The need for attention to confirmation bias and confounding in the field of plastic pollution and wildlife impacts: comment on “Clinical pathology of plastic ingestion in marine birds and relationships with blood chemistry”
Roman, L and Gilardi, K and Lowenstine, L and Hardesty, BD and Wilcox, C, The need for attention to confirmation bias and confounding in the field of plastic pollution and wildlife impacts: comment on 'Clinical pathology of plastic ingestion in marine birds and relationships with blood chemistry', Environmental Science and Technology, 55, (1) pp. 801-804. ISSN 0013-936X (2021) [Contribution to Refereed Journal]
In their recent paper, Lavers et al. 2019 measured blood chemistry parameters in fledgling seabirds, finding a link between ingested plastic and growth, calcium, uric acid, and cholesterol. The manuscript suggests varied toxicological interpretations and concluded that superficially healthy animals may still experience negative health consequences of ingesting plastic debris: an influential message with dire consequences for wildlife globally. While well intended, we posit that the interpretations reveal confounding and confirmation bias and highlight a growing trend within the field of plastic pollution and wildlife impacts. By evaluating recent publications, we provide a framework for assessing confounding and biases in plastic and wildlife research, and we reinterpret Lavers et al.’s results to suggest a nutrition explanation.
As bite-sized plastics increasingly invade our oceans, the past decade has seen a surge in studies seeking to understand how plastic affects the health of humans and wildlife.(1−3) Scientists perceive hundreds of impacts from exposure to plastic, but recent reviews highlight a growing problem in a field aiming to reveal subtle effects: not all claims are supported by evidence, and fewer by evidence in realistic exposure scenarios.(1,2,4) A spate of critical opinions have recently been published, detailing poor quality control(4−7) and a lack of harmonization(8) among the growing pains of the emerging field of plastic and wildlife impacts. In this correspondence, we will not reiterate these criticisms but add to the conversation by positing that mismatches between perception of threats and measurable impacts may be symptoms of unconscious confirmation bias and confounding. In the arms race to publish significant and citable conclusions,(9,10) we argue that real-world context and objective interpretation of results are lacking in the plastic wildlife field. Going beyond what is "good" and "bad" science, there is a danger that confirmation bias may lead to the publication of spurious results, which could be misleading for the field and the public, and erode reputational standing and relevance with policy makers facing pressure to act. By highlighting recent examples, we hope to improve researcher and reviewer practices in the disciplines that study plastic and wildlife interactions.
Research seeking to understand the impact of plastic on wildlife primarily encompasses observational studies due to the challenges of experimental research on wild animals. Confirmation biases can be detected by scrutinizing three aspects of a research study. First and primarily, confounding: are there factors other than plastic that might influence the response and are alternative explanations explored? Second, statistical significance (or lack of): are data dredging approaches employed and are results subject to unbiased, balanced critique? Finally, context: are the conclusions justified given potential confounding, statistical significance, and the environmental/ecological context of the study?