Cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) are complex procedures with a very steep a learning curve (LC). This study evaluates the LC of CRS and HIPEC in a single-center experience of peritoneal surface malignancies (PSMs). Approximately 140 to 150 cases were necessary for the acquisition of competence in CRS and HIPEC with adequate radicality and acceptable safety. Eighty to 100 cases were necessary to assure short-term prognostic gains in rare PSMs. This article highlights how LC and continuous monitoring of surgical performance is critical in evaluating the credibility of emerging and already established PSM centers.
- •
Cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) is a complex procedure with a very steep learning curve (LC).
- •
Sequential probability ratio test with risk adjustment was used to evaluate the LC of CRS and HIPEC in a single-center experience of peritoneal surface malignancies (PSM).
- •
Approximately 140 to 150 cases were necessary for the acquisition of competence in CRS and HIPEC with adequate radicality and acceptable safety. Eighty to 100 cases were necessary to assure short-term prognostic gains in rare PSMs.
- •
This study highlights how LC and the continuous monitoring of surgical performance is critical in evaluating the credibility of emerging and already established PSM centers in the world.
Introduction
The advent of the cytoreductive surgery (CRS) associated with hyperthermic intraperitoneal chemotherapy (HIPEC) in the early 1990s has dramatically changed the treatment of peritoneal surface malignancies (PSM). The combined treatment has been suggested as the standard of care for pseudomyxoma peritonei (PMP), peritoneal mesothelioma (PM), and peritoneal carcinomatosis arising from colorectal cancers.
The concept of proficiency in surgical procedures is multidisciplinary and encompasses a wide range of perioperative aspects. The expertise comprises variables related to indication of the treatment, surgical technique, management of HIPEC parameters, and short-term surgical and oncologic results. The few studies published thus far on learning curve of CRS and HIPEC have reported a steep and long-lasting process until the acquisition of sufficient competence.
Although originally developed for use in quality control studies in the realm of manufacturing, the sequential probability ratio test (SPRT) has been largely used in medicine to monitor the safety of medical interventions such as interventional cardiology, cardiac surgery, emergency medical services, and other procedures It offers an advantage by allowing formal hypothesis testing. By providing a graphic summary of changes in performance with time, SPRT can detect deteriorated or improved performance. Therefore, it is helpful to analyze the learning curve of surgical procedures.
The authors recently conducted an analysis on learning using this statistic for the CRS and HIPEC in a series of 420 cases of PSM treated in the Istituto Nazionale dei Tumori (NCI), Milan, Italy. One hundred forty-nine and 137 cases were necessary to assure acceptable reductions of the risk of severe morbidity and incomplete cytoreduction.
The present study attempted the following:
- 1.
To reevaluate the learning curve of the same surgical team in CRS and HIPEC in an expanded case mix. As in the previous study, changes in short-term surgical outcomes in a case sequence composed of a variety of PSMs were reassessed. Evolution of short-term oncologic outcomes were evaluated separately in subsets of pseudomyxoma peritonei (PMP) and peritoneal mesotheliomas (PM).
- 2.
To monitor the surgical performance of the same team in terms of G3-5 morbidity along the timeline after the acquisition of expertise (surveillance with audit intent).
- 3.
To discuss the utility and implications of the present results for emerging PSM programs and already active ones.
Patients and methods
All patients were treated under an institutionally approved protocol and provided written informed consent. The eligibility requirements for treatment were as follows: histologically confirmed diagnosis of peritoneal surface malignancy judged resectable clinically and radiologically; age younger than 75 years; no distant metastasis; adequate renal, hematopoietic, and liver functions; and performance status according to Eastern Cooperative Oncology Group (ECOG) classification.
Data were extracted from a prospectively collected institutional database of the PSM program of the NCI in Milan. In total, 462 CRS and HIPEC procedures performed from August 1995 to February 2012 were included in the study.
