key: cord-0052108-epvki6p1
authors: Ingwersen, Peter; Holm, Soeren; Larsen, Birger; Ploug, Thomas
title: Do journals and corporate sponsors back certain views in topics where disagreement prevails?
date: 2020-10-29
journal: Scientometrics
DOI: 10.1007/s11192-020-03743-8
sha: 96db1c2b60c86121116ed96cff8dc6c4c71aa960
doc_id: 52108
cord_uid: epvki6p1

The article focuses on scientific disagreement about the use of statin-related drugs in the prevention of cardiovascular events. The study forms part of an exploration of the broader principle of research polarization, foremost in medicine. The hypothesis is that statin-positive and statin-critical researchers publish in different committed central journals, and that they are financially supported by different dedicated corporate sources. Methodologically we use Web of Science (WoS) analytic tools to perform publication analysis of a time series covering 1998–2018 in three seven-year windows. For each window data is captured based on sets of known statin-positive and statin-critical articles and researchers, and their primary and secondary co-authors. Standard deviation is used as a focused normalization and visual instrument together with Spearman’s correlation coefficient in order to compare frequency distributions of statin-positive and critical journal and sponsor articles. Z-test p-values are used to assess the probability of error concerning the distributions. Findings at general topical level showed that a few journals consistently and significantly occupied top positions, 2 of which, American Journal of Cardiology and Circulation, published articles from both positions. Besides, Journal of the American College of Cardiology served as a major publisher of statin-positive research from 2005, as did European Heart Journal from 2012, replacing American Journal of Cardiology at the top. From 2012 Atherosclerosis and European Journal of Preventive Cardiology served as top-publishers of statin-critical articles. Two central US funding agencies, US Department of Health Human Services and National Institutes of Health (NIH), operated at general topical level across the time series, but the agencies played only a minor role in the divergent research positions. From 2005 statin-positive as well as statin-critical research was mainly sponsored by multinational pharmaceutical companies, predominantly Merck, AstraZeneca and Pfizer. In conclusion, the initial hypothesis about dedicated journals and sponsors was entirely substantiated at the general topical level and at the journal level of research disagreement, but not at sponsor level. Distinct dedicated journals were extracted separately from the 2 divergent statin positions. Since the WoS coverage of sponsor data 1998–2004 was sporadic sponsor data are analyzed from 2005. Only from 2012 the WoS sponsor coverage of the topic is consistently at 60%.

In several fields and specialties, particularly in the health sciences, critical disagreements or even polarization (Ploug and Holm 2015) exist where groups of researchers have opposite views, e.g. on the application of drugs, the use of health monitoring programs or the adverse effects of interventions. Topical examples include the use of statin drugs as primary prevention against cardiovascular events by diminishing cholesterol, HPV (human papilloma virus) prevention by vaccination or breast cancer screening.

The current study forms part of an ongoing scientometric project about polarization in the sciences. The idea is to attempt to establish a methodology which considers a variety of parameters characterizing selected groups of research articles made available by citation indexes, such as Web of Science (WoS) (Clarivate Analytics 2020) or Scopus (Elsevier 2020) . Such parameters are co-authorships; research groups; sponsorships; publishing journals; research institutions; specific title and abstract terms or concepts; specific indexing keywords; and citations (Rousseau and Egghe 2018) . The aim is to provide journal editors and publishers, research groups, researchers, sponsors and research institutions with a tool that may inform about actual authors, research groups, journals or sponsors that are driving certain research views or stand-points. For instance, the discovery of specific actors promoting a certain critical or positive view of a case, eventually over time, is of interest to the community. From a scientometric perspective such a tool consists of (1) an information retrieval mechanism to extract relevant data in a given time period associated with the topic in question from a given database (e.g., WoS) and (2) a range of well-known analytic instruments, such as frequency distributions; a time series; co-term/author/citation analyses; bibliographic coupling; clustering and network analyses.

In this study we explore the statin issue, including the encapsulation of statin-positive vs. statin-critical research by means of co-author analyses and the application of frequency distributions of journals and sponsors in the form of a time series. The assumption behind co-author analysis is that co-authors to a large extent share the same research perspective. The frequency distributions provide ranked lists of actors (journals and sponsors; positive and critical) which can be compared. Our hypothesis is that aside from the many journals and sponsors that support statin research, irrespectively of viewpoint, one may disclose statin-positive research published in some committed high-ranking journals and being sponsored by a few dedicated corporate units different from other high-ranking journals and sponsors dedicated to statin-critical research.

