low response rates, both these studies showed widespread and poorly managed educational relationships with the industry. Most participants expressed a belief that they had an adequate separation from industry, while concurrently most believed that unconscious bias in favour of a drug could arise from education sponsorship.

Category 2: Investigations of financial ties among influential cancer physicians

As shown in Table 3, 11 studies analysed financial ties with the pharmaceutical industry among influential cancer physicians; nine with medical oncologist participants [16, 28–35], and two with haematologist participants [36, 37].

Table 3.

Financial ties among influential cancer physicians (authors of clinical trials and guidelines): key results.

First author	Pertinent results	Authors’ conclusions	Summary finding
Article authors			Authors and editors of high-impact cancer research frequently have financial conflicts of interest with the industry.
Cherla et al. [36]	-79% discordance rate for among authors of haematology research. -Haematologists in 'incomplete disclosure' category received the highest median payments (US$30,812 (IQR 7122–145,087) of the five tested specialty groups (P < 0.001) in the same category. - Haematologists in 'complete disclosure' category received the highest median payments (US$106,183 (IQR 2000–200,000), but not statistically significantly higher than other groups (P = 0.34) in the same category	There is a high rate of discordance between self-reported and industry-reported conflicts of interest among authors.
Jagsi et al. [28]	-29% of oncology studies had conflicts of interest -17% of oncology studies declared industry funding. -Medical oncologist corresponding authors more likely to have conflicts (45%) than other departments (P < 0.001). -North American studies more likely to have conflicts than other locations (P < 0.001). -Studies with conflicts of interest more likely to show positive OS outcomes (P = 0.04).	There are conflicts of interest within a significant minority of oncology studies.
Wayant et al. [34]	-263 (76.5%) oncologist authors received any payment. -Median value of payments: -General: $2,828 (IQR $0–$19,628) -Associated research (eg grants): $164,644 (IQR $0-$551,926). -110 (32%) did not disclose payments from trial sponsors.	Majority of authors had FCOI, with a high rate of non-disclosure.
Journal editors
Liu et al. [30]	-35/67 investigated journal editors (52.2%) received payments. - For J Clin Oncol: - Mean general payment was $5,957 (SD $9474), median $228 (IQR 0–$8752) - Mean research payment was $160,304 (SD $307,252), median $31,999 (IQR 0–$144,198). - For Journal of the National Cancer Institute: - Mean general payment was $5154 (SD $13,576), median 0 (IQR 0–$519) - Mean research payment was $76,223 (SD $323,500), median 0 (IQR 0–0). - Cumulative mean payments for the 67 editors showed mean payment $5249 (SD $13,098), median $11 (IQR 0–$811) - Highest individual payment $57,282.	Journal editors commonly receive payments from the pharmaceutical industry.
Haque et al. [35]	- 80% of investigated editors had received NRPP - Mean payments $106,778 per editor over 6 years from 2013–2018 - 77% of journals had an editor with NRPP above $100,000 - Total NRPP increased over time from $1,732,240 in 2013 to $7,992,980 in 2018 - Mean NRPP per editor correlated with journal impact factor (Pearson’s r = 0.43, P = 0.02)	Journal editors frequently receive NRPP. Higher impact factors are associated with higher payments.
Riechelmann et al. [32]	- For editorials in J Clin Oncol, 27/59 authors (45.8%) declared COI. 30% declared honoraria and 27% declared consultancy fees.	Authors of oncology editorials commonly have financial ties with the pharmaceutical industry.
Clinical trial investigators
Riechelmann et al. [32]	- For clinical trials investigated in J Clin Oncol, at least one author had conflicts of interest in 198/289 (68.5%) - 150/289 first authors (51.9%) - 150/289 (51.9%) senior authors - 1544/3031 (50.9%) other authors declared COI. - Plurality for each group declared consultancy fees - (93 (32%), 58 (20%) and 107 (4%), respectively. - Authors of clinical trials in North America more likely to have personal conflicts of interest compared to Europe (OR 2.9, P = 0.002) - Authors of industry-funded trials more likely to have COI compared to non-profit trials (OR 13.8, P < 0.001).	Clinical researchers in cancer commonly have conflicts of interest, particularly when research is funded by industry.
Authors of clinical practice guidelines, members of representative societies and advocates to funding bodies
Lexchin et al. [29]	-173 (66.3%) submissions to the pan-Canadian Oncology Drug Review (pCODR) had financial conflicts of interest (FCOI). -119 (45.6%) had FCOI with the submission’s drug manufacturer. -For 'fund' recommendations, the majority of clinicians had no FCOI (18/27). - For 'do not fund' recommendations, the majority of clinicians had a FCOI (10/13), with a plurality having a FCOI with the drug’s manufacturer (6/13), P = 0.027.	There are widespread financial conflicts of interest among clinicians submitting to the pCODR.	Authors of cancer clinical practice guidelines and clinicians advocating for funding of cancer medicines often have financial conflicts of interest with the industry.
Harada et al. [37]	- 94.5% of Japanese haematology guidelines authors received at least one payment for either speaking, consulting or writing - Mean payment (2016 and 2017 combined): $48,040 (SD $41,441) - Median payment (2016 and 2017 combined): $31,553 (IQR $11,456–$75,125).	FCOI among Japanese haematology guidelines authors are widespread and may inappropriately influence decisions
Mitchell et al. [31]	-108 National Comprehensive Cancer Network (NCCN) guidelines authors (86%) had at least one FCOI. - Mean general payments were $10,011 (range 0–$106,859) - Mean research payments were $236,066 (range 0–$2,756,713). - General payments received by 84% of authors - Research payments received by 47%.	NCCN authors had a high value of research payments and a high frequency of general payments.
Moynihan et al. [16]	- 30 (80%) leaders of the American Society of Clinical Oncology (ASCO) received payments. - Medians: - General: $21,138 (10,548–89,399) [3rd highest of all groups analysed] - Research: $510,746 (37,237–1,830,666) [highest of all groups analysed]	There are extensive financial relationships between industry and the leaders of professional associations.
Saito et al. [33]	- 255/326 (78.2%) authors of Japanese cancer clinical practice guidelines received any payment - 3947 payments made totalling $3,444,193, with majority ($2,696,777 (78.3%)) for speaking. - Median payment $3,233 (IQR $506–$10,873) - Mean payment $10,565 (SD $20,059).	Most guideline authors in Japan receive payments from the industry.