CRS and HIPEC
The technique of CRS has been described elsewhere. In brief, the goal of the surgical cytoreduction was to remove all visible tumor with a diameter greater than 2.5 mm by means of diaphragmatic, parietal anterior, and pelvic peritonectomy with greater and lesser omentectomy. Depending on disease involvement, multiorgan resections were performed, including removal of gallbladder, spleen, left/right or total colon, and uterus with annexes.
Peritoneal carcinomatosis was quantified according to the peritoneal cancer index (PCI). The mean PCI was 22 (range 0–39). Completeness of cytoreduction was classified as follows : CC-0, no residual disease (RD); CC-1, RD of 0 to 2.5 mm; CC-2, RD of 2.5 mm to 2.5 cm; and CC-3, RD greater than 2.5 cm. Incomplete cytoreduction was defined as RD greater than 2.5 mm (CC-2/3).
HIPEC was performed with a closed-abdomen technique using an extracorporeal device (Performer LRT; RAND, Medolla [MO], Italy), which maintained the intra-abdominal temperature at 42° to 43°C. Chemotherapy schedules were as follows: cisplatin (CDDP; 25 mg/m 2 /L) and mitomycin C (MMC; 3.3 mg/m 2 /L) for colorectal cancer, gastric cancers, and pseudomyxoma peritonei; and CDDP (43 mg/L of perfusate) and doxorubicin (Dx; 15.25 mg/L of perfusate) for ovarian cancer and peritoneal mesothelioma. Patients older than 70 years and those who had undergone previous chemotherapy received a 30% dose reduction of both drugs. The same surgical team guided by the same principal operator (D.M.) operated on all of the patients.
In the postoperative period, patients were assisted in an intensive care unit for at least 48 hours. In the successive in-hospital stay, patients were cared for by a multidisciplinary team comprising internists with a nutritional science specialization, anesthetists, rehabilitation professionals, and psychologists.
Study Parameters
Adverse events (morbidity, anastomotic leak, systemic toxicity) were graded according to the NCI CTCAE v3 criteria. When more than 2 complications occurred during the same period in 1 patient, the complication was considered as more serious for analysis. Anastomotic leak was defined according to the criteria established elsewhere. The procedure-related mortality (PRM) was defined as death occurring during the in-hospital stay after CRS and HIPEC. G3-5 morbidity was defined as the combination of surgical and medical G3-5 morbidities, PRM, and G3-5 systemic toxicity. To assess surgical outcomes according to case sequence, they were divided into 6 subgroups of 70 for cases 1 to 420, while cases 421 to 462 were designated as another subgroup.
The outcomes of interest were rates of incomplete cytoreduction, G3-5 morbidity, and short-term oncologic failure. Potential confounders to the outcomes of interest were identified among the following clinicopathological parameters: age, sex, ECOG performance status, Charlson comorbidity index (combined condition and age-related score), previous systemic chemotherapy, previous surgical score (PSS), body mass index (BMI), preoperative serum albumin, tumor histotype, PMP histologic subtype (disseminated peritoneal adenomucinosis [DPAM], peritoneal mucinous carcinomatosis [PMCA]), PM subtype (biphasic/sarcomatoid vs others), PCI, bowel anastomosis, number of peritonectomy procedures, operating time, CDDP dose for HIPEC, subgroups of case sequence, incomplete cytoreduction rate, and HIPEC drug schedule.
The risk modeling and elaboration of SPRT curves for oncologic outcomes were done separately for PMP and PM. Short-term oncologic failure was defined as emergence of recurrence and/or death within 1 year from the combined procedure for PMP patients and within 6 months for PM patients. Therefore, PMP cases with less than 1 year of follow-up and PM cases with less than 6 months of follow-up that did not present recurrence and/or death were excluded.
Statistical Analysis
One-way analysis of variance (ANOVA) and the χ 2 test were used to analyze changes of continuous and discrete variables, respectively, according to subgroups of case sequence.
