Statistics from Altmetric.com
We read with great interest the article of Avunduk and coworkers,1 who conducted a study in using confocal microscopy to evaluate Aspergillus fumigatus keratitis in treated and untreated rabbit eyes. They concluded that “confocal microscopy is a rapid and sensitive diagnostic tool for both the early diagnosis and non-invasive follow up of fungal keratitis.” In order to justify the statement, two issues of concern on the early diagnosis have to be addressed.
The first is about the sensitivity in having positive diagnosis in the untreated eyes in the first experiment. On day 2, all 14 samples were smear and culture positive for Aspergillus fumigatus, therefore confocal microscopy could not demonstrate any superiority in early diagnosis in terms of sensitivity. On days 14 and 22 their conclusion that “confocal microscopy was more sensitive than culture technique” also could not be drawn unless the authors could enlighten us with supplementary data on the percentages of positive culture in those periods together with their p values.
Another issue is about computation of statistical values in the second experiment. The authors implied that topical and orally treated eyes had significantly lower positive culture growth than the control group receiving no treatment on days 14 and 22 by listing p values of 0.002 and 0.003. However, in performing the χ2 analysis again with the data provided, we can only achieve p = 0.391 and p = 0.280 on day 14 and p = 0.308 and p = 0.237 on day 22. We would suggest that statistical differences cannot be demonstrated in these parts of study, at least, with such a sample size.
We thank Dr Fan and coworkers for their letter and interest in our article.
The conclusion drawn by us was that confocal microscopy was a rapid and sensitive diagnostic tool for both early diagnosis and non-invasive follow up of fungal keratitis, not that it was superior to culture and corneal biopsy staining techniques in the early stage of fungal keratitis. It is rapid compared to culture and biopsy staining techniques, since we were able to detect fungal hyphae in all rabbit eyes 2 days after fungal inoculation, but at least 2–3 days had to elapse to determine any fungal growth on Sabouraud’s agar. Moreover, O’Day et al1 reported that about one fourth of fungal cultures became positive only after 2 weeks. Confocal microscopy is also rapid compared to biopsy staining, since to perform calcofluor staining some time had to elapse. As stated in our article2 “Although in our model Sabouraud’s agar and corneal biopsy techniques showed similar sensitivity (100%) in the early stage, confocal microscopy appears to have a definitive advantage in the later stages of infection, since not all cases of fungal keratitis could be cultured.” In the abstract we wrote that “on days 14 and 22 confocal microscopy was more sensitive than culture technique in both treated and untreated animals, since not all cases of fungal keratitis could be cultured.” We think the conclusion drawn is valid in light of the data provided in the study. In the second experiment, six rabbits were treated with topical fluconazole, seven were treated with oral fluconazole, and seven were left untreated. On day 14, we observed hyphal fragments (broken in treated corneas and full size in untreated ones) in each of 20 corneas by confocal microscopy. However, only eight of 20 scrapings grew Aspergillus fumigatus on Sabouraud’s agar culture. The difference between groups was statistically significant as is given in the text by utilising χ2 test. Similarly, on day 22 confocal microscopy revealed hyphal fragments in 14 corneas out of 20 (three in the topically treated, four in the orally treated, and seven in the untreated groups). At this stage only five corneal scrapings grew fungus on culture. The difference was statistically significant again as given in the article by utilising the χ2 test. Thus, superiority of confocal microscopy over culture technique on days 14 and 22 in treated and untreated rabbits was supported well by the data presented in the article.
In the result section, we were attempting to determine the efficacies of topical and oral fluconazole treatment by culture. However, p values were not correct as a result of an error. The errors escaped both our and the reviewer’s attention. However, this part of the results section does not contain any information that could affect any conclusion drawn as a result of study data. Actually, this part was not directly linked to the main aim of the study. The authors wish to thank to Dr Fan and coworkers for their careful attention.
The correct p values are given here: on day 14 (p = 0.383 and p = 0.296); on day 22 (p = 0.342 and p = 0.279).
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.