# br of the number of FRa

100) of the number of FRa + ovarian cancer lesions confirmed by both fluorescent light and the pathology and/or IHC (TP) over the num-ber of FRa + ovarian cancer lesions confirmed by fluorescent light (TP + FP, where FP=False Positive). Healthy (non-cancerous) tissue was not removed as part of this study, and therefore, a true negative (TN) sample was not available.

Since the correlation was unknown for the ovarian tumor lesions from the same subject, 2 different statistical models were used, sepa-rately for sensitivity and PPV [10]:

• Proc Glimmix in SAS for a binomial distribution with logit link func-tion, without a random effect for patient and assuming no correlation among multiple lesions in a single patient, a Generalized Linear Model (GLM).

• Proc Glimmix in SAS for a binomial distribution with logit link func-tion, with a random effect for the patient allowing for the possibility of some correlation among lesions within a single patient, Generalized Linear Mixed Model (GLMM).

At least 135 FRα + lesions were required to have an 80% chance that the lower boundary of the boundary 95% confidence interval for sensi-tivity would be above 85% assuming the “true” sensitivity is 92% or more. Assuming 7 individual tumor lesions per patient, it Oxidopamine hydrochloride was estimated that 20 patients with FRa + ovarian cancer would provide 140 FRa + le-esions for excision and testing. The sample size calculations assumed le-sions within each patient are uncorrelated. All statistical analyses were performed using Statistical Analysis Software SAS® (SAS Institute Inc., Cary, NC, version 9.4.)

+ ovarian cancer lesion, 1 did not undergo fluorescence imaging, 4 did not receive study drug, and 7 were enrolled only in the safety pop-
ulation after the efficacy endpoint had been met.

Table 1 shows the demographic and oncologic characteristics of the study cohort. The mean age of the women who participated was

64 years, ranging from 37 years to 82 years. Patients were primarily white (79.5%), and the median BMI was 25.3 g/m2. Most patients had In-ternational Federation of Gynecology and Obstetrics (FIGO) stage ≥ III ovarian cancer (70.3%) of serous histology (61.4%) with a typical perito-neal disease distribution.

The total length of time patients were exposed to fluorescence imag-ing pre-resection ranged from 2 to 23 min. The maximum duration of exposure to fluorescence imaging, which included any post-resection imaging, was 46 min. No patient discontinued from the study due to an adverse event related to the imaging system.

3.1. Treatment efficacy in the mITT population

A total 225 lesions were obtained from the 29 patients, comprising the mITT efficacy population. 171 lesions were positive by both OTL38 NIR and FRa and were counted as true positives (Table 2). There were 23 lesions that fluoresced but did not test positive for both FRa and ovar-ian cancer, which were counted as false positives. Lesions that did not fluoresce but were FRa and ovarian cancer positive equaled 28 false neg-atives, and there were 3 true negative lesions that did not fluoresce and were negative for both FRa and ovarian cancer. The GLM analysis, as-suming no correlation of lesions within patient (excludes a random ef-fect for patient), resulted in an estimated sensitivity of 85.93%, with a 95% lower boundary CI = 81.19. The estimate for PPV was 88.14% with a 95% lower boundary CI = 83.59. The GLMM analysis that ac-counts for possible correlation of lesions within the patient estimated a sensitivity of 97.97%, with a 95% lower boundaryCI = 87.75. The

Table 1

Demographic and tumor characteristics of the safety population (n = 44).

Characteristic
Intent to treat (n = 44)

Age (years)