Background

“Science is built of facts the way a house is built of bricks: but an accumulation of facts is no more science than a pile of bricks is a house” (Henri Poincaré).

Fake news and alternative facts have become commonplace in these so-called “post-factual times.” What about medical research? Are scientific facts fake as well? A wide range of scientific results have been shown to be false [1]. Even much cited studies don’t hold up and are hard to replicate [2,3,4,5,6,7]. Initially strong effects of clinical interventions reported in highly cited articles are frequently contradicted [8]. Scientific results are fashioned by who finances research [9] and by researchers’ ties to industry [10]. Spoof research is frequently accepted [11], and scientific truth and objectivity is challenged [12, 13]. All this fuels the claim that scientific facts are suspect and that science is in crisis [14].

One source of crisis in science is when facts are based on confirmative empirical testing [15] or that research hypotheses, models, and approaches are directed by strong interests. The latter can be observed in polarised fields of research. Polarisation occurs when “reputable scientists hold radically opposed views leading to the segregation of the scientific community into groups in part constituted by their opposition to other groups in the field. Polarisation goes beyond mere disagreement. It occurs when researchers begin (1) to self-identify as proponents of a particular position that needs to be strongly defended beyond what is supported by the data and (2) to discount arguments and data that would normally be taken as important in a scientific debate” [16]. In polarised research scientists come to engage in facting interests instead of revealing interesting facts.

Main text

How then are we to identify and address such “polarised facts?” One approach is to reveal polarised research fields and to put polarisation on par with other forms of conflicts of interests in scientific publishing. Let me use mammography screening as an example to illustrate how polarised facts can be investigated. In this field there are two main points of disagreement: a) What is the benefit of mammography screening, e.g., in terms of reduced breast cancer mortality, and b) what is the harm of this type of screening, e.g., in terms of overdiagnosis? Some researchers tend to claim that the mortality reduction is high, while the overdiagnosis rate is low [17], while others claim that the mortality reduction rate is moderate, while overdiagnosis is high [18]. What is at stake is the risk/benefit-ratio in a utilitarian perspective. Hence, one way to illustrate the polarisation in this field is to scrutinize the divergence in the Overdiagnosis to Mortality Reduction Ratio (OMRR), that is, the ratio of overdiagnosis over the rate of mortality reduction. “Overdiagnosis is the term used when a condition is diagnosed that would otherwise not go on to cause symptoms or death” [19]. Mortality from breast cancer is defined as deaths with breast cancer coded as the underlying cause of death and mortality reduction is defined in terms of reduced breast cancer mortality in a screened group compared to a non-screening group in the assessment of a screening program.

Accordingly, the research questions of this brief study are: What is the OMRR in publicly funded mammography screening programs of women aged 50–69 years old? How is this related to the corresponding authors’ attitudes towards screening? A straight forward literature search identifies 8 studies who have addressed the first question. The studies and their results are shown in Table 1.

Table 1 Overdiagnosis to mortality reduction ratio (OMRR) for various studies and the corresponding author’s attitudes to mammography screening as assessed by experts in polarised research (1: Very negative to screening, 2: Negative to screening, 3: Neutral to screening, 4: Positive to screening, 5: Very positive to screening)
Full size table

In order to assess the researchers’ attitudes to screening specific questions suggested to identify “polarised conflict of interest.” were adapted to this particular case and were sent to the corresponding authors of the identified publications. However, the corresponding authors found it difficult to answer the questions. As expected, “researchers within a polarised group in a polarised field may not themselves be able to identify the field as polarised or see themselves as belonging to a polarised group”. In order to overcome this problem, two experts on polarised conflict of interest were asked to classify the corresponding authors of the identified publications. Inclusion criteria for these experts were that they were experts on science ethics in general and polarised research in particular, and exclusion criteria were if they had been involved in mammography screening programs or their primary evaluations. The research question and the included articles were not revealed. For a description of the literature search, the questions to the authors, and the questions to the experts, and a discussion of the applied methods, see Additional file 1. The classification of the corresponding interests is given in Table 1.

The correlation between the OMRR and the authors’ attitudes to screening as assessed by experts in polarised research was strong (R = 0.9). The scatter plot is shown in Fig. 1.

Fig. 1
figure 1

Scatter plot of the relationship between OMRR and attitudes to screening. (1: Very negative to screening, 2: Negative to screening, 3: Neutral to screening, 4: Positive to screening, 5: Very positive to screening)

Full size image

This indicates that research results in this field are strongly formed by professional interests and attitudes to screening. The same effect of “facted interests” may be observed in other fields of polarised research, and in-depth studies are needed and encouraged.

Of course, personal interest can be a good thing in science. It can motivate important and ground-breaking research. However, it can also bias judgments, cloak influences, direct methodological choices, and skew the presentation of results. Moreover, framed facts can influence important health policy decisions. Accordingly, it is crucial to acknowledge that this represents “genuine conflicts of interest” threatening “the objectivity of science” [12] and trust in science. This can be done by a) making researchers state their “polarised conflict of interest” when submitting manuscripts, b) making reviewers explicitly assess polarisation, and c) apply external experts to assess polarisation when reviewers (and/or editors) are too ingrained in the research to be able to make the assessment.

Polarisation may be a general trend resulting from disagreements on research methodology or assessment of evidence (according to GRADE or other). However, it may also result from self-interest [20], intellectual laziness [21, 22], mental shortcuts, or hyper-partisanism [23]. Moreover, emotional conflicts of interests are more difficult to handle than financial conflicts [24].

While philosophers of science and sociologists long have revealed the challenges of value-laden facts and underscored the constitutive value of disinterestedness in science [12], it is high time we scientists acknowledge this in practice.

Conclusion

Scientists appear to engage in facting interests as much as in revealing interesting facts. Published research on mammography screening for breast cancer illustrates the problem of science being directed by strong professional interests, where some researchers continuously publish positive results while others publish negative results on the same issue – even when based on the same data. Analysing this as polarised research may provide a way to address an important issue threatening to undermine trust in scientific results and medical researchers. Hence editors should a) make researchers state their “polarised conflict of interest” when submitting manuscripts, b) make reviewers explicitly assess polarisation, and c) apply external experts to assess polarisation when reviewers (and/or editors) are too ingrained in the research to be able to make the assessment.

How exactly to assess polarised conflict of interest may need more elaboration and collaborate work. However, Table 2 suggests some questions to ask when assessing polarised conflict of interest. This is a first step illustrating methodological and empirical feasibility.

Table 2 Relevant questions to ask when assessing polarised conflict of interest
Full size table