We have chosen to analyze the statin issue covering the period 1998-2018 across three seven-year periods. Simvastatin was originally produced by Merck Company and the patent ran out in 2006. Fundamentally, the research community currently agrees to apply statins (Simvastatin or Atorvastatin and similar drugs) as secondary prevention, that is, in cases where patients already have experienced heart failure and other serious cardiovascular events (Godlee 2014; Demasi 2018) . However, statin drugs used as primary prevention in people that shows no or minor indications of cardiovascular problems, has given rise to much disagreement (Olsson 2009; Redberg and Katz 2017; Akyea 2019) .

For the period 1998-2018 we searched Web of Science (WoS) for scientometric analyses of disagreement or polarization in research work in order to find similar studies. We found that scientometric analyses of divergence phenomena concentrate on publication patterns and gender issues, demographic and topical divergence, management positions and productivity or differences in citation and publishing patterns. One study investigated the spread of research supported by one funding agency (Folbe et al. 2014) . Vinnik et al. (2012) aimed at identifying factors predicting high-quality cardiovascular research. Very few scientometric studies deal with different research views on the same topic, e.g. Nestorowicz and Anacka (2019) on migration literature, and we have discovered no attempts to do such analyses in medical topics.

We have used WoS, Science Citation Index and its analytic tools as the source for collecting data (Clarivate Analytics 2020). Statin-positive and statin-critical authors were initially detected through researcher statements in classic media, such as the BBC, and other science associated sources in the public domain. We applied levels of standard deviation (STDV) as a normalization tool in order visually to compare the frequency distributions of journals and sponsors within each 7 year period, regardless of their size. Non-parametric statistical tests, i.e., Spearman's Rho, were applied to assess the strength of difference between statin-positive and critical distributions.

Following the "Methodology" section, the paper provides general findings on the statin topic, followed by findings associated with the journals and sponsors for each period and comparisons between statin-positive and critical research. The paper ends with a concluding discussion of study biases, limitations and findings.

Initially, we isolated a basic set from WoS on the topic 'simvastatin OR atorvastatin OR statin*', combined with 'cardiovascular' covering 1998-2018, Table 1 . This retrieval profile covers a large variety of cardiovascular issues, such as heart failure and other coronary issues and include adverse phenomena, also in patients with diabetes or renal diseases. The set was then divided into three seven-year sets, 1998-2004; 2005-2011; 2012-2018, to form a time series on the topic. For each period set we applied WoS analytic tools to produce ranked frequency distributions of the publishing journals and supporting institutions (named sponsors) associated with the topic, Tables 2 and 3. We applied 3 independent data sources in order to capture statin-positive and statincritical articles from each analysis window: (1) knowledge of known statin-positive/critical researchers derived from various media; (2) manually monitoring and selecting top-cited articles from each of the 3 period sets in a systematic way; (3) manually checking review articles from within each period set retrieved by search terms like 'controversy', 'debate', 'risk*' 'efficacy', 'benefit*', 'adverse effect*', 'primary prevention'.