COI conflicts of interest, FCOI financial conflicts of interest, NRPP non-research pharmaceutical payments, J Clin Oncol Journal of Clinical Oncology.

Six studies looked at the authors or editors of oncology or haematology trials or journals, to assess ties in these groups [28, 30, 32, 34–36]. These demonstrated financial relationships were reported by between 29 and 80% of clinicians, with one study showing that 79% of haematologists’ financial ties were disclosed incompletely in the published literature [36]. Medical oncologist authors were more likely than any other specialty to have financial ties, and incomplete disclosure of relationships in 32% of cases [28, 34].

Five studies looked at financial ties among the authors of oncology clinical practice guidelines, leaders of representative societies and clinicians advocating for cancer drug funding, again consistently noting that these ties are widespread [16, 29, 31, 33, 37]. In 2014, 84% of National Comprehensive Cancer Network guidelines authors for four common cancers received general payments, while 78.2% and 95%of Japanese oncology and haematology guidelines authors respectively received non-research payments between January 1, 2016 and September 30, 2017 [31, 33, 37]. Lexchin found that 66.3% of submissions to the pan-Canadian Oncology Drug Review had some declared conflict, and 44.5% of all submissions had a financial conflict with the submission’s drug manufacturer between 2016 and 2019 [29]. Among oncologist leaders of the American Society of Clinical Oncology, ~80% received either general or research payments between 2017 and 2019 [16].

Category 3: Investigations of associations between industry exposure and prescribing

As shown in Table 4, nine studies assessed associations between pharmaceutical industry exposure and prescribing. Four of these had medical oncologist participants [27, 38–40], and five had both haematologists and oncologists [26, 41–44]. All were based in the US, and eight used the Open Payments database as an exposure against Medicare prescribing data as a dependent variable.

Table 4.

Associations between industry exposure and prescribing: key results.

