Sample size requirement for digital image analysis of collagen proportionate area in cirrhotic livers.

Aims:    The requirements for adequate cirrhotic liver biopsy size have not been established for quantitative fibrosis measurements (collagen proportionate area: CPA). We evaluated the CPA of virtual biopsies in cirrhosis to elucidate (i) the amount of tissue required to achieve reliable CPA measurements and (ii) the effect of aetiology on sample size requirements.

Method And Results:    A total of 120 cirrhotic tissue blocks (six aetiologies) were studied. A representative 100 mm(2) region was selected from each block and a reference CPA measured. Each image (n = 120) was divided into 100 × 1 mm(2) images; CPA was measured for each 1 mm(2) and virtual biopsies of different sizes were created from the 1 mm(2) components. For each virtual biopsy size the probability that the virtual biopsy CPA would be within 5% of the reference CPA was calculated. There were 441 000 virtual biopsies. Biopsy size versus probability plots indicated that, for 90% probability that the virtual biopsy CPA can be expected to be within 5% of the reference CPA, 22-28 mm(2) of analysable tissue is required depending on liver disease aetiology; and that a 75% probability level requires a biopsy with 12-15 mm(2) of analysable tissue.

Conclusion:    The sample size required for a given probability level depends on the aetiology of cirrhosis, and this should be taken into account when judging the reliability of cirrhotic liver biopsy CPA.