1. We consulted Wikipedia (UK) and other classic media associated with scientific issues, such as TV documentaries, newspapers and Scandinavian medical journals not indexed in WoS, in order to find statin-positive and statin-critical researchers and articles. For instance, the well-known and highly cited statin-positive researcher Collins R was extracted from a BBC health program (BBC 2013); later BBC programs have also been dedicated the use of statin treatments. The Swedish medical journal Laekaertidningen provided us with the researcher, Olsson AC (2009) and other statin-critical or uncon-vinced international researchers from article references. This mode of data gathering from non-academic sources is not to be regarded as systematic but helped to get hold on the divergent issues in question. From the selected key researchers, we extracted their articles and actual co-authors, providing us with two sets of statin-positive/critical articles. We name these small sets of authors 'Key-Researcher Sets'. 2. Each of the 3 period sets were sorted by citations and top-cited articles manually checked for statin-positive and critical/uncertain statements found in titles, abstracts or conclusions. Because older articles in each set may obtain more citations than younger articles the lower limit of citations to articles monitored was set to 50 during 1998-2011 and ≥ 20 citations for 2012-2018. An article was regarded statin-positive if it promotes the use of statin treatment in cases of low LDL/HDL values, in primary prevention of cardiovascular events and in cases where the benefits of treatment are emphasized to outdo observed adverse phenomena. Most articles of this kind were cohort or placebocontrolled studies and meta-analyses. Statin-critical articles were items pointing to adverse issues interpreted as more serious for patients than benefits, arguing against the use of statins in primary prevention or casting doubt about or pointing to outcomes of studies where statin treatment did not demonstrate substantial effects. Cohort analyses as well as opinion papers, review articles and letters to the editor/comments and replies considering studies constituted many of such items. Discussions of biases in cohort vs. placebo studies were often interwoven into the argumentation. The selected statinpositive/critical articles and their authors were checked against the already captured two Key Researcher Sets of co-authors retrieved from each topical period. 3. Further, we identified some few additional authors to articles not retrieved previously deemed statin-positive or critical from title words and abstract statements traced in WoS during the three analysis periods defined by the search terms outlined above.

The 2 groups of statin-positive/critical co-author data per analysis period, originating from the 3 modes of data capture, were searched against the 3 original period sets of topical publications. The resulting 2 × 3 set combinations form the first round of co-authors searching. For each set WoS analytics was used to produce 6 new ranked distributions of co-authors. The distributions were sorted alphabetically in order to detect previously found author names. New authors not detected previously were extracted and added to the previous author search profiles by Boolean OR. The lower limit of author frequency on the co-author distributions was set to ≥ 4. As an extra benefit we could carry out name form control and selection of less used name forms for the same authors (e.g., Ridker PM & Ridker P). This second and final round of co-author searching augmented the size of the original statin-positive/critical sets. Each of the 2 × 3 sets was refined by WoS analytics to exclude statin-positive authors from the statin-critical sets and vice versa. A third round of co-authorship searching was avoided for fear of extracting non-relevant authors and because the authors on the lists tended to be the same as on previous lists. The total number of statin-positive articles was twice as large as the number of critical articles for the entire period (817 vs. 393). For each analysis period we used WoS analytic tools to produce frequency distributions of journals and sponsors from the 2 divergent sets of articles, Tables 4, 5, 6 and 7.

Thus, first we compared the journal or sponsor rank distributions in the 3 original sets on the general statin topic, Tables 2 and 3, with the statin-positive and the statin-critical distributions, Tables 4, 5 and 6, 7, respectively. Secondly, we carried out time series analysis as well as comparisons between statin-positive and critical journal and sponsor distributions. Since the data are non-parametric, we applied one-tailed Z-test when comparing the statin-divergent distributions with the original topical sets, but only to demonstrate the calculated degree of effect and probability of error (the z and p values). We do not state anything about the (non)significance of the statin-divergent sets. We are aware of the significance discussions by Schneider (2015) and Amrhein et al. (2019) and recognize that by using our described methodology several biases are introduced in the study. In order to compare statin-positive journal and sponsor distributions we used Spearman's Rho and visual comparisons.

To visually detect if a (group of) journal(s) or sponsor(s) stand out on the general and particular distribution lists observed for the selected periods we applied standard deviation (STDV) statistics, so that data elements (analysis units) located by frequency equal to or above average frequency + 3-5 STDV in distributions are marked across the time series. As such the STDV indicator functions as a focused normalization instrument in the comparisons. WoS analytics allows analysis of max. 100,000 units per set of articles, which in the current study did not pose a problem, since the number of units in all analyses was below that limit. Sponsor name forms were checked for different versions, resulting in altered frequencies for some sponsors. WoS controlled and verified the indexed journal names. Table 1 demonstrates the development of research production 1998-2018 on the statin topic. The productivity was slow and without highly cited articles and few review articles during the initial period 1998-2004 compared to later. The following 7 year period saw a vast increase in productivity-from a total of 1413 to 4809 items and more than tripling from 300 to 1070 review articles. 29 articles became highly cited 2005-2011. As noted above the patent held by Merck for Simvastatin ran out in 2006, making the drug free to pursue and study further by various companies and state agencies from then on. In the third and most recent period the research production increased further with respect to all document types. Many meta-analyses, cohort and large-scale clinical studies also emerged. Tables 2 and 3 demonstrate the distributions of mainstream journals and sponsors used in statin research. We plotted the frequencies and the average plus triple, quadruple and ≥ 5 × STDV values, based on all elements of the distributions. Table 2 As can be observed in Table 3 WoS does not index all sponsoring agencies in the initial analysis period. Only 8% of the records contain sponsors, hence the short list. No commercial companies are indexed. The indexing begins to be more effective from 2007. For sponsors the analyses therefore concentrated on the two 7 year periods, 2005-2011 and 2012-2018, in which the coverage of sponsors is substantially higher, 30%-59%, respectively. Thus, for the period 1998-2011 Table 3 only demonstrates clues as to top supporters.

