OBJECTIVE: To determine whether providing previous cytology and histology findings alters the accuracy of conventional cervical cytology reading or changes reading times. STUDY DESIGN: Each of 9 cytologists read 9 batches of 8 routinely referred Pap smears (total, 648 slides), with history (H) and without history (NH), at an interval of no less than 4 weeks. Each batch was read blind to the result of reading under the other strategy and to histology. Histologic cervical intraepithelial neoplasia 2 or more severe was the reference standard. Accuracy of reading was assessed across all thresholds using receiver operating characteristic (ROC) curves and by sensitivity and specificity at a cytology threshold of possible low grade squamous intraepithelial lesion (consistent with atypical squamous cells of undermined significance). RESULTS: Areas under the ROC curve, sensitivities and specificities were similar if read with or without history, except for 1 reader for whom reading with history increased the area under the ROC curve from 0.716 to 0.833 (increase of 0.117, p = 0.017) and the sensitivity from 0.57 to 0.79 (increase of 0.22, p = 0.014), without any significant change in specificity. Accuracy varied between subgroups defined by age and by the severity and timing of previous abnormalities, but the results of the comparison of accuracy in H and NH did not vary by subgroup. Mean reading times were 8.2 (H) and 7.9 (NH) minutes per slide, a difference of 0.34 minutes (p = 0.083). Differences in mean batch times (H-NH) between readers ranged from -0.08 to 1.0 minutes, the largest difference being for the reader whose accuracy increased. CONCLUSION: An accurate history might improve accuracy for some cytologists.
OBJECTIVE: To determine whether providing previous cytology and histology findings alters the accuracy of conventional cervical cytology reading or changes reading times. STUDY DESIGN: Each of 9 cytologists read 9 batches of 8 routinely referred Pap smears (total, 648 slides), with history (H) and without history (NH), at an interval of no less than 4 weeks. Each batch was read blind to the result of reading under the other strategy and to histology. Histologic cervical intraepithelial neoplasia 2 or more severe was the reference standard. Accuracy of reading was assessed across all thresholds using receiver operating characteristic (ROC) curves and by sensitivity and specificity at a cytology threshold of possible low grade squamous intraepithelial lesion (consistent with atypical squamous cells of undermined significance). RESULTS: Areas under the ROC curve, sensitivities and specificities were similar if read with or without history, except for 1 reader for whom reading with history increased the area under the ROC curve from 0.716 to 0.833 (increase of 0.117, p = 0.017) and the sensitivity from 0.57 to 0.79 (increase of 0.22, p = 0.014), without any significant change in specificity. Accuracy varied between subgroups defined by age and by the severity and timing of previous abnormalities, but the results of the comparison of accuracy in H and NH did not vary by subgroup. Mean reading times were 8.2 (H) and 7.9 (NH) minutes per slide, a difference of 0.34 minutes (p = 0.083). Differences in mean batch times (H-NH) between readers ranged from -0.08 to 1.0 minutes, the largest difference being for the reader whose accuracy increased. CONCLUSION: An accurate history might improve accuracy for some cytologists.