The risk-adjusted SPRT plot was used to chart, across the case sequence, changes in the rates of incomplete cytoreduction, rates of G3-5 morbidity, and short-term oncologic failure. To elaborate the SPRT, 4 parameters were defined: estimated probabilities of outcomes of interest for each case, a prespecified odds ratio (OR) for the outcomes of interest, and type I and II error rates. Probability of type I and type II (α and β) error were both set at 0.05. From these, 2 control limits (h0 and h1) and the cumulative sum of log-likelihood ratio with risk adjustment were calculated according to equations outlined by Rogers and colleagues.
The risk predictions for incomplete cytoreduction, G3-5 morbidity, and short-term oncologic failures were elaborated by multivariate analysis, using the logistic regression model. Clinical factors were selected as covariates when P values were less than .15 on univariate analysis. Using backward stepwise selection, independent factors were identified. The probability of event was calculated, which was adjusted with independent risk factors of incomplete cytoreduction, G3-5 morbidity, or short-term oncologic failures from the logistic regression model. Discrimination was measured by the area under the receiver-operating characteristic curve (AUC), with values of 0.5 representing no discrimination and 1.0 representing perfect discrimination. Model fitness was assessed by the Hosmer-Lemeshow goodness-of-fit test, with P values of greater than .05 indicating acceptable fit.
When creating the RA-SPRT curve, each case was plotted in sequence along the x-axis. When a success occurred (no G3-5 morbidity, optimal cytoreduction, or short-term oncologic success) a log-likelihood ratio with risk adjustment ( s ) was subtracted from the cumulative score. When a failure occurred, the constant 1 − s was added to the cumulative score. Thus, an ascending slope in the RA-SPRT line indicated deterioration of performance, whereas a descending slope indicated improvement of performance.
An unacceptable OR was set for increase in rates of incomplete cytoreduction and G3-5 morbidity, respectively, at 1.8 and 1.4, adopting the same boundary lines of the authors’ previous experience of LC analysis. In brief, these figures were obtained considering the highest and lowest rates of these outcomes available in the literature. The upper and lower control limits were respectively defined as h1 (reject line) and h0 (accept line). OR was prespecified as 2 for short-term oncologic failure, for both PMP and PM.
If the RA-SPRT curve crossed the upper decision limit (h1) from below, this meant that the actual OR for outcome was equal to or higher than the prespecified OR with the probability of type I error of 0.05. If the line crossed the lower decision limit (h0 or accept line) from above, this indicated that the actual OR for the outcome being studied is less than the unacceptable OR with the probability of type II error of 0.05. When the line was between h0 and h1, no statistical inference could be made. Expertise was estimated to be achieved at point(s) in which the curve of RA-SPRT crossed the accept line (h0) for outcomes of interest. At the time of crossing to the lower boundary, the graph was reset to 0 to start the surveillance of surgical performance with audit intent.
Once the expertise is acquired, the surveillance continues to monitor any eventual deterioration of the performance. The RA-SPRT curve could assume a positive slope in the case of enrollment of new surgeons in the surgical team, launch of new protocols using more toxic or more effective treatments, and the introduction of changes or innovations in surgical strategy based on new data obtained from other centers.
Evolution of predicted risk of incomplete cytoreduction and G3-5 morbidity along the case sequence was assessed with the moving-average (MA) method. An MA order of 10 was used.
The applied statistical software was SPSS 18.0 (SPSS Inc, Chicago, IL), while the MA method and RA-SPRT model were calculated by Excel version 2003 (Microsoft Corporation, Redmond, WA). Statistical significance was set at P <.05.
Patients and methods
All patients were treated under an institutionally approved protocol and provided written informed consent. The eligibility requirements for treatment were as follows: histologically confirmed diagnosis of peritoneal surface malignancy judged resectable clinically and radiologically; age younger than 75 years; no distant metastasis; adequate renal, hematopoietic, and liver