The 5 same organizations, 2 US health agencies and three pharmaceutical companies including the patent owner Merck Company, stand out as actual top sponsors of research on statin drugs related to cardiovascular events 2005-2018 (≥ av. + 5 × STDV). A huge gap exists between the two US agencies and again after NHLBI (2005 ( -2011 ( ) and AMGEN (2012 ( -2018 . The spread of individual sponsors increases dramatically from the second to the third analysis period, almost tripling-from N = 1576 to N = 4688, Table 3 , mainly owing to the indexing policy of WoS. 3760 sponsors are single article sponsors constituting a very long distribution tail, hence the lower STDV compared to the 2005-2011 period. This phenomenon can be explained by the fact that many articles are sponsored by more than 1 organization. AMGEN constitutes a new player in the research. Large cohort, clinical and meta-analysis studies are published and often highly cited during this recent period. Table 4 shows that Circulation and American Journal of Cardiology acted as top-journals for station-positive publications, 1998-2011, only partly corresponding to the overall toppattern for the same periods, Table 2 . Circulation and Journal of the American College of Cardiology served as top-journals 2005-2018 with European Heart Journal and Atherosclerosis also entering as top-journals during the third period, Table 4 . This pattern at the top was quite different from that shown in Table 2 for the identical periods. Note that PLOS One, listed 3 2012-2018, Table 2 , did not appear on the top-30 list of statin-positive journals during this period, Table 4 . Hence, during the period 2005-2018 the Journal of the American College of Cardiology, and European Heart Journal 2012-2018, were regarded specific to statin-positive research. Owing to their high frequencies it is highly improbable that the three top-journals 2005-2018, located in the 4 × STDV + zone, might be replaced by lower frequency journals, given the fairly low statistical uncertainties for the 2 periods (p = 0.02 and 0.056, Table 4 ). Table 5 Table 5 only serves as clues to or indications of the top-sponsors. The low rank positions of the top mainstream research supporters, Table 3 , i.e., the 2 national US health agencies, and the apparent competition between several commercial players, with Merck Company only ranked 3 during 2012-2018, constituted the most interesting observations. Note also the emergence of Amgen as a statin-positive major sponsor during the most recent period-in line with its sudden appearance, Table 3 . Table 6 demonstrates that American Journal of Cardiology and Atherosclerosis served as the most important journals publishing statin-critical research articles 2005-2018, with European Journal of Preventive Cardiology ranked 3 2012-2018. The 2 latter journals covered 36% of the total statin-critical output during the third analysis period and the statistical uncertainty is quite small: p = 0.015. Thus, it is most likely that the top-journals 2012-2018 do not shift rank positions. Compared to the journals publishing mainstream statin research, Table 2 , European Journal of Preventive Cardiology stands out as the most specific statin-critical journal in the recent period, followed by International Journal of Cardiology.

The 3 pharmaceutical companies, Merck Company, Pfizer and AstraZeneca topped the 2 distributions, Table 7 , as in the case of statin-positive supporting institutions, Table 5 Table 4 vs. Table 6 .

During the most recent period, 2012-2018, the Spearman´s correlation coefficient is even weaker compared to the previous period, Rho = 0.13254. The 2 opposite distributions are very different. 11 top-30 statin-critical journals were not found among the top-30 statin-positive ones and 13 statin-positive journals did not appear on the list of statin-critical journals. During this period Circulation, Journal of The American. College of Cardiology, European Heart Journal and American Heart Journal shift from neutral top-journals 2005-2011 to serve as the dominant statin-positive journals, with very few published statin-critical articles, Table 8 . Lancet assists the statin-positive position, ranked 6. Heart is the highest ranked statin-critical journal (ranked 4) not found on the positive list.

