Authorship:
David L Phelps*, Julia Balog †#, Mona El-Bahrawy*, Abigail VM Speller#, Sadaf Ghaem-Maghami*^ and Zoltan Takats#^
* Imperial College London, Hammersmith Campus / † Waters Research Center, Hungary / # Imperial College London, South Kensington / ^Co-corresponding authors
| Short Abstract Ovarian cancer is common and 5-year survival is 43.5%. Surgeons use intra-operative frozen section for tissue identification but this is time-consuming and expensive. Microscopic non-descript lesions during surgery, which may be cancer can be difficult to correctly identify. The near-real-time tissue identification abilities of the iKnife were tested in this study. We have shown for the first time that ovarian, peritoneal and fallopian tube tissues have unique REIMS spectral signatures, which can be used to accurately classify tissue histopathology. Leave-one out patient cross-validation resulted in 100% sensitivity and 100% specificity in the separation of normal and cancerous ovary (n=189). |
Long Abstract
Introduction
Ovarian cancer is common with a 2% lifetime risk of developing the disease(1). 95% are epithelial adenocarcinoma and this group of malignant tumours includes: high grade serous, endometrioid, clear cell and mucinous, with high grade serous the commonest variant by far. Many clinicians and scientists now regard cancers of the fallopian tubes, peritoneum and ovaries similarly, as they present and metastasise alike and are clinically managed in much the same way(2).
Standard treatment for ovarian cancer is surgical debulking followed by chemotherapy. However, some centres treat with neo-adjuvant chemotherapy followed by delayed primary debulking. This approach renders tumour deposits difficult to identify during surgery due to fibrosis after chemotherapy. It is universally accepted that debulking to no macroscopic residual disease yields the best prognosis. In a large study overall survival for stage IV disease was 54.6 months in women with no macroscopic residual disease, compared to 23.9 months with >10mm residual disease(3). Radical surgery is not without risk and morbidity is higher for these patients.
Often the histopathology is not known before surgery. Surgeons rely upon intra-operative frozen sections which involves sending a small piece of the tumour to the histopathologist for microscopic analysis. This process is time consuming, open to inter-observer variability and is expensive. The surgeon therefore relies upon experience to guide how radical the resection should be. The lack of confidence in determining the nature of non-descript lesions can lead to more radical (perhaps unnecessary) surgery. The near-real-time tissue identification abilities of the REIMS intelligent knife (iKnife) lends itself very well to this challenge.
This research serves as a proof of principle study to test the iKnife’s ability to accurately diagnose ovarian, fallopian and peritoneal histopathology.
Methods
All samples were stored frozen tissue. A Xevo G2-S Q-TOF mass spectrometer was used coupled with a modified handheld monopolar surgical diathermy, and a Force Triad Covidien generator. The technology analyses the surgical diathermy smoke for ionic species by extracting the smoke from the surgical field through PTFE tubing attached to the hand-piece. The smoke is passed across to the atmospheric inlet of the mass spectrometer using a Venturi air jet pump driven by medical air or nitrogen gas. Isopropyl-alcohol (IPA) with 0.2 ml/min rate is administered directly to the inlet capillary in order to facilitate ionisation. In total 146 samples (130 separate patients) of normal ovary, fallopian tube and peritoneum, plus malignant and borderline tumours of the ovary were processed. The processed fragment of remaining tissue was formalin fixed, catalogued and sent to histopathology for confirmation of histopathological diagnosis. Ethical approval was granted through the Research Ethics Committee.
The spectra and histopathological diagnosis, are used to populate a home-built SQL database. Background subtraction of the spectra is performed within this software in addition to lock mass correction and averaging of the peak intensities from multiple spectra within each burn within the m/z 600-900 range. Tissue groups are then selected for statistical analysis using principal component analysis (PCA) plots and linear discriminant analysis (LDA). To ensure the highest accuracy, each sample is updated in the database with the post processing histopathological diagnosis. If no tumour is present in a sample expected to have tumour, it is reclassified as normal.
Results
The goal of this study was to generate a proof of concept dataset and preliminary database, followed by in vivo sampling within the theatre during surgery. The POC data included a total number of 486 sampling points taken from the 146 tissue bank samples. The acquired spectra featured multiple species, including triglycerides (TGs) in the m/z range 850-1000, fatty acids (FA) in the 150-400 mass/charge region and phospholipids (PL) in the 600-900 m/z region. PLs featured a unique “fingerprint” in different normal and benign tissues and different type and grade of cancers. Multivariate models were created in order to show the separation between normal and cancerous tissue of different type and grade. The leave-one out patient cross-validation resulted in 100% sensitivity and 100% specificity in the separation of normal and cancerous ovary (n=189). The sensitivity and specificity didn’t change when including normal fallopian tube and peritoneum to the model (n=291), the overall correct classification rate was 92.7%. Our algorithm separated normal ovary, benign lesions and borderline ovarian cancers (n=146) well on the LDA plot. Our next step is moving on to fresh ex vivo samples and to intraoperative use of the instrument during tumour debulking surgeries.
Our data has proved that iKnife is capable of separating normal ovary, fallopian tube, peritoneum, benign lesions and different type and grade of cancerous ovarian tissue. This and further data will be used to populate a spectral reference library created for intraoperative tissue classification. In gynaecology surgery our novel iKnife method has the potential to significantly cut operating times by mitigating the need for frozen sections and may reduce margin status and improve identification of non-descript lesions. The potential for identification of necrotic, inflamed tissue during delayed debulking surgery after neo-adjuvant chemotherapy is a very exciting prospect, as it could facilitate on-table decision making resulting in improved survival rates and patient outcomes.
References
(1)NICE. The recognition and initial management of Ovarian Cancer, clinical guideline 122. NICE; 2011.
(2)Mutch DG, Prat J. 2014 FIGO staging for ovarian, fallopian tube and peritoneal cancer. Gynecologic oncology. 2014;133(3):401-4.
(3)Wimberger P. Influence of Residual Tumor on Outcome in Ovarian Cancer Patients With FIGO Stage IV Disease. Annals of Surgical Oncology. 2010:964-9.