First author	Drug(s)	Payments and prescribing	Associationa	Authors’ conclusions	Summary finding
Prostate cancer
Bandari et al. [27]	Denosumab	- Industry payments associated with higher prescription costs ($69,620 vs $60,732, P < 0.001). - Total industry payments per physician correlated with higher prescription costs (ρ = 0.10, P = 0.003).	Positive	Weak association between payments and prescribing.	Payments have a positive or neutral association with prescriptions for prostate cancer
Bandari et al. [38]	Enzalutamide	- total payment per oncologist associated with total number of prescriptions (ρ = 0.31, P < 0.01), - number of payments to oncologists higher for prescribers than non-prescribers (median (IQR) 3 (2–8) vs 2 (1–3), P < 0.01, median total payment (IQR) $59 (25–143) vs $31 (17–65), P < 0.01).	Positive	Positive correlation noted between industry payments and prescriptions of enzalutamide.
	Abiraterone	- total payment per oncologist not associated with number of prescriptions (ρ = 0.06, P = 0.15). - number of payments to oncologists higher for prescribers than non-prescribers (median 3 (IQR 1–5) vs 2 (IQR 1–4), P < 0.01), median total payment amount (IQR) $72 (26–114) vs $56 (22–106), P < 0.01)	Positive (number of payments)	Neutral (total payment amount)	Correlation between number of payments and prescriptions of abiraterone, but not total payment amount
Mitchell et al. [41]	Enzalutamide and abiraterone	- No relationship between general payments and prescriptions for prostate cancer (RR 0.97, 95% CI 0.93–1.02).	Neutral	Null finding for prostate cancer drugs
Renal cell cancer
Mitchell et al. [41]	Axitinib, everolimus, pazopanib, sorafenib and sunitinib	- For US physicians who received general payments in all three investigated years, high rates of prescribing that manufacturer’s drugs in renal cell cancer (RR 1.81, 95% CI 1.58–2.07)	Positive	The strongest association between payments and prescriptions exists for physicians who receive payments consistently.	General payments have a positive or neutral association with prescriptions for renal cell cancer
Mitchell et al. [43]	Sunitinib, pazopanib and sorafenib	Odds of prescribing drug for renal cell cancer from manufacturer providing payment: - OR 1.84 (95% CI 1.25–2.70, P = 0.02) for research payments (n = 32 (9.0%)) - OR 2.05 (95% CI 1.34–3.14, P = 0.001) for general payments (n = 89 (25.1%)) - Combined OR 1.84 (95% CI 1.25–2.70, P = 0.002).	Positive	For one drug assessed, payments associated with increased prescribing
	Sunitinib	General payments associated with more prescribing of sunitinib (50.5% vs 34.4%, P = 0.01),	Positive
	Pazopanib	No association between payments and prescribing	Neutral
	Sorafenib	No association between payments and prescribing	Neutral
Chronic myeloid leukaemia
Mitchell et al. [41]	Dasatinib, imatinib and nilotinib	- For US physicians who received general payments in all three investigated years, high rates of prescribing that manufacturer’s drugs in chronic myeloid leukaemia (RR 1.22, 95% CI 1.08–1.39)	Positive	The strongest association between payments and prescriptions exists for physicians who receive payments consistently.	General payments generally have a positive association with prescriptions for chronic myeloid leukaemia
Mitchell et al. [43]	Dasatinib, imatinib and nilotinib	Odds of prescribing drug for chronic myeloid leukaemia from manufacturer providing payment: - OR 1.16 (95% CI 0.89–1.53, P = 0.27) for research payments (n = 38 (3.8%)) - OR 1.29 (95% CI 1.13–1.47, P < 0.001) for general payments (n = 879 (39.5%)) - Combined OR 1.31 (95% CI 1.14–1.48, P < 0.001).	Positive	For two drugs assessed, general payments associated with increased prescribing. For one drug, a negative association may explained by the introduction and promotion of newer agents at the time.
