Dear Editor

I read with interest the article by Zhang et al published in the September 2006 issue of the Journal of Clinical Pathology (J Clin Pathol 2006;59:958–964.). The authors wrote on fascin, an actin-binding protein, as a potential biomarker for early-stage oesophageal squamous cell carcinoma. There are a number of methodological issues with the research and paper that potentially weakens their findings and interpretation.

Firstly, the design of the study is suboptimal. The authors used two different patient groups for different aspects of the work. 102 archival materials from 2001 to 2003 were used for the immunohistochemical staining and specimens from 49 patients with oesophageal cancer operated upon in 2003 and 2004 were used for the western blot and real-time RT-PCR analyses. However, the results from the two different groups were presented and compared as if from the same study population. Methodologically, it would have been better to use the 2003/2004 patient cohorts for the whole research.

Immunohistochemical positivity was defined as cells showing >5% cytoplasmic staining in the methodology section. Later in the result section, the authors said they graded positivity as mild, moderate and severe but the readers are not told on what basis this grading was based. For the RT-PCR analysis, they used beta-actin as the internal control (i.e. house-keeping gene). Given that fascin is an actin binding protein, surely the potential for interaction between the target gene and internal control gene therefore exists.

Lastly, whilst the authors gave a good description of the statistical analysis carried out, I think they have employed wrong tests to analyse the data. Chi-squared test was used to analyse proportions but given that some of the data items were very small (<10) as presented in Table 2, Fischer’s exact test would have been more appropriate in those situations. Also, for the continuous data presented in Tables 3 and 4 they used student’s t-test to analyse the data, thereby assuming normality. However, looking more closely at the data, there is a strong suggestion that the data presented are not normally distributed with standard deviations as high (even higher) than the mean concentrations/levels of fascin ⁽¹⁾. This is actually not a surprise, as it is generally known in statistics that data obtained from ratios or concentrations are usually not normally distributed. Therefore, the authors should either have log transformed ⁽²⁾ their data before using student’s t-test or use a non-parametric test such as Mann-Whitney test for the statistical analysis.

The above methodological and statistical issues call into question the interpretation of the data generated and presented and therefore, weakens the conclusions reached. It would be interesting to hear the authors’ responses to the above points.

Yours truly

Dr Olorunda Rotimi MRCPath, PGDip Health Research

References

1. Altman DG, Bland JM. Detecting skewness from summary information. BMJ 1996;313:1200

2. Bland JM, Altman DG. Transforming data. BMJ 1996;312:770