Ovarian cancer is a disease with a high mortality. Worldwide, approximately 200.000 women receive a diagnosis of ovarian cancer annually; of these 75% are at an advanced stage and 140.000 women die of this disease each year. Although response to primary treatment is high, most patients have recurrent disease and become resistant to treatment resulting in this high mortality.
The diagnostic evaluation of a woman with suspected ovarian cancer includes a physical examination, an ultrasonography, an abdominal computed tomography (CT-scan), and measurement of serum tumour marker CA-125 and CEA. Standard treatment of ovarian cancer consists of primary cytoreductive surgery followed by six courses of taxane- and platinum-based systemic chemotherapy. The goal of cytoreductive surgery is to resect all macroscopic tumour or at least to reduce the largest tumour residuals to less than a centimetre. When the diagnostic evaluation suggests that it is impossible to accomplish complete cytoreduction or when a patient is unable to sustain extensive surgery, than interval debulking surgery (IDS) could be an alternative. IDS implies three cycles of neoadjuvant chemotherapy (NACT) followed by cytoreductive surgery and another three courses of chemotherapy. However if at primary debulking it is not possible to remove all disease to at least residuals of less than one centimetre another laparotomy will be performed after three courses of chemotherapy.
The aim of this review was to investigate if laparoscopy is more accurate to diagnose extensiveness of disease than standard staging. If so, a useless and unnecessary primary laparotomy can be avoided and these patients can start immediately with neoadjuvant chemotherapy followed by interval surgery.
In total we identified 7 studies which reported on 6 cohorts of patients. In these studies 364 patients underwent a laparoscopy to evaluate the extensiveness of disease in the abdomen. All studies concluded that if at laparoscopy it was thought that removing macroscopic to at least residual disease of less than one centimetre was not feasible this was correct. However, even when performing a laparoscopy there are still patients primarily operated who have an unsuccessful debulking.
Quality of the evidence
A limitation of this review is that only two studies performed the laparoscopy and the laparotomy in all patients. The other studies only performed a laparotomy when at laparoscopy it was thought that no optimal result was feasible. Most studies suffer therefore from verification bias, which makes it impossible to draw conclusion on sensitivity of this test. Three studies develop or validated a prediction model including laparoscopy. Using a prediction model does not increase the sensitivity and will result in more unnecessarily explored patients.
Laparoscopy is a promising test, but the low number of studies and the differences between the included studies do not allow firm conclusions to be drawn from these data. Due to a difference in prevalence, there is a wide range in negative predictive values between studies. Two studies verified all patients. These imply a high specificity of laparoscopy in diagnosing resectability and have a good sensitivity. Both studies show that the use of criteria for unresectable disease will result in no patients inappropriately unexplored. However, there will still be patients undergoing unsuccessful primary laparotomy. Using a prediction model does not increase the sensitivity and will result in more unnecessarily explored patients, due to a lower specificity.
The presence of residual tumour after primary debulking surgery is the most important prognostic factor in patients with advanced ovarian cancer. In up to 60% of cases, residual tumour of more than 1 cm is left behind, stressing the necessity of accurately selecting those patients who should be treated with primary debulking surgery and those who should receive neoadjuvant chemotherapy instead.
To determine if performing an open laparoscopy after the diagnostic work-up of patients suspected of advanced ovarian cancer is accurate in predicting the resectability of disease.
We searched MEDLINE, EMBASE, The Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Register of Diagnostic Test Accuracy Studies, MEDION and ISI Web of Science to February 2013. Furthermore, we checked references of identified primary studies and review articles.
We included studies that evaluated the diagnostic accuracy of laparoscopy to determine the resectability of disease in patients who are suspected of advanced ovarian cancer and planned to receive primary debulking surgery.
Two review authors assessed the quality of included studies using QUADAS-2 and extracted data on study and patients' characteristics, index test, target condition and reference standard. Data for two-by-two tables were extracted and summarised graphically. Sensitivity and specificity and negative predictive values were calculated.
We included seven studies reporting on six cohorts. Between 27% to 64% of included patients per study were positive on laparoscopy (too extensive disease to warrant laparotomy) and between 36% to 73% were negative (disease suitable for debulking laparotomy). Only two studies avoided partial verification bias and provided data to calculate sensitivity and specificity, which did not justify meta-analysis. These two studies had a sensitivity of 0.70 (95% confidence interval (CI) 0.57 to 0.82) and 0.71 (95% CI 0.44 to 0.90); however, the specificity of both studies was 1.00 (95% CI 0.90 to 1.00). In these two studies there were no false positives, i.e. no patients for whom laparoscopy indicated that major surgery would not be successful and should be avoided, whereas, in reality the patient could be successfully operated upon. Negative predictive values (NPV), for those patients who were diagnosed with having not too extensive disease correctly identified were 0.75 (95% CI 0.55 to 0.86) and 0.96 (95% CI 0.56 to 0.99) due to a different prevalence. Although the studies did report sufficient data to calculate NPVs, we judged these estimates too heterogeneous to meta-analyse.
Three studies described the development or validation of a prediction model with a clear cut-off for test positivity. Sensitivity and specificity of these prediction models were 0.30 to 0.70 and 0.89 to 1.00, respectively. However, one of these studies suffered from partial verification bias.