	Dasatinib	General payments associated with more prescribing of dasatinib (13.8% vs 11.4%, P = 0.02).	Positive
	Nilotinib	General payments associated with more prescribing of nilotinib (15.4% vs 12.5%, P = 0.01).	Positive
	Imatinib	General payments associated with less prescribing of imatinib (72.4% vs 75.5%, P = 0.02)	Negative
Lung cancer
Mitchell et al. [41]	Afatinib and erlotinib	- For US physicians who received general payments in all three investigated years, high rates of prescribing that manufacturer’s drugs in lung cancer (RR 1.69, 95% CI 1.58–1.82).	Positive	The strongest association between payments and prescriptions exists for physicians who receive payments consistently.	General payments have a positive association with prescriptions for lung cancer
Opioids
Eisenberg et al. [40]	All opioids	- For oncologists specifically, the introduction of any of four policy changes results in 1% fewer days prescribing opioids (% difference −0.935 [SE 0.1320])	Positive	Conflict of interest policies can reduce inappropriate opioid prescribing.	Payments generally have a positive association with opioid prescriptions, and changes to institutional policies can improve opioid prescribing practice
Hadland et al. [42]	All opioids	- Of the 8053 US haematologist/oncologists who prescribed opioids in 2015, 832 (10.3%) received ≥1 opioid-related industry payment in 2014, totalling $432,345. - Haematologist/oncologist individuals who received no payments had a mean 108 (SD 127) opioid claims in 2015, compared to a mean of 165 (SD 183) claims for individuals who received ≥1 payment. - Difference in claims attributable to receiving any payment calculated as 3.4% (95% CI 0.6–6.2%).	Positive	Results add to previous studies demonstrating the effect of payments on prescribing.
Hollander et al. [44]	All opioids	- Haematologist-Oncologists who accepted $100 or more in opioid-related gifts were more likely to be in the highest quartile of opioid prescribing the next year than those who received nothing (aOR 1.46 [CI 1.03–2.07]).	Positive	Larger gifts are related to an increased likelihood of prescribing opioids.
Zezza et al. [39]	All opioids	For oncology opioid prescriber subgroups investigated (oncologists receiving payments vs comparison oncologists): Cohort 1: Mean expenditures for opioids (annual): $10,285 (2013), $18,723 (2014/15) vs $9,963 (2013), $10,545 (2014/15). DiD $7,856, SE 3,542, P = 0.0266. Mean daily doses: 2945 (2013), 3245 (2014/15) vs 2732 (2013), 2779 (2014/15). DiD $253, SE 104, P = 0.0149. Mean daily dose expenditure: $3.35 (2013), $5.76 (2014/15) vs $3.55 (2013), $3.71 (2014/15). DiD $2.24, SE 0.77, P = 0.0036. Cohort 2: Mean expenditures: $9355 (2013/14), $11,477 (2015) vs $8,764 (2013/14), $9,608 (2015). DiD $1,278, SE 1053, P = 0.2248. Mean daily doses: 2,582 (2013/14), 2734 (2015) vs 2,476 (2013/14), 2518 (2015). DiD 110, SE 62, P = 0.0749. Mean daily dose expenditure: $3.66 (2013/14), $4.28 (2015), $3.48 (2013/14), $3.68 (2015). DiD $0.42, SE 0.51, P = 0.4169.	Positive	Payments from industry influence opioid prescribing.
All prescriptions
Perlis et al. [26]	All drugs	- Prescribing costs increased linearly with quintiles of payments (P < 0.0001) [range: Q1 = $90.36; SE: 110.35 to Q5 = $1803.23; SE: 11.37] and compared with no payment.	Positive	Payments associated with higher prescribing costs and more frequent branded prescribing.	Payments have a positive association with cancer drug prescriptions

