Abstract： Objective To observe the missed diagnosis of invasive carcinoma under the microscope (ICUM) in high grade squamous intraepithelial neoplasia (HSIL), and analyze associated factors influencing missed ICUM. Methods A retrospective study was performed on patients diagnosed with HSIL by colposcopy-guided biopsy and treated with loop electrosurgical excision procedure(LEEP)at the First Affiliated Hospital of Nanjing Medical University, from December 2014 to December 2016. They were non-pregnant, ≤50 years old and the cervical volume without obvious enlargement and exogenous surface without and ulcerative lesions. A total of 283 cases with early cervical cytology results, never received cervical traumatic treatment or cervical biopsy in another hospital before, and their colposcopic images were clear enough to reevaluate. The ultimate pathological diagnosis was based on the higher-level pathological diagnosis between the results of cervical biopsy and LEEP to evaluate ICUM missed in HSIL and the risk factors. Results (1) Among the 283 cases with HSIL diagnosed by colposcopy-directed biopsy,44 cases (15.5％,44/283) were missed diagnosis of ICUM, which consisted of 29 cases Ⅰa1, 4 cases Ⅰa2 and 11 cases Ⅰb1 in the ultimate pathology.(2)Analysis of associated factors for missed ICUM:univariate analysis showed that,as the age increased, the risk of missed ICUM also increased(the rates of missed diagnosis for<30, 30-39, 40-50 years were 7.7％, 11.5％, 22.0％, respectively;χ2=6.254, P=0.012 by trend test). The more the number of high-grade features, the higher risks(the rates of missed diagnosis for 1, 2, 3, 4 high-grade features were 10.2％, 17.6％, 23.8％, 30.8％, respectively;χ2=7.686,P=0.006 by trend test). The locations of HSIL were only endocervical, only ectocervical and mixed, the risk increased by this sequence(2.8％, 5.1％, 28.7％; χ2=26.193,P<0.01 by trend test). The rate of missed diagnosis for not completely visible squamocolumnar junction(SCJ)was higher than that of the completely visible one(22.3％ vs 2.1％;χ2=19.680, P<0.01). The rate of missed diagnosis was higher for existing atypical vessels than those without(60.7％ vs 10.6％;χ2=48.279, P<0.01). The rate of missed diagnosis for visible lesion size≥40 mm2 was higher than that of<40 mm2(27.3％vs 4.2％;χ2=28.921, P<0.01). The rate of missed diagnosis for the proportion of visible lesion size in ectocervical size ≥0.75 was higher than that of <0.75 (83.3％ vs 14.1％;P<0.01). The rate of missed diagnosis for the maximum linear length of visible lesion≥10 mm was higher than that of<10 mm(46.9％vs 9.0％;χ2=44.473, P<0.01). But the different severity of cervical cytology before colposcopy was not associated with missed ICUM(P>0.05). Multivariable analysis found that visibility of SCJ, atypical vessels, visible lesion size and maximum linear length of visible lesion were associated with missed diagnosis of ICUM(all P<0.05). Conclusions The diagnostic value of HSIL by colposcopy is limited. Meanwhile, for the patients who are ≤50 years old with HSIL diagnosed by cervical biopsy, invisibility of SCJ, atypical vessels, visible lesion size and maximum linear length of visible lesion evaluated by colposcopy are the independent risk factors of missed ICUM. Thereby, it is necessary to take active intervention for HSIL with these risk factors.