Overview

Why was this needed?
  • Open access (OA) to research published by pharmaceutical companies can improve transparency and foster trust.¹
  • However, there is no publicly available automated tool to assess OA publication rates across pharmaceutical companies, or those in other sectors.
What did we do?
  • We updated a previously described methodology² by introducing a new approach to identifying the type of publication (e.g. journal article, review, letter).
  • The new approach aimed to obtain a cleaner data set by correctly identifying publication types and excluding certain formats.
    • Conference abstracts and letters were excluded, because these are typically less likely to be OA and therefore outside the scope of most OA policies.
  • OA rates from 2019 and 2020 for published articles that were supported by the top 20 pharmaceutical companies were analysed with the new method (see 'Methods' section for more details).³
    • Any Open Pharma Member/Supporter companies outside the top 20 were also included.
What did we find?
  • As expected, fewer articles were identified with the updated method (5093) than the original method (6900) for 2019 (Figure 1).
  • Overall OA rates in 2019 determined using the updated method were higher (76%) than the original method (69%).
  • This suggests that abstracts and letters were successfully excluded.
  • OA rates for individual companies determined using the new methodology are shown in Figure 2.
Figure 1: Publications from 2019 identified as OA articles using the original and updated methods. Footnote: OA, open access.
Figure 2: The proportion of OA pharma-supported articles distinguished by the updated methodology in 2019 and 2020.
Bubbles represent total number of articles associated with the company. Grey bubbles represent 2019 OA rates. Coloured bubbles represent 2020 rates, with Open Pharma Members and Supporters marked in red (other companies analysed in blue).
AZ, AstraZeneca; BMS, Bristol Myers Squibb; GSK, GlaxoSmithKline; J&J, Johnson, Johnson and Janssen; OA, open access.
Table 1: The proportion of pharma-supported OA articles in 2019 and 2020 identified using the updated methodology.
2019 2020
Overalla, % 76 77
Minimum, % 67 56
Maximum, % 95 90
ᵃMean

Methods

Identify pharmaceutical-company-associated publications in Lens.org⁵
With tag ‘medicine’
Export DOIs
Run DOIs through Embase⁶, to select only articles
Exclude all other publication types
Remove companies with < 10 articles
Get OA status via Unpaywall⁷
By running DOIs of articles through search engine
DOI, digital object identifier; OA, open access

Publications analysed

All % values represent proportion of total publications identified in the initial search.
Ineligible articles are conference papers, letters, reviews, editorials, errata, notes (as categorized by Embase).
OA, open access.

Ipsen case study

  • Since 2019, Ipsen has had an OA mandate for all affiliated scientific publications.⁸
  • An internal audit has shown that Ipsen met its commitment to publish all research OA in 2019.⁹
  • However, this automated approach suggested an OA rate of 95% in 2019 (n = 37) and 89% in 2020 (n = 44) (Figure 3).
  • Article-level analysis identified the reasons why seven publications were incorrectly labelled as not being OA.  
    Incorrect article classification by Embase: publication incorrectly assigned as a journal article, so should have been excluded from the analysis
    Incorrect OA status by Unpaywall: articles OA but incorrectly categorized by Unpaywall

    Feedback given to Unpaywall as a result of this analysis has helped improve their algorithms
    Incorrect author affiliation: articles listed with Ipsen affiliation when, in fact, the research was conducted at a previous institution; so therefore should have been excluded
This case study confirms the internal audit finding of 100% OA rate for Ipsen in 2019 and 2020.
Figure 3: Ipsen-affiliated articles (2020) identified by the updated methodology - reasons for articles being incorrectly identified as not OA.
OA, open access.