IQR interquartile range, RR relative risk, CI confidence interval, SD standard deviation, OR odds ratio, DiD difference in differences.

^aNote: Positive = payments associated with more prescribing; Negative = payments associated with less prescribing; Neutral = no statistically significant association between payments and prescribing.

Four studies assessed potential associations between industry payments and prescription rates of anticancer drugs. Two studies showed a small or negligible association between payments and prescription rates, although quality assessments raised concerns about both of these due to the identification and control of confounders as well as the validity of the exposure assessment [27, 38]. Prescribing outcomes were measured 1 year prior to exposure assessments in both studies, raising concerns about the validity of the outcome assessment; the results of these studies should therefore be interpreted with caution.

In contrast, Mitchell et al. assessed the effect of payments prior to prescriptions in two studies, with an analytical focus on general payments [41, 43]. In both of these, for almost all cancer subtypes tested there were higher odds of prescribing specific manufacturers’ drugs when oncologists received general payments by that company, if these were received consistently in the years prior. A single negative association for imatinib was potentially explained by contemporaneous introduction and promotion of nilotinib, made by the same manufacturer for the same indication.

Three studies assessed the broad cost of prescriptions following industry payments. Perlis et al found that haematologists and oncologists, combined as one group, had the highest relative non-research payments received of any specialty, with prescription costs increasing in a statistically significant linear fashion across all five quintiles of payments [26]. Hadland et al. and Zezza et al. both investigated the relationship between payments and the costs of opioid prescriptions, with Hadland specifically looking at non-research payments [39, 42]. Using different methodological approaches, both studies showed that cancer physicians who received payments related to opioids had higher overall opioid prescription costs, particularly when payments were consistent over several years. Hollander et al. looked at opioid-related gifts rather than payments per se, and again found higher levels of opioid prescribing among haematologists and oncologists when a greater value of gifts was received [44].

Eisenberg et al. performed the only study assessing the effect of introducing institutional marketing restriction policies on subsequent opioid prescribing [40]. This showed a small (1%) but significant difference in the percentage days of opioid prescribing between the period before and after the introduction of the policies, although it is unclear how these policies were enforced across different centres, meaning these results should be interpreted with caution.

Synthesis of results

We did not perform any quantitative synthesis (i.e. meta-analysis). This was partly due to the majority of studies being observational in nature, without measures of effect, and partly due to the heterogeneous design of the few studies that did measure effect. We were additionally therefore unable to quantitatively estimate differences in the magnitude or direction of outcomes based on study quality.

Discussion

Key findings

This systematic review found strong evidence that cancer physicians frequently receive both general and research payments from the pharmaceutical industry or maintain financial conflicts of interest. When compared to other specialties, studies consistently show that cancer physicians receive payments at the highest or near highest rate of any specialty group. We found further evidence that ‘key opinion leader’ oncologists and haematologists (i.e., those whose positions within authoritative bodies are likely to influence broader practice) receive these payments at especially high amounts, suggesting a risk of bias internationally in the formation of clinical guidelines and high-impact journal publications.

Eight studies assessed prescribing practice of cancer physicians associated with payments from industry, and one looked at valuable gifts rather than payments. All of these found an association with prescribing, with either higher prescribing costs or preference for sponsors’ drugs over others, particularly in the context of general payments. These findings are consistent with a recent review assessing the relationship between payments and physicians across all specialties [7]. All the assessed studies in this category in our review took place in the United States, which is explained by the greater ease of accessing prescribing data in this study population for these drugs than other jurisdictions. No studies directly assessed patient outcomes, and only a single trial assessed the effect of limiting marketing on subsequent prescribing [40].

Notably, all the studies of associations with prescribing practice related to drugs that are either orally or subcutaneously administered, due to limitations in the prescription data available for analysis. They did not assess the prescription of intravenous anticancer medicines, including expensive novel agents such as immune checkpoint inhibitors.

Only two studies asked cancer physicians directly about their knowledge and beliefs around interactions with the industry, in Italy and Australia [18, 19]. The generalisability of both was limited by low response rates to the distributed surveys. Both suggested that oncologists believed that education by industry could lead to an unconscious bias in favour of the companies’ products on the part of prescribers.

Strengths and limitations

This is the first systematic review directly assessing relationships between the pharmaceutical industry and cancer physicians specifically. It was strengthened by our clear methodological approach in line with the Joanna Briggs Institute Reviewer’s Manual. Our search strategy underwent review by academic librarians at two institutions to enhance its validity, following PRESS guidelines, and our reporting followed PRISMA guidelines.

