2014-2-Sadeghi-letter

JRHS 2014; 14(2): 167-168

Copyright© Journal of Research in Health Sciences

Letter to the Editor

Comment on:

Prevalence of Diabetes in people aged >30 years: The results of screening program of Yazd province, Iran in 2012

Received: 01 January 2014, Revised: 19 January 2014, Accepted: 19 January 2014, Available online: 05 February 2014

Dear Editor-in-Chief

We read with great interest the article published in JRHS journal by Lotfi et al.1 They discussed on the prevalence of diabetes in people aged 30 years using results of screening program in Yazd Province. Authors found that prevalence of diabetes in women (25.3%) was more than men (9.2%). It seems that high prevalence of type 2 diabetes in women due to the large portion (more than 90 percent) of women who have participated in the study were 50 years and above. In other words, the mean age of women was higher than men (Table 21; diabetic women = 1235, total number of diabetic patients 50 years = 1206). If this is confirmed by the authors, the high prevalence of diabetes among women in present study is due to selection bias and not volunteer bias that has cited in the discussion. In fact, it was better that sampling was done better, or analysis should be standardized based on province or national population. Authors only in this way could claim that obtained measurements in the study are related to people aged 30 yr. Thus, the prevalence of diabetes in the general population and in subgroups (gender) will differ from the reported facts and figures.

Authors have mentioned that total of 10 variables that had significant P-value in univariate analysis were entered into the logistic regression models. These 10 variables are not clear and with regard to the results, residence place was not significant in the univariate model (P=0.36). However this variable was entered into the multiple models and also was significant. So there is a contradiction. Authors reported that the variables were significant in the multiple models and one of these variables was gestational diabetes. As authors know, this factor is limited to subgroups of women, but the reported results and analysis was performed for the total population including men and women. Is it possible to enter variable in the final model while it is limited to one subgroup?

In the methodology section they reported that 3966 people at risk were not screened (lost). The inference is that 2790 people were female. In the second paragraph of results, the authors have analyzed data on 7676 women. This means that also missed women were included in the analysis. Is this inference correct? If correct, is it possible to evaluate the association between diabetes and risk factors related to women?

Furthermore authors reported that the final analysis was performed on a total of 11,027 people and total of 4886 women were studied (Table 1). But findings of 7676 screened women have been reported in the second paragraph of results.

It is important that the term "prevalence" in place of "prevalence rate" would be replaced in the text. As the authors are well aware of the fact that prevalence is a proportion rather than rate.2 

Using the term of "case control" in the study that is actually comparative cross sectional is not correct. It is important to know groups as defined in Table 2, as an alternative to case and control groups, diabetic and non-diabetic groups should have been written.3 

Finally, in Table 2, it was desirable that findings related to crude odds ratios and adjusted odds ratios were reported in the presence of all variables (both significant and non-significant variables). So audiences could have a proper perception and interpretation of the presented results.4 

Masoumeh Sadeghi (MSc)a*, Mohsen Dehghani (MSc)b, Ali Akbar Haghdoost (MD, PhD)c

a Research Center for Modeling in Health, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran

b Department of Epidemiology, School of Public Health, Shahroud University of Medical Sciences, Shahroud, Iran

c Research Center for Modeling in Health, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran

Correspondence to: Masoumeh Sadeghi (MSc), E-mail: sadeghi.masoume@gmail.com

References

  1. Lotfi MH, Saadati H, Afzali M. Prevalence of diabetes in people aged 30 years: The results of screening program of Yazd province, Iran, in 2012. J Res Health Sci. 2014;14(1):87-91.
  2. Kenneth J. Rothman. Epidemiology: an introduction. New York: Oxford University Press; 2012.
  3. Vandenbroucke JP, Von Elm E, Altman DG, Gotzsche PC, Mulrow CD, Pocock SJet al. Strengthening the reporting of observational studies in epidemiology (STROBE): explanation and elaboration. Gac Sanit. 2009; 23(2):158.
  4. Nasirian M, Sadeghi M, Haghdoost AA. Principle of data analysis and proper presentation of results. Journal of Isfahan Medical School. 2001; 27(100):645-658. [Persian].

Reply

I would like to thank authors for showing interest in our article. With regard to the stated comments, some explanations are provided as follows. 

In respect to being high prevalence of diabetes amongst women compared to men; our study was a cross-sectional type that was mainly based on the data already collected. Cross-sectional studies are often prone to various bias and the comment about the high prevalence of diabetes in women due to selection bias, may not be ignored but since that women with a previous history of diabetes compared to known diabetic men had higher welcoming to a diabetes screening program (24.1 versus 7.4), the higher prevalence of diabetes in women can be attributed to emphasizing their health, concluded due to volunteer bias. 

As for residence place variable, entered into the logistic regression model. To find the predictive and effective factors on diabetes, the variables that have significant P-values and also those that have no significant P-values but were clinically important such as residence place, they also were entered into the regression model. 

Regarding gestational diabetes variable in the multiple model; the given comment cab be correct, but we performed a separate analysis for the men and women. However, it could have been mentioned at the footnote of the regression table.

In relation to the comment said in the second paragraph of results, findings of 7676 screened women were related to the descriptive analysis, but the final analysis was performed on a total 4886 women. (Data on 2790 missed women were not included in the analysis). 

The term "prevalence" in place of "prevalence rate”; the suggested comment can be used, so there is no conflict and term of prevalence has to be replaced instead of prevalence rate.

In the case of using the term of “diabetic and non-diabetic groups” instead of "case control", we agree to this comment.

Hassan Saadati (MSc)a

a Department of Biostatistics & Epidemiology, Health Faculty, Shahid Sadoughi University of Medical Sciences, Yazd, Iran

E-mail: hasansaadati1391@gmail.com 



JRHS Office:

School of Public Health, Hamadan University of Medical Sciences, Shaheed Fahmideh Ave. Hamadan, Islamic Republic of Iran

Postal code: 6517838695, PO box: 65175-4171

Tel: +98 81 38380292, Fax: +98 81 38380509

E-mail: jrhs@umsha.ac.ir