Selective cervical cytology screening: discriminant analysis approach.

Cervical cancer is one of the leading malignancies seen in Indian women. It has been well established that organized cervical cytology screening program is the mainstay for control of cervical cancer. It is not possible to carry out cervical cytology screening for masses in India due to paucity of human and financial resources. Hence there is a need for development of an alternate strategy to concentrate on women with high risk. In the present communication attempt was made to define a high risk group based on sociodemographic factors, viz. age, parity, education and clinical features. A total of 67,000 women were screened of which in 250 malignancy was detected. The rate of malignancy was observed to be high in women above 40 years (10.5/1000) with more than two children (6.1/1000) and in illiterate group (4.9/1000) as compared to women below 40 years, more than 3 children and illiterate group. Similarly, the rates were higher in women with clinical diagnosis of cervical erosion which bled on touch, unhealthy cervix and suspicious looking cervix, malignancy rates were 17.1, 24.7, 263.2 (per 1000), respectively. An attempt was made to study the combined effect of all the six factors (sociodemographic and clinical) by employing the technique of linear discriminant analysis to find out the discrimination power between the normal and malignant women. Discriminant score thus obtained would help to classify the case for subjecting to cervical cytology. It was observed that the model containing sociodemographic and clinical variables was able to classify 69% of malignant cases correctly. When the clinical variables were dropped from the model, the sensitivity dropped to 65%. The above exercise indicated that based on the discriminant score even in the absence of facilities for clinical examination of women, it may be possible to identify women of high risk group for subjecting them to cervical cytology screening.