However, the review had two major limitations. First, our initial title and the abstract screen were performed by a single reviewer, which may have led to the inadvertent omission of relevant texts. Second, the review was limited by its specificity. By focusing directly on cancer physicians, studies were excluded in which cancer physicians were assessed but not reported as a specified subgroup. This, therefore, limited the breadth of results that could be included in the analysis.

It is additionally possible that some studies may have been missed due to the specificity of our search strategy, given that several studies were identified through the in-citation review, though the risk of this was minimised through our additional post-hoc search described in the Methods. All studies identified also occurred in high-income countries, limiting the applicability of the results to low- and middle-income countries.

How results relate to other data

The findings of this study are consistent with previous systematic reviews assessing relationships between the industry and physicians in general [2–6]. However, this review has demonstrated that relationships with the industry are more common and more lucrative for cancer physicians than other specialty groups. It has additionally identified that ‘key opinion leader’ cancer physicians are specific targets of influence for the industry.

Meaning of results

This review has shown consistent evidence that cancer physicians are targeted by the pharmaceutical industry, and often more intensively than other specialists, and some evidence that there is a high likelihood their prescribing is influenced as a result. The results also demonstrate that cancer physicians frequently either have little awareness of this or little resolve to alter their behaviour. The mandatory disclosure of payments from industry in several jurisdictions internationally has exposed ethically dubious relationships. While there is some evidence to suggest disclosing payments may lead advisors to avoid these [45], there are no real-world data so far that suggest disclosures have led cancer physicians to reduce their acceptance of payments from the pharmaceutical industry.

There is therefore a need for policy to manage these relationships. At the very least, cancer physicians in influential positions, such as guideline authors and journal editors, should be discouraged or prohibited from accepting general payments from the industry. At least one previous study suggested that US Food and Drug Administration Oncology Drug Advisory Committee recommendations are not associated with financial conflicts of interest, although its interpretation is limited by an unclear number of clinicians on the Committee [46].

Implications and future research

This is an area of ongoing research and investigation. No studies assessed the effect of industry interactions in a controlled, randomised manner, and only a single study looked at behaviour change following alteration of institutional policies [40]. While not impossible, performing a randomised trial would be practically very difficult, as the research question is one of unconscious behaviour in the standard practice of independent practitioners. Notably, controlled trials have been used in other specialties to assess the role of educational interventions on subsequent behaviour, for example in psychiatry residents [47].

A reasonable alternative would be to perform further trials of mandated decreased interaction with industry, such as the Eisenberg study, with a focus on lucrative anticancer drugs rather than opioids. This review also demonstrated a clear paucity of quantitative research exploring the knowledge and beliefs of cancer physicians. If issues with recruitment could be overcome, studies could be performed to understand why cancer physicians as a group interact with the industry to such an extent. This would be valuable to help formulate management policies globally.

Conclusions

The power of cancer physicians to prescribe anticancer medicines is more lucrative to the pharmaceutical industry than any other specialty group. It is therefore imperative to understand how the industry attempts to influence these physicians so that later research can focus on strategies to avoid or, at minimum, manage these interactions to the benefit of patient care.

In this review, consistent evidence was found internationally that cancer physicians maintain financial conflicts of interest with the pharmaceutical industry, particularly when in positions that are likely to influence wider practice. Additional evidence was found that these interactions are likely to affect prescribing practice in a negative way. There is limited evidence that cancer physicians acknowledge and understand that interactions with industry may lead to bias, but no studies assessed or discovered any intent to change the current level of interactions that occur. More studies are needed to investigate how these interactions affect practice.

Author contributions

All authors contributed to the protocol development, selection of studies, interpretation of data and final manuscript. AP undertook the literature searches and extracted data. BM confirmed the extracted data. AP, BM and AF performed the quality appraisals.

Funding

AP was supported by a University of Sydney Postgraduate Award.

Data availability

No additional data is available.

Competing interests

In 2020, Barbara Mintzes acted as an expert witness for Health Canada in a legal case related to the marketing of an unregistered product in Canada. There are no other conflicts to declare.

Ethics approval and consent to participate

No ethics approval was necessary as all data analysed exist in the public domain.

Consent to publish

Not applicable.

Supplementary information

The online version contains supplementary material available at 10.1038/s41416-021-01552-1.