The problem of retractions and contentious papers

Failing to correct erroneous scientific and medical research makes fools of us all

A truth that’s told with bad intent Beats all the lies you can invent William Blake (1757–1827)

On 9 November, Pfizer and BioNTech announced that their vaccine was ‘more than 90 per cent effective in preventing Covid-19 in participants without evidence of prior SARS-CoV-2 infection…’ The Lancet’s Editor-in-Chief Dr Richard Horton responded: “Publishing interim results through a press release is neither good scientific practice nor does it help to build public trust in vaccines.

An announcement should come with full publication of a peer-reviewed research paper in a scientific journal.”

But how reliable are peer-reviewed research papers? In a piece for a national newspaper, I cited an article by Dr Nicolas Guéguen in Archives of Sexual Behaviour (2015, 44: 2227–2235) – “High heels increase women’s attractiveness” – which the journal later retracted (2020, 49: 85) for “serious methodological weaknesses and statistical errors”.

And recently, Prof Katrina Bramstedt considered in the Journal of Medical Ethics “The carnage of substandard research during the Covid-19 pandemic:
A call forquality”, noting that “[a]s of 31 July 2020, 19 published articles and 14 preprints about Covid-19 have been retracted, withdrawn, or an expression of concern has been issued.”

I felt annoyance with Dr Guéguen; embarrassment that my interest in high heels (yes, I could have rephrased that better) had ended with misinformed readers; and wondered how I may have influenced sales of peep-toe stilettoes. But nobody died.

Prof Bramstedt’s findings, however, invite the inference that publishing false results could be fatal. Yet both episodes illuminate a spectrum, from mistakes to fraud, that blights scientific discourse.

The website www.retractionwatch.com monitors research misconduct and is cited both by Profs Bramstedt and Avenell et al who described “An investigation into the impact and implications of published papers from retracted research…” in BMJ Open (2019;9:e031909).

They examined 12 randomised clinical trial reports tainted by misconduct, and found that the flawed reports “were cited in 1,158 publications, including 68 systematic reviews, meta-analyses, narrative reviews, guidelines, and clinical trials”, commenting that “[c]orrection of these distortions is slow, uncoordinated, and inconsistent”.

The “carnage of substandard research”, and the “slow, uncoordinated, and inconsistent” correction of distortions identified by Profs Bramstedt and Avenell et al illustrate the extent to which we can be treated like
fools.

Bryan Appleyard warned in his Understanding the Present (1993): “Scientists need to be observed and criticised more than any other members of society… [and be]… as morally and philosophically answerable as the rest of us.” He’s right.

An instructive lesson arises from the burgeoning evidence base confirming the effectiveness of low-carbohydrate diets in addressing diabetes, and endorsed by, for example, the Government of Western Australia, the American Diabetes Association, and Diabetes Canada.

Yet opponents promote the outdated low-fat high-carbohydrate approach to weight and diabetes management. When Prof Tim Noakes challenged them, he ended up in court in 2015… and won.

In his Lore of Nutrition (2017), co-authored by Marika Sboros, two papers whose errors attracted notoriety are highlighted. One – the largest ever trial of a low-fat diet – saw Howard et al (JAMA, 2006; 295: 655–666) investigate its effect on cardiovascular disease risk in 48,835 post-menopausal women.

During his 2012 Centenary Debate in Cape Town with one of the study’s architects Prof Jacques Rossouw, Prof Noakes exposed the absence of a vital line on Figure 3 identifying a 26 per cent increased risk of further heart problems among participants with pre-existing cardiovascular disease, but “if you missed the key sentence buried in the article and only looked at this table, you would never know about the study’s only significant finding”.

I put this to Prof Rossouw, who replied: “Noakes picked on a typesetting error in Figure 3…. We should have seen that during proofreading, but we did not, and to the best of our knowledge, Noakes is literally the only one out of the thousands who have read this paper to have raised this as an issue.”

But one out of thousands (including the report’s 47 co-authors) or not, when errors are identified they should be corrected, thereby avoiding Blake’s minatory admonition (above), and affirming civilised values.

A second paper, central to charging Prof Noakes with “disgraceful conduct”, was a review by Naudé et al (PLoS One, 2014; 9: e100652), which found little difference in weight loss and cardiovascular risk factor changes when overweight adults, with or without type 2 diabetes, were randomised to low-carbohydrate diets and isoenergetic balanced weight loss diets.

However, Harcombe and Noakes (S Afr Med J, 2016;106:1179-1182) found 14 material errors and asked whether these constituted “mistake or mischief”. In 2018 the PLoS One Editors (PLoS One, 2018; 13: e0200284) published a five-page correction to the Naudé review. I have twice emailed Dr Naudé to ask why the editors – and not the authors – issued these corrections, but have not received a reply.

It is time to cut through the dissembling and obfuscation that risk tainting the honourable pursuit of science and turn to plain speaking. If mistakes remain uncorrected, we become spectators to unfairness, which diminishes us all.