SciELO - Scientific Electronic Library Online

 
vol.84 issue1Viral respiratory infections: behind simplicity lies intelligenceData with unexpected values should be checked author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Jornal de Pediatria

Print version ISSN 0021-7557

J. Pediatr. (Rio J.) vol.84 no.1 Porto Alegre Jan./Feb. 2008

http://dx.doi.org/10.1590/S0021-75572008000100017 

LETTERS TO THE EDITOR

 

Statistical and epidemiological methods in prevalence studies: odds ratio vs. prevalence ratio

 

 

Dear Editor,

When reading the article by Rodrigues et al.,1 "The association between cardiorespiratory fitness and cardiovascular risk in adolescents," it was possible to find some errors regarding the application of statistical and epidemiological methods, as well as their respective interpretations; however, in our opinion, this does not invalidate the findings, but it can depreciate the scientific method of that study, which is not desirable for the authors, readers or even Jornal de Pediatria, which is such an important journal.

That article is a result of a cross-sectional study, i.e., a prevalence study, situation in which association estimates are preferentially calculated by prevalence ratio (PR) or, less adequately, by odds ratio (OR) and their respective 95% confidence intervals (95%CI). That is so because, in this type of study, it is not possible to determine incidence. Analyzing the results, one can see that OR and relative risk (RR) were used as association estimators. Both are inadequate, since it is known that OR overestimates strength of association,2,3 and RR cannot be estimated, since it is the possibility of calculating prevalence, and not incidence.4

Another important aspect verified was that the 95%CI of estimators used was not published. In our opinion, their visualization is of great help in a proper analysis of results, because it allows estimating adequacy of sample size and verifying statistical significance of the association, besides being an extra aspect when searching for causal inference.

We hope to have contributed with our observations and suggest that, in studies of that nature, preference is given to estimation of strength of association by PR, always showing its 95%CI. Therefore, applicability and interpretation of statistical tools used in epidemiological studies are adequate to their purposes.

We stress the importance of that article and reinforce that reported inadequacies do not diminish the merit, nor invalidate results; presentation forms only need to be corrected.

 

 

No conflicts of interest declared concerning the publication of this letter.

Altacílio Nunes
Doutor. Professor adjunto, Departamento de Medicina Social, Universidade Federal do Triângulo Mineiro (UFTM), Uberaba, MG, Brazil. Coordenador do Internato de Pediatria, Medicina, Universidade de Uberaba, Uberaba, MG, Brazil.

 


 

Authors' reply

 

 

Dear Editor,

We read and appreciated the contributions sent by Professor Altacílio Nunes concerning the article "The association between cardiorespiratory fitness and cardiovascular risk in adolescents"1 about application of statistical methods, an area in which we have much to learn.

The authors would like to clarify that, in that type of study, odds ratio (OR), although numerically higher, as shown in Tables 3 and 4,1 follows relative risk (RR) and is a good estimate for it.2 Such method (RR) has been recommended as first choice to determine exposure risk to a certain disease, and perhaps the most adequate to the objectives of that study.

However, it is inadequate in cross-sectional research studies, such as that being discussed here. Option for not using prevalence ratio (PR), although it can be used, is due to the fact that, in cross-sectional studies,3 OR allows identification of possible associations in which PR may lead to false conclusions.2

The authors acknowledge not drawing attention to the fact that, although RR has lower values for investigated associations, due to methodological inadequacy, those that should be considered are OR values, an estimated approximation of RR. The authors do not acknowledge inadequacy of using OR in their research study.

 

 

No conflicts of interest declared concerning the publication of this letter.

Anabel Nunes Rodrigues
Doutor. Professor adjunto, Faculdade Salesiana de Vitória, Vitória, ES, Brazil

 


 

References (letters to the editor)

1. Rodrigues AN, Perez AJ, Carletti L, Bissoli NS, Abreu GR. The association between cardiorespiratory fitness and cardiovascular risk in adolescents. J Pediatr (Rio J). 2007;83:429-35.         [ Links ]

2. Thompson ML, Myers JE, Kriebel D. Prevalence odds ratio or prevalence ratio in the analysis of cross sectional data: what is to be done? Occup Environ Med. 1998; 55; 272-7.         [ Links ]

3. Pearce N. Effect Measures in Prevalence Studies. Environ Health Prospect. 2004;112:1047-50.         [ Links ]

4. Gordis L. Epidemiology. 2nd ed. Philadelphia: W.B. Saunders Company; 2000.         [ Links ]

 

Referências (resposta dos autores) / References (authors' reply)

1. Rodrigues AN, Perez AJ, Carletti L, Bissoli NS, Abreu GR. The association between cardiorrespiratory fitness and cardiovascular risk in adolescents. J Pediatr (Rio J). 2007;83:429-35.         [ Links ]

2. Rumel D. "Odds ratio": algumas considerações. Rev Saude Publica. 1986;20:253-8.         [ Links ]

3. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research. Beltmont, CA; Lifetime Learning; 1982.        [ Links ]