For Sponsors 2012-2018, Tables 5 vs.7, Rho = 0.52403. By normal standards, this association between the 2 variables is considered quite weak and not significant. Compared to the much lower Rho coefficients for journals more sponsor units overlapped the 2 distributions. An additional reason for a higher sponsor Rho was the fact that the same group of identical pharmaceutical companies topped both distributions. However, it is interesting to observe, Table 7 , that the Merck Company ranked far ahead as statin-critical sponsor 2012-2018, compared to its third position as a statin-positive supporter after AstraZeneca and Pfizer, Table 5 . Further, Novo Nordisk (ranked 9, Table 7 ) and Shering Plough Corporation (ranked 11) exclusively supported statin-critical research, 2012-2018, not being ranked among top-30 in the Table 5 distribution for the same period. The highest ranked unique statin-positive sponsors were NIH National Heart Lung Blood Institute, NHLBI (ranked 9, Table 5 ) and Medical Research Council, UK (ranked 12). 7 top-30 statin-critical sponsors were not found among the top-30 statin-positive ones and 8 statin-positive sponsors did not appear on the list of statin-critical journals.

Our methodology contains some possible biases of subjective nature. We applied four kinds of sources in order to collect data in our study.

(1) The main data source in the study consisted of articles extracted by common information retrieval. In our case the retrieval profile contained the terms ((simvastatin OR atorvastatin OR statin(s)) AND cardiovascular). The profile was searched in WoS, that is, Science Citation Index limited to 1998-2018 and to articles only. The reasons for using WoS are twofold: The database coverage is high in the health sciences (Ingwersen and Lynge 2004) and powerful analytic tools are available (Clarivate Analytics 2020). This retrieval mode was objective and exhaustive, covering a variety of associated health problems without these being mentioned directly in the profile, e.g. diabetes, heart failure, renin issues, adverse events, etc. The initial set on the statin topic was then divided into 3 sets covering 1998-2004; 2005-2011; and 2012-2018 . This data capture mode contains no biases.

(2) Finding known protagonists of statin-positive and critical/doubtful research positions, named Key-Researchers. This kind of initial known-person or known-item searching is well-known in the information retrieval discipline as a starting point (Ingwersen and Jarvelin 2005). Typically, such researchers may indeed be retrieved from the classic media, such as TV documentaries or newspapers/magazines, to where scientific debates have penetrated because they are seen to be of public interest. Such areas might concern climate change, energy issues, immigration, etc. Among health issues the pro et contra of the application of face masks or vaccine reliability in relation to Covid-19 constitute actual topics that might be investigated scientometrically in the near future.

In the actual study concerning the topic ´statins used to prevent cardiovascular events' BBC and other news media as well as national medical magazines/journals were used. Searching the 'key-researchers' in WoS provided sets of articles and consequently also the co-authors of Key-Researchers. The co-author names were put aside to be added later in search mode 3). As stated above in the "Methodological" section the assumption behind applying co-authorship analysis is that co-authors commonly share scientific views on a published issue (Bates 1979) .

(3) For each 7 year analysis window WoS analytics made it possible to sort the retrieved article records according to citations. Logically, the most interesting articles to check are the most used (= cited) articles (Rousseau and Egghe 2018) . By going through the highly cited abstract titles and conclusions (≥ 50 citations, 1998-2011; ≥ 20 citations, 2012-2018) it was possible (subjectively) to assess if articles belonged to mainstream, statin-positive or critical positions. We carried out double-checking. Only if the assessments agreed an article was extracted as positive or critical. If in doubt the article belonged to mainstream research. For each article deemed statin-positive or critical all authors were extracted and added to the Key-Researcher co-author sets from search mode 2) into extended author-dependent search profiles. Previously selected authors were monitored for appearance in the highly cited abstracts. The extended author profiles were then combined logically (by Boolean AND) with the three period-dependent topical sets, providing statin-positive and critical author-derived sets of articles for each period. Due to its interpretative nature subjectivity biases exist in this search mode.