Strengths and limitations

Strengths
  • Full article-level analysis for Ipsen revealed the methodology to be broadly accurate, and identified reasons for misclassification.
  • Uses publicly available data, meaning that it is objective.
  • Reproducible and automatable.
General limitations
  • Automated classifications result in some errors.
    • AI (artificial intelligence) subject tagging of articles as ‘medicine’ is not always correct.
    • Lens.org sometimes fails to identify author affiliations correctly.
    • Unpaywall occasionally misclassifies articles.
    • Embase also makes occasional errors in article type classification – published congress abstracts can be classified as journal articles.

Conclusions

Future directions
  • This analysis could be used to monitor funder compliance with OA policies.
    • As OA policies become compulsory, this analysis will give an objective measure of compliance for funders and could be used to develop a platform that will allow companies to benchmark the OA rates of their publication portfolios internally.
  • Full automation could allow the easy assessment of changes in OA rates over time.
  • Can be easily extended to analyse OA type (green, gold, etc.) of articles, or copyright licence type (CC BY, CC BY-ND, etc.).
References
  1. Ross JS et al. Am J Public Health 2012;102:72–80.
  2. Freeman H et al. Curr Med Res Opin2021;37(Suppl1):21–39.
  3. Sagonowsky E. The top 20 pharma companies by 2020 revenue. 2021. Available from:  https://www.fiercepharma.com/special-report/top-20-pharma-companies-by-2020-revenue (Accessed 23 December 2021).  
  4. Open Pharma. About Open Pharma. Available from: https://openpharma.blog/about-us/ (Accessed 23 December 2021).
  5. Lens.org. Available from: https://www.lens.org/ (Accessed 23 December 2021).
  6. Ovid – Embase. Available from: https://ovidsp.dc2.ovid.com/ (Accessed 23 December 2021).
  7. Unpaywall. Simple query tool. Available from: https://unpaywall.org/products/simple-query-tool (Accessed 23 December 2021).
  8. Ipsen Pharma. Ipsen commits to making all its published scientific research freely accessible to everyone. Available from: https://www.ipsen.com/ipsen-commits-to-making-all-its-published-scientific-research-freely-accessible-to-everyone/ (Accessed 13 January 2022).
  9. Page S et al. Open sesame! Evaluation of an open access commitment on manuscripts affiliated to a mid-sized pharmaceutical  company. 2020 European Meeting of the International Society for Medical Publication Professionals (ISMPP); 21–22 January 2020, London, UK.
Funding
This work was produced by representatives of Open Pharma. Open Pharma is a multi-sponsor collaboration facilitated by Oxford PharmaGenesis Ltd and receives sponsorship funding from Alexion Pharmaceuticals, Inc., AstraZeneca Pharmaceuticals LP, Boehringer Ingelheim International GmbH, F. Hoffmann-La Roche AG, Galápagos NV, Gilead Sciences, Inc., GlaxoSmithKline Biologicals SA, Ipsen Biopharm Ltd, Janssen Global Services LLC, John Wiley & Sons Ltd, Novartis Pharma AG, Novo Nordisk A/S, Pfizer Inc., Takeda Development Center Americas, Inc., and UCB Biopharma SRL.
Disclosures
EB (https://orcid.org/0000-0001-8982-0627), TK (https://orcid.org/0000-0001-6152-7365) and TR (https://orcid.org/0000-0003-0221-0098) are employees of Oxford PharmaGenesis. SB (https://orcid.org/0000-0001-8072-5690) is an employee of and may hold stock in Galápagos NV. LM (https://orcid.org/0000-0003-4555-8123) is an employee and stockholder of Alexion, AstraZeneca Rare Disease. WG (https://orcid.org/0000-0002-9733-6330) is a former employee of Ipsen, now employed by Bristol Myers Squibb. VP (https://orcid.org/0000-0002-7444-6027) is an employee and stockholder of Takeda Pharmaceuticals.
Acknowledgements
We thank the Members and Supporters of Open Pharma for their input and insightful discussions surrounding our analysis.
Presented at the 2022 Annual Meeting of the International Society for Medical Publication Professionals (ISMPP), 9-11 May 2022, Washington DC, USA