We read with interest the valuable article by Castillo-Tokumori and colleagues published in the International Journal of Infectious Diseases (
Castillo-Tokumori et al., 2017
). The authors aimed to describe community-acquired urinary tract infections (CA-UTI) caused by extended-spectrum beta-lactamase (ESBL)-producing Escherichia coli and its risk factors. They conducted a case–control study involving 67 patients with ESBL and 105 patients with non-ESBL E. coli CA-UTI. The authors found that chronic corticosteroid use was significantly associated with ESBL CA-UTI, which is problematic.The authors reported (in their Table 2) that there was a statistically significant association between chronic corticosteroid use and ESBL CA-UTI (crude odds ratio (OR) 8.39, 95% confidence interval (CI) 0.96–73.45; p = 0.023). However, there are issues regarding the CI and p-value. In a statistically significant association, the CI should not cross unity (1) and the P-value should be less than 0.05 (
Rothman et al., 2008
). In the aforementioned association, the CI has crossed 1, whereas the reported P-value is less than 0.05. We examined this association and found it to be non-significant (correct P-value = 0.06) (Table 1).Table 1Univariate association between chronic corticosteroid use and ESBL CA-UTI using ordinary and penalized logistic regression.
ESBL | Non-ESBL | |
---|---|---|
Corticosteroid use | ||
Yes | 5 | 1 |
No | 62 | 104 |
Estimated OR (95% CI) | ||
Ordinary logistic regression | 8.38 (0.95–73.45) | |
Penalized logistic regression | 4.42 (0.86–22.46) |
ESBL, extended-spectrum beta-lactamase; CA-UTI, community-acquired urinary tract infection; OR, odds ratio; CI, confidence interval.
Furthermore, the authors reported a large OR with a wide CI for the association between chronic corticosteroid use and ESBL CA-UTI in the univariate (OR 8.39, 95% CI 0.96–73.45) and multivariate models (OR 24.32, 95% CI 2.39–246.92), which is questionable. The researchers indicated that the large effect estimates and wide CIs resulted from the sparse data, as there were insufficient numbers of observations in the different strata of the independent and dependent variables (
Greenland et al., 2016
, Greenland and Mansournia, 2015
). The sparseness of the data would have been severe in the multivariate models, since the number of strata is increased in these models (Greenland et al., 2016
).The data provided by Castillo-Tokumori et al., on the association between chronic corticosteroid use and ESBL CA-UTI, are clarified in Table 1. The sparseness of the data was confirmed, as the number of observations was low and the adjusted OR and 95% CI were larger and wider, respectively, compared to the corresponding univariate OR and 95% CI. Hence, the bias due to sparse data should be removed using appropriate and efficient statistical methods, such as penalization via data augmentation (
Greenland et al., 2016
, Ayubi and Safiri, 2017
). We re-analyzed the crude OR using the penalization method with log-f (2,2) prior distribution, and the estimations improved remarkably (Table 1). The adjusted OR could also be corrected using this method, but the individual data would be needed. Hence, we propose that Castillo-Tokumori et al. re-analyze their adjusted estimates using the method introduced herein, in order to obtain unbiased and valid estimates. It is expected that the adjusted unbiased association between chronic corticosteroid use and ESBL CA-UTI would be diluted and non-significant.Funding
None.
Conflicts of interest
None of the authors has a conflict of interest to disclose.
Author contributions
EA and SS designed the study. EA and SS drafted the manuscript. Critical revision was done by EA and SS.
References
- “Predictors of failure after single faecal microbiota transplantation in patients with recurrent Clostridium difficile infection: results from a three-year cohort study”; methodological issues.Clin Microbiol Infect. 2017; 23: 890
- Worrisome high frequency of extended-spectrum beta-lactamase-producing Escherichia coli in community-acquired urinary tract infections: a case–control study.Int J Infect Dis. 2017; 55: 16-19
- Penalization, bias reduction, and default priors in logistic and related categorical and survival regressions.Stat Med. 2015; 34: 3133-3143
- Sparse data bias: a problem hiding in plain sight.BMJ. 2016; 352: i1981
- Modern epidemiology.Lippincott Williams & Wilkins, 2008
Article info
Publication history
Published online: November 11, 2017
Received:
April 12,
2017
Corresponding Editor: Eskild Petersen, Aarhus, DenmarkIdentification
Copyright
© 2017 The Author(s). Published by Elsevier Ltd on behalf of International Society for Infectious Diseases.
User license
Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0) | How you can reuse
Elsevier's open access license policy

Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0)
Permitted
For non-commercial purposes:
- Read, print & download
- Redistribute or republish the final article
- Text & data mine
- Translate the article (private use only, not for distribution)
- Reuse portions or extracts from the article in other works
Not Permitted
- Sell or re-use for commercial purposes
- Distribute translations or adaptations of the article
Elsevier's open access license policy
ScienceDirect
Access this article on ScienceDirectLinked Article
- Worrisome high frequency of extended-spectrum beta-lactamase-producing Escherichia coli in community-acquired urinary tract infections: a case–control studyInternational Journal of Infectious DiseasesVol. 55
- PreviewThere has been a sustained and dramatic increase in community-acquired urinary tract infections (CA-UTI) caused by extended-spectrum beta-lactamase (ESBL)-producing bacteria over recent years. Despite this, no studies have been performed in low- or middle-income countries. The main objective of this case–control study was to describe ESBL CA-UTI and its risk factors.
- Full-Text
- Preview
- Response to “Worrisome high frequency of extended-spectrum beta-lactamase-producing Escherichia coli in community-acquired urinary tract infections: a case–control study; methodological issues”International Journal of Infectious DiseasesVol. 66
- PreviewWe, the authors of the original article, have decided not to implement the statistical method ‘penalization via data augmentation’ to our study, as suggested by Erfan Ayubi and Saeid Safiri. We have investigated whether this method could generate a valid odds ratio (OR) regarding corticosteroid use (as a risk factor); unfortunately, we would have to assume unreal data before applying the method, as we need a tentative OR a priori. The necessary data are currently unavailable, as we found this association incidentally.
- Full-Text
- Preview