A CRITICAL EVALUATION OF THE HAZEWINKEL CRITERIA FOR THE OPTICAL DIAGNOSIS OF SESSILE SERRATED LESIONS (SSL) AT THE BEGINNING OF A LEARNING PROCESS

Aims:

1. To describe how the main characteristics of the Hazewinkel criteria for SS are identified by a group of non-experienced endoscopists;

2. To identify which combination of characteristics identifies with more reliability a SSL in a learning background.

Methods:

Prospective study in the setting of a population-based CRC screening program. Six endoscopists attended a short session on optical diagnosis of SSL. For every lesion all endoscopists described the presence of the Hazewinkel criteria (cloud appearance, irregular shape, indistinct limits and black dots) and categorized lesions following the NICE classification. The presence of ≥2 criteria was considered diagnostic of SSL.

Results:

A total of 2505 lesions were included. Among them, 116 (4.6%) SSL were identified [median size (SD) 4 (6.2); proximal location 68 (58.6%)]. Accuracy of ≥2 criteria for the diagnosis of SSP was 0.93. Overall positive predictive value (PPV) was 0.25 without differences among endoscopists, while the NICE PPV for adenoma was 0.84. The frequency of identification of each criterion in every SSLs was: cloud-like surface 45 (38.8%), irregular shape 27 (23.3%), indistinctive borders 30 (25.9%) and black dots 14 (12.1%). All criteria were more prevalent in SSL > 10 mm. The proportion of SSL diagnosed in lesions harboring each criteria combination is summarized in Tab. 1.

Tab. 1:

Proportion of SSL diagnosed in lesions harboring each criteria combination

	NICE1	NICE2	Size 1 – 10 mm	Size > 10 mm
Cloud-like +irregular shape	28.6	12.5	5.6	28.6
Cloud-like + indistictive borders	50.0	-	36.4	100
Cloud-like+black dots	25.0	-	20.0	-
Cloud-like+irregular shape+indictinctive borders	35.3	25.0	22.9	57.1

Conclusions:

1. A great proportion of SSL does not have the optical diagnosis criteria;

2. The identification of the Hazewinkel criteria improves with size and NICE1 lesions.

3. Cloud-like surface is the most prevalent characteristic found in SSL.