To an extent they were controlled by means of inter-assessor agreement but, like for inter-indexer agreement, bias cannot completely be avoided (Tell 1969; Jones 1983 ).

Search mode 3) constituted a first round of co-author retrieval, resulting in 2 × 3 novel and larger sets than retrieved from the Key-researchers and their co-authors alone in mode 2). These sets were again analyzed for additional co-authors by means of WoS analytic tools. Not previously observed authors were added to the search. However, to be certain that statin-critical Key-researchers and direct co-authors did not form part of the statinpositive sets as false-positive, such researcher names were excluded from the final sets. The same treatment was done with respect to the statin-critical sets. The twice repeated co-author searching, and the exclusion, served as instruments for enlargement of the sets in a controlled manner. It was possible that some of the new co-author names might lead to other scientific positions than intended. However, their number would be small. The final sets to be analyzed in terms of journal and sponsor distributions constituted approximately 10% of the total retrieval output associated with statins and cardiovascular events (1210/12,563 articles).

(4) At this point we experimented with searching the three original period sets by means of specific keywords considered to signal positive or critical positions. However, this retrieval mode did only supply a few additional articles published in each period compared to the outcomes from other search modes. Typically, the new articles were opin-approx. 40% of the research on this topic has no external sponsors, but in reality is supported by internal funding. In other topics this percentage might be different. At the general topical analysis level, the findings showed a huge gap between 2 central US state funding agencies and the remaining sponsors, in addition to a gap between 3 commercial companies as top sponsors and other supporting organizations during the 2 recent periods. All 5 sponsors were located above the quadruple STDV zone. From normalization and display points of view the application of STDV was shown to be valuable. The spread of individual journals publishing statin drugs applied to cardiovascular cases increased dramatically, more than tripling during the analysis period 1998-2018. With respect to sponsors during the period 2011-2018 a long tail of frequency-one institutions indicated that many research projects were sponsored by more than 1 supportive agent.

Our hypothesis stated that statin-positive and statin-critical scientists primarily published in and were sponsored by 1 or a few committed major journals and supporting actors. The idea was that such dedicated journals and sponsors back a position and could be observed consistently across 2-3 analysis windows. Commercial or public health reasons were hypothesized to be the motivation for the involved actors.

At general journal level the findings were consistent, showing that Atherosclerosis and Circulation were the most important mainstream journals in the overall topic across all 3 analysis windows. They were consistently located in the triple-quadruple + STDV zones. American Journal of Cardiology acted as top-journal 1998-2011 but dropped to rank 7 during the third period, replaced at the top by PLOS One. The latter journal did not appear in the top STDV zones, regardless of its position on the use of statins. Interestingly, Atherosclerosis appeared outside the top STDV zones in the statin-positive journal distributions and did only appear as top-statin-critical journal during the 2012-2018 period. Journal of American College of Cardiology entered the quadruple + STDV zone among the statinpositive sources but was hardly detected in the statin-critical distributions.

Circulation and American Journal of Cardiology served to a large extent both the statin-critical as well as statin-positive research over the period. The findings indicated that Journal of the American College of Cardiology and European Heart Journal, with American Heart Journal as runner up 2012-2018, can be regarded as specific to statin-positive research 2005-2018. PLOS One, listed 3 2012-2018 in mainstream research did not appear on the top-30 list of statin-positive journals and very low on the statin-critical distribution during this period. The European Journal of Preventive Cardiology and International Journal of Cardiology served as the highest ranked sources 2012-2018 publishing statincritical research. However, no journals were located consistently among the top statin-critical journals across 2 consecutive periods. The assumption about position-dedicated journals consistently observed across 2 or more periods only held for statin-positive journals. In the statin-critical case the assumption was true only for single periods. Notwithstanding, the correlations between the statin-positive/critical journal distributions decreased dramatically over the 3 analysis windows 1998-2018, as demonstrated by means of Spearman's Rho. Thus, the statin-positive and critical journal distributions were increasingly very different.

At the general topical level, the statin research was consistently backed by 2 major US health funding agencies, United States Department of Health Human Services and National Institutes of Health, USA, (NIH), followed by central pharmaceutical companies. However, the 2 US state agencies played minor roles in supporting statin-positive or critical research. During recent periods 2005-2018 Medical Research Council, UK, appeared in the distributions. All the commercial top companies sponsored both statin-positions, but to different degrees. The assumption of sponsors uniquely dedicated a scientific position consistently over time could not be verified by the study, perhaps owing to lack of sponsor coverage in WoS 1998-2011. However, the study showed that approx. 80% of the research 2012-2018 was sponsored by external actors, regardless of statin research position, Tables 5 and 7, that is, a figure 20 percentage points higher than the average sponsor coverage for that period.

In conclusion our hypothesis was entirely substantiated at the general topical mainstream level for journals and supporting organizations. It was partly verified in relation to the disagreeing positions. Some distinct journals central to the two disagreeing positions were uncovered, but that was not the case for sponsors. Observations showed that many top-journals and top-commercial sponsors backed both positions, in the case of sponsors often in an asymmetric way. The major limitation of the study was associated with the sporadic sponsor coverage until 2012. In terms of generalization we claim that the study's research question and methodology are transferrable to other topics demonstrating scientific disagreements. However, with the sponsor limitation in mind, we recommend making 2012 as the analysis starting point in further investigations using WoS as data resource.

Sub-optimal cholesterol response to initiation of statins and future risk of cardiovascular disease

Abandon statistical significance

Information search tactics

From Trust me -I am a Doctor

Statin wars: Have we been misled about the evidence? A narrative review

Scientific inquiry into rhinosinusitis: Who is receiving funding from the National Institutes of Health?

Statins and the BMJ

The turn: Integration of information seeking and retrieval in context

Daekningsgraden i Science Citation Index af dansk sundhedsvidenskabelig forskning 1998

How do we index? A report of some ASLIB informatics group activity

Mind the gap? Quantifying interlinkages between two traditions in migration literature

Varning för simvastatin 80 mg

Conflict of interest disclosure and polarization of scientific communities

Statins for primary prevention: The debate is intense, but the data are weak

Becoming metric-wise: A bibliometric guide for researchers

Null hypothesis significance tests: A mix-up of two different theories-the basis for widespread confusion and numerous misinterpretations

Document representation and indexer consistency

From abstract to impact in cardiovascular research: Factors predicting publication and citation

ion papers below the citation thresholds applied. In addition, this retrieval mode might introduce a new subjective bias owing to the interpretation of the keywords in context. Consequently, this retrieval mode is not to be recommended.The z-tests served to demonstrate the degree of uncertainty or probability of error associated with the data sets containing statin-positive or critical journals as well as sponsors. According to Amrhein, Greenland and McShane (2019) if applying null-hypothesis and significance testing, the researcher should define the alpha value and assess the sample parameters prior to a study with a given population. We did not apply null-hypothesis and statistical significance testing. However, we performed such additional calculations. We regarded each of the 2 statin-positive and critical sets of articles retrieved per period as samples of the 3 original sets. Table 9 shows the sample parameters in a z-test at p = 0.05, i.e., the minimum number of units and maximum mean values necessary in samples from each original period set of journals and sponsors (the given populations), Tables 2-3. Alpha was set to 0.05 since all distributions in the samples had steep start frequencies. A probability of error at 5% might not replace the top-ranked units in a distribution by lower ranked units due to the high frequencies of the former. However, their mutual order might indeed change.With respect to statin-positive and critical journals, Tables 4 and 6, Table 9 demonstrates that in most cases the conditions for p ≤ 0.05 became satisfied, mainly due to a higher number of units which compensated for too high mean values. During the period 2012-2018 the statin-positive journals provided a p value on the borderline (p = 0.06). For sponsors the conditions were far from being satisfied, with p values ≥ 0.30. The implication was that the probability of error became very high (from 30% upwards) and the top-10 sponsors in each distribution might very well change rank order or in some cases be replaced by middlerange frequency units. However, the overall finding that the commercial sponsors backed the 2 opposite statin positions holds. Albeit, the largest bias in the study derives from the low sponsor coverage 1998-2011 in WoS (< 30%), which influences both conditions in non-parametric tests negatively, given known populations. A test of WoS coverage of the topic 'statin use in cardiovascular events' 2012-2019 shows an annual coverage of approx. 60% of the records that contains sponsor data. This consistency leads to the conclusion that