PROTOCOL DESCRIPTION

The chemotherapy arm of the trimodal protocol consists of 6 cycles of chemotherapy – 2 or 3 neoadjuvant cycles and 4 or 3 adjuvant cycles. A neoadjuvant cycle consists of intra-arterial perfusion of cisplatin 100 mg/m ² on day 1 and intravenous doxorubicin 25 mg/m ² on days 1 to 3. To avoid direct brain or retina perfusion through an internal carotid artery access, the internal maxillary artery (IMA), a branch of the external carotid artery is cannulated. A second intra-arterial cycle is administered in 3 weeks. Three to 4 weeks after completing the second cycle, MRI or CT of the orbits is obtained to assess tumor response, which typically shows visible shrinkage of tumor volume, in some cases downstaging the tumor to a surgically resectable mass. However, if the posterior tumor margin shrinkage in the orbital apex is incomplete, rendering tumor resection margin clearance less likely, a third cycle is administered to achieve further cytoreduction. The decision to implement the third cycle is also guided by the patient’s tolerance or side effects to the initial cycles. After 2 (or possibly 3) intra-arterial chemotherapy cycles and following hematologic recovery, orbital exenteration is performed without delay to prevent tumor progression. Surgery is followed by adjuvant radiotherapy 2 to 4 weeks later with concomitant weekly IV cisplatin 20 mg/m ² for radiosensitization. Following chemoradiation therapy, 3 or 4 cycles of adjuvant IV cisplatin 100 mg/m ² on day 1 and intravenous doxorubicin 20 mg/m ² on days 1 to 3 every 3 weeks are administered, so patients would optimally receive a total of 6 cycles of systemic chemotherapy.

PARTICIPANTS

The clinical records of 19 consecutive patients with LGACC treated with neoadjuvant IACC were retrieved from the Bascom Palmer Eye Institute (BPEI) and University of Miami Hospital and Clinics. The study period extended from 1988 to 2021. The American Joint Committee on Cancer TNM Classification , 7 ^th edition for tumor staging was added. Seven patients were treated consecutively by conventional therapies between 1967 and 1984, with information gathered for the 2013 study remaining unchanged. This previously reported cohort of patients constituted the institutional comparator of the treatment group (Supplementary Figures 1-3). Disease-related mortality was defined as death due to LGACC. The 2 previously reported cases with alternate histology were excluded.

All the surviving participants from the 2013 study were contacted in January 2021 in a telephone interview regarding their overall medical status, with particular emphasis on disease-free survival. Referring ophthalmologists or medical oncologists were contacted for medical status confirmation, supplemental information relating to late sequelae of chemotherapy complications, or cause of death. The cause of death was confirmed by medical records and the primary treating physician. Information on locoregional recurrence, distant metastases, disease-free survival time, and medical status were updated.

STATISTICAL ANALYSES

The efficacy of treatments with Kaplan-Meier survival analyses for 3 endpoints was compared: mortality from all causes (overall survival), mortality due to LGACC with patients censored at death due to other reasons (disease-specific survival), and time to recurrence (progression-free survival). There is some controversy in the oncology literature regarding the most appropriate assessment of mortality: all-cause or disease-specific. This is a particular issue for small retrospective studies wherein 2 considerations come into play: (1) the uncertainty of obtaining accurate cause of death information from contacting relatives and (2) the reduction in statistical power associated with fewer endpoints in cause-specific mortality. For both reasons, all-cause mortality is preferred but disease-specific mortality is also reported. The survival time was calculated as the number of days from protocol initiation to the endpoint, last follow-up examination, or contact date. Statistical significance was assessed with the log-rank test. The reporting of progression-free survival and overall survival are well accepted standards for assessing cancer therapies in oncologic literature and they are universally utilized to compare data across studies. Instead of demonstrating overall survival advantages, especially for diseases with multiple lines of treatment or long-term survivorship, which may take years to confirm, progression-free survival has been used as a surrogate endpoint for which new treatments for malignant disease have been FDA approved. These endpoints are therefore reported here, consistent with clinical trial assessments.

Patients were divided into 2 groups receiving IACC: group 1 and group 2, depending on the integrity of the lacrimal artery. The latter group included 1 patient (case 8) with an intact lacrimal artery who received neoadjuvant therapy but deviated from protocol by refusing exenteration and adjuvant treatment. Survival was measured by specific cumulative proportions illustrated in the Kaplan-Meier curves, the benchmark statistic for interpreting survival. In addition to standard median follow-up (F/U) time-all for all patients, median F/U time-alive only for patients surviving at the end of the time of this report was also reported. Because LGACC mortality manifests only after many years, recently treated patients with short follow-up contribute little to treatment efficacy assessment. The median F/U time-alive is a gauge for identifying studies including patients with short follow-up.

Cumulative proportions were compared with the log-rank test of statistical significance. The risk ratio associated with positive tumor margins on each endpoint was studied with univariate Cox survival regression stratified by group and employing the likelihood ratio test for statistical significance.