lack of “statistically significant” association does not exclude causality
TRANSCRIPT
The Spine Journal 10 (2010) 944–946
Letters to the Editor
Lack of ‘‘statistically significant’’ association does notexclude causality
To the Editor:
This series of four systematic reviews aimed to evaluatethe causal relationships between several occupational fac-tors and low back pain [1–4]. The criteria for the interpre-tation of causality are based on the well-known ‘‘aspects ofassociation’’ discussed by Bradford-Hill. The authors usedfive ‘‘criteria’’ in their analysis: statistical association;dose-response relationship; experimental evidence; tempo-ral relationship; and biological plausibility. In the selectionof literature the authors did, however, exclude reports onbasic sciences and biomechanics, and only studies with atleast 30 subjects were included. All exposures under inves-tigation were related to biomechanical and physiologicalfactors. This limitation to epidemiologic studies excludesa huge amount of experimental laboratory studies that haveshown evidence of short-term effects between mechanicalexposures and outcomes related to low back pain, whenevaluated by these five aspects of causality. In addition, ep-idemiologic studies can hardly provide a logical conclusionon biological plausibility, whereas some biological expla-nations were borne in mind while planning the epidemio-logic study or discussed in the article.
Conclusions such as ‘‘It is unlikely that awkward occu-pational postures are independently causative of LBP in thepopulations of workers studied’’ are obviously incorrect be-cause they are based on a very mechanical comparison ofstudies with different designs, populations, and methods.The correct conclusion could be that the information onthe evidence of causality in these studies, performed in in-dustrial countries, is insufficient. In agricultural work in de-veloping countries, where women work for 12 hours a day,the results would obviously be completely different.
Bradford-Hill’s ‘‘aspects’’ related to causality have oftenbeen wrongly referred to as ‘‘criteria’’ for causality. Thisfallacy in the interpretation of causality has been discussedin several epidemiologic articles [5]. Of the nine ‘‘aspects’’,only the temporal relationship between the cause(s) and ef-fect(s) seems to be a necessarydalthough insufficientdcriterion, and none of the other aspects are either necessaryor sufficient criterion for proof. It seems it is not possible tofind a set of criteria or procedures that will formally provethe causality between the proposed cause(s) and effect(s) inepidemiology.
1529-9430/$ - see front matter � 2010 Elsevier Inc. All rights reserved.
Checklists and rating scales are commonly used to eval-uate the quality of original reports and the level of evidencein systematic reviews. Use of these tools should not beautomatic but critical, keeping in mind that the scales andcutoff limits for decisions are arbitrary, and that there isno theoretical ground to generalize them as golden stan-dards. Causality or lack of causality cannot be concludedby scales. Neither does lack of statistical associationexclude causality.
Systematic reviews are highly respected as a summary ofevidence onmedical questions and arewidely used for healthpolicy decisions. Therefore, the authors of these reviews, andthe reviewers recommending publication, should be aware ofthe fallacies related to the statistical inference and logic ofcausal conclusions in epidemiologic studies to prevent themfrom conveying the wrong summary statement.
References
[1] Wai EK, Roffey DM, Bishop P, et al. Causal assessment of occupa-
tional bending or twisting and low back pain: results of a systematic
review. Spine J 2010;10:76–88.
[2] Roffey DM, Wai EK, Bishop P, et al. Causal assessment of awkward
occupational postures and low back pain: results of a systematic
review. Spine J 2010;10:89–99.
[3] Roffey DM, Wai EK, Bishop P, et al. Causal assessment of occupa-
tional sitting and low back pain: results of a systematic review. Spine
J 2010;10:252–61.
[4] Roffey DM, Wai EK, Bishop P, et al. Causal assessment of occupa-
tional standing or walking and low back pain: results of a systematic
review. Spine J 2010;10:262–72.
[5] Ward A. The role of causal criteria in causal inferences: Bradford
Hill’s ‘‘aspects of association’’. Epidemiol Perspect Innov 2009;6:2.
Esa-Pekka Takala, MD, PhDHelsinki, Finland
doi:10.1016/j.spinee.2010.07.008
Reply
We wish to thank Dr Takala for his interest in our man-uscripts [1–8] and understand that our findings may appearcounterintuitive to some readers.
The starting point for undertaking these studies was theconfusion surrounding the potential role that specific occu-pational physical activities might have in the development