LETTER TO THE EDITOR

Appearance of methicillin-resistant Staphylococcus aureus (MRSA) sensitive to

gentamicin in a hospital with a previous endemic distinct MRSA

Fei Gao1 & David Machin1,21Division of Clinical Trials & Epidemiological Sciences, 11 Hospital Drive, Singapore 169610, Singapore; 2Nordic School of

Public Health, Nya Varvet, Box 12133, SE-40242 Goteborg, Sweden

Accepted in revised form 16 October 2003

We have read the article of Sopena et al. [1] withinterest and would like to comment on two aspects oftheir presentation.

The first concerns the number of GS-MRSA andGR-MRSA infected cases observed in 1996 as sum-marised in their Figure 1. This raises the intriguingquestion as to whether or not there is a seasonalinfluence to these infection types. Using the methodssummarised by [2], the estimated peaks for GR andGS are mid March and 7 months later in late Octoberrespectively. Both are statistically significant usingMardia’s test with 2 degrees of freedom (v2 ¼ 13.0,p-value ¼ 0.002, v2 ¼ 13.7, p-value ¼ 0.001). The testcompares the observed distribution of cases over theyear with the null hypothesis of a uniform distribu-tion. A comparison of these two peak dates usingmethods described in [3] suggests that there is a real(rather than chance) difference between them (p-value<0.001). It is of some interest therefore to ask theauthors if such a pattern was repeated in later years.

Our second comment concerns the interpretationof the large number of statistical comparisons madeof potential risk factors. For example, the authorsreport 26 (53.1%) from 49 cases with GR-MRSAwere nasal carriers, whilst for GS-MRSA this was 14(34.1%) from 41 cases. They conclude that the dif-ference was not statistically significant. However,using the standard test for the comparison of pro-portions gives v2 ¼ 3.23, p-value ¼ 0.07 which isquite close to the conventional 0.05. The corre-sponding 95% confidence interval (CI) for the 19.0%difference in nasal involvement rates is from )1.2 to39.1%. Taken together these results are very sugges-

tive of a greater proportion of nasal carriers withGR-MRSA and this may have clinical relevance al-beit not firmly established. Thus in order to bestinterpret data we suggest that as a routine, as hasbeen recommended for many years (see [4]), exactp-values and CIs should be reported for each com-parison made.

References

1. Sopena N, Garc��a-Nunez M, Prats R, et al. Appearanceof methicillin-resistant Staphylococcus aureus (MRSA)sensitive to gentamicin in a hospital with a previous

endemic distinct MRSA. Eur J Epidemiol 2001; 17: 325–330.

2. Machin D, Chong SF. On the detection of the seasonalonset of disease. J Epidemiol Biostats 1999; 3: 385–394.

3. Gao F, Seah SK-L, Foster PJ, Chia K-S, Machin D.Angular regression and the detection of the seasonalonset of disease. J Cancer Epidemiol Prev 2002; 7: 29–

35.4. Altman DG, Gore SM, Gardner MJ, Pocock SJ. Sta-

tistical guidelines for contributors to medical journals.

Br Med J 1983; 286: 1489–1493. Revised in Chapter 14of Altman DG, Machin D, Bryant TN, Gardner DG(eds), Statistics with Confidence, 2nd ed., London: BMJPublications, 2000.

Address for correspondence: Fei Gao, Division of ClinicalTrials & Epidemiological Sciences, National Cancer Centre,11 Hospital Drive, Singapore 169610, SingaporePhone: +65-6236-9442; Fax: +65-6536-5503

E-mail: ctegfe@nccs.com.sg

European Journal of Epidemiology 19: 497, 2004.� 2004 Kluwer Academic Publishers. Printed in the Netherlands.