Hostname: page-component-848d4c4894-2pzkn Total loading time: 0 Render date: 2024-06-01T13:11:19.661Z Has data issue: false hasContentIssue false
  • English
  • Français

Opinion: on the importance of maintaining the functional form of explanatory variables

Published online by Cambridge University Press:  04 August 2022

Florian Zapf ,
Warwick Butt  and
Siva P. Namachivayam Open the ORCID record for Siva P. Namachivayam [Opens in a new window]
Florian Zapf
Affiliation:
Cardiac Intensive Care Unit, The Royal Children’s Hospital, Melbourne, Victoria, Australia
Warwick Butt
Affiliation:
Cardiac Intensive Care Unit, The Royal Children’s Hospital, Melbourne, Victoria, Australia Clinical Sciences, Murdoch Children’s Research Institute, Melbourne, Victoria, Australia Department of Paediatrics, University of Melbourne, Melbourne, Victoria, Australia Department of Critical Care, University of Melbourne, Melbourne, Victoria, Australia
Siva P. Namachivayam*
Affiliation:
Cardiac Intensive Care Unit, The Royal Children’s Hospital, Melbourne, Victoria, Australia Clinical Sciences, Murdoch Children’s Research Institute, Melbourne, Victoria, Australia Department of Paediatrics, University of Melbourne, Melbourne, Victoria, Australia Department of Critical Care, University of Melbourne, Melbourne, Victoria, Australia
*
Author for correspondence: Siva P. Namachivayam, FCICM, MBios, Cardiac Intensive Care Unit, The Royal Children’s Hospital, Melbourne, Victoria, Australia. E-mail: siva.namachivayam@rch.org.au
Get access Rights & Permissions [Opens in a new window]

Abstract

In medical research, continuous variables are often categorised into two or more groups before being included in the analysis; this practice often comes with a cost, such as loss of power in analysis, less reliable estimates, and can often leave residual confounding in the results. In this research report, we show this by way of estimates from a regression analysis looking at the association between acute kidney injury and post-operative mortality in a sample of 194 neonates who underwent the Norwood operation. Two models were developed, one using a continuous measure of renal function as the main explanatory variable and second using a categorised version of the same variable. A continuous measure of renal function is more likely to yield reliable estimates and also maintains more statistical power in the analysis to detect a relation between the exposure and outcome. It also reveals the true biological relationship between the exposure and outcome. Categorising a continuous variable may not only miss an important message, it can also get it wrong. Additionally, given a non-linear relationship is commonly encountered between the exposure and outcome variable, investigators are advised to retain a predictor with a linear term only when supported by data. All of this is particularly important in small data sets which account for the majority of clinical research studies.

Keywords

Cardiac surgerykidney injurycreatininecategorisationlinearity
Type
Original Article
Information
Cardiology in the Young , Volume 33 , Issue 8 , August 2023 , pp. 1337 - 1341
Check for updates
Copyright
© The Author(s), 2022. Published by Cambridge University Press

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

Sutherland, SM, Byrnes, JJ, Kothari, M, et al. AKI in hospitalized children: comparing the pRIFLE, AKIN, and KDIGO definitions. Clin J Am Soc Nephrol 2015; 10: 554–561.CrossRefGoogle ScholarPubMed
Altman, DG, Royston, P. The cost of dichotomising continuous variables. BMJ 2006; 332: 1080.CrossRefGoogle ScholarPubMed
Royston, P, Altman, DG, Sauerbrei, W. Dichotomizing continuous predictors in multiple regression: a bad idea. Stat Med 2006; 25: 127–141.CrossRefGoogle ScholarPubMed
Naggara, O, Raymond, J, Guilbert, F, Roy, D, Weill, A, Altman, DG. Analysis by categorizing or dichotomizing continuous variables is inadvisable: an example from the natural history of unruptured aneurysms. AJNR Am J Neuroradiol 2011; 32: 437–440.CrossRefGoogle ScholarPubMed
Selvin, S. Statistical power and sample size calculations. Statistical Analysis of Epidemiological Data, 3 rd edn. Oxford University Press, 2004; Book Chapter: 75–92.CrossRefGoogle Scholar
Greenland, S. Avoiding power loss associated with categorization and ordinal scores in dose-response and trend analysis. Epidemiology 1995; 6: 450–454.CrossRefGoogle ScholarPubMed
Buettner, P, Garbe, C, Guggenmoos-Holzmann, I. Problems in defining cutoff points of continuous prognostic factors: example of tumor thickness in primary cutaneous melanoma. J Clin Epidemiol 1997; 50: 1201–1210.CrossRefGoogle ScholarPubMed
Del Priore, G, Zandieh, P, Lee, MJ. Treatment of continuous data as categoric variables in Obstetrics and Gynecology. Obstet Gynecol 1997; 89: 351–354.CrossRefGoogle ScholarPubMed
MacCallum, RC, Zhang, S, Preacher, KJ, Rucker, DD. On the practice of dichotomization of quantitative variables. Psychol Methods 2002; 7: 19–40.CrossRefGoogle ScholarPubMed
Shaw, A, Swaminathan, M, Stafford-Smith, M. Cardiac surgery-associated acute kidney injury: putting together the pieces of the puzzle. Nephron Physiol 2008; 109: p55–60.CrossRefGoogle ScholarPubMed
Blinder, JJ, Goldstein, SL, Lee, VV, et al. Congenital heart surgery in infants: effects of acute kidney injury on outcomes. J Thorac Cardiovasc Surg 2012; 143: 368–374.CrossRefGoogle ScholarPubMed
Alabbas, A, Campbell, A, Skippen, P, Human, D, Matsell, D, Mammen, C. Epidemiology of cardiac surgery-associated acute kidney injury in neonates: a retrospective study. Pediatr Nephrol 2013; 28: 1127–1134.CrossRefGoogle ScholarPubMed
Morgan, CJ, Zappitelli, M, Robertson, CM, et al. Risk factors for and outcomes of acute kidney injury in neonates undergoing complex cardiac surgery. J Pediatr 2013; 162: 120–127 e1.CrossRefGoogle ScholarPubMed
Royston, P, Altman, DG. Approximating statistical functions by using fractional polynomial regression. Journal of The Royal Statistical Society: Series D (The Statistician) 1997; 46: 411–422.Google Scholar
Royston, P, Sauerbrei, W. Building multivariable regression models with continuous covariates in clinical epidemiology--with an emphasis on fractional polynomials. Methods Inf Med 2005; 44: 561–571.Google ScholarPubMed
Bennette, C, Vickers, A. Against quantiles: categorization of continuous variables in epidemiologic research, and its discontents. BMC Med Res Methodol 2012; 12: 21.CrossRefGoogle ScholarPubMed
Cohen, DS. The cost of dichotomization. Applied psychological measurement 1983; 7: 249–253.CrossRefGoogle Scholar
Greenland, S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology 1995; 6: 356–365.CrossRefGoogle ScholarPubMed
van Walraven, C, Hart, RG. Leave ‘em alone - why continuous variables should be analyzed as such. Neuroepidemiology 2008; 30: 138–139.CrossRefGoogle ScholarPubMed
Royston, P, Sauerbrei, W. Chapter 3: Handling categorical and continuous predictors. multivariable model-building: A pragmatic approach to regression analysis based on fractional polynomials for modeling continuous variables. John Wiley & Sons Ltd 2009: 58.Google Scholar
Supplementary material: File

Zapf et al. supplementary material

Zapf et al. supplementary material

Download Zapf et al. supplementary material(File)
File 22.8 KB