Application of an automated DNA-imager in cervical cancer screening

Abstract

In cervical screening programmes, Papanicolaou test (Pap test) is the key screening tool. However Pap test is difficult to implement in low-resource region. Introduction of an economic, cost-effective and less skill demanding equipment is hence a potential direction of advance in cervical screening methodology.

Cervical carcinogenesis involves genetic instability which leads to chromosomal aneuploidy. Evaluation of aneuploidy may hence provide information for identifying cancer and precursor cells. An automated DNA-image-cytometry system (DNA-imager) capable of quantitating the DNA content of cells has recently been developed. To evaluate the efficacy of DNA-imager in cervical cancer screening, a total of 483 residual ThinPrep liquid-based cytology (LBC) samples after diagnosis were retrieved and evaluated by the DNA-imager. The high risk human papillomavirus (HPV) status of the atypical squamous cells of undetermined significance (ASC-US) samples has been tested as a parallel study. According to established criteria, 423 out of the 483 samples were satisfactory for downstream analysis. The samples were designated “Normal”, “Suspicious” or “Abnormal” according to their DNA aneuploidy and proliferation activity. Significantly more high grade lesion samples (HSIL and SCC) were designated as “Abnormal” by DNA-imager than were lower grade lesion samples (Negative, ASC-US, AGC, ASC-H, and LSIL) (94.19% vs 51.04%, p<0.0001). For detecting ≥HSIL, DNA-imager achieved high sensitivity and specificity (94.19% and 48.96%) using “Abnormal” as cut off. Adopting a more stringent definition of “Suspicious” or “Abnormal” would increase the sensitivity to 100% but decreased the specificity to 24.33%.

Regarding ASC-US triage, DNA-imager achieved a sensitivity and specificity of 80.00% and 71.29% respectively when “Abnormal” was used as test positive to predict cases with ≥HSIL follow-up in the next two year. The sensitivity increased but the specificity decreased to 90.00% and 34.65% respectively if “Suspicious or above” was used as indicator of undesirable follow-up. HR-HPV test, on the other hand, was able to identify all cases with ≥HSIL upon follow-up (sensitivity = 100%) but the specificity was only 15.84%. Among these ASC-US sample, test positivity of the two tests showed poor concordance with each other (Cohen’s κ = 0.062 and 0.074 respectively for “Suspicious or above” or “Abnormal”, respectively). Our findings suggested that DNA-imager may be a useful tool for automated primary screening of cervical cancer

3