According to results of a study published online in Annals of Surgical Oncology, experienced cancer surgeons “demonstrated a clinically unacceptable rate of missed peritoneal metastases” when using lesion appearance to distinguish benign peritoneal lesions from metastases during laparoscopy for gastrointestinal cancers.
Although the identification of peritoneal metastases is a critically important component of operative staging in gastrointestinal cancer, it is complicated by the common presence of benign peritoneal lesions, even in patients without cancer. In many cases, an assessment of the gross physical characteristics of a peritoneal lesion, performed during surgery, is used to make determinations regarding whether to perform a biopsy.
This study involved 35 patients with biopsy-proven gastrointestinal cancer, excluding esophageal, hepatic, and colorectal cancers, who underwent a video-recorded laparoscopic operation performed by a single surgeon at the Lahey Hospital and Medical Center in Burlington, Massachusetts, between January 1, 2014 and September 30, 2017.
Two deidentified, still images were prepared for each peritoneal lesion included in the study (ie, 1 image showed a clear view of the lesion while the other was an image of the lesion next to a surgical instrument). Each of the lesions was biopsied during the operative procedure. Ten experienced gastrointestinal oncologic surgeons who were not informed that biopsies of the lesions had been performed classified the lesions as either benign or metastatic based on the appearance of the images.
In addition, 3 experienced investigators blinded to biopsy results subjectively characterized the appearance of the imaged lesions, using digital image-processing software ImageJ, for evaluations of lesion size, color, and gradient magnitude (ie, changes in luminosity at the edge of each lesion). The lesions were also evaluated using a deep neural network, machine-learning model.
Of the 87 lesions included in the study, 82% and 18% arose from the parietal peritoneum and visceral peritoneum, respectively; at final pathologic examination, 68% and 32% of lesions were determined to be benign or metastatic, respectively.
In the group of experienced gastrointestinal oncologic surgeons, 36% of peritoneal metastases were misclassified as benign.
Based on subjective evaluations of the lesion images made by the experienced investigators, multivariate analyses identified 3 independent predictors of the metastatic nature of a lesion (degree of nodularity, border transition, and degree of transparency [P <.03 for all variables]).
Respective area under the receiver operating characteristics curves (for which values of 1 and 0.5 would indicate 100% discrimination and no discrimination between benign and metastatic lesions, respectively) were 0.82 (95% CI, 0.72-0.91) for the multivariate model incorporating subjective assessments of lesion characteristics, 0.66 (95% CI, 0.54-0.78) for evaluations of the gradient magnitude of a lesion using ImageJ software, and only 0.47 (95% CI, 0.38-0.57) for use of the machine-learning model.
“The implications from the findings of this study for current practice are that all visible peritoneal nodules/lesions need to be biopsied in the setting of a staging laparoscopy,” the study authors noted in conclusion.
- Schnelldorfer T, Ware MP, Liu LP, et al. Can we accurately identify peritoneal metastases based on their appearance? An assessment of the current practice of intraoperative gastrointestinal cancer staging [published online March 25, 2019]. Ann Surg Oncol. doi: 10.1245/s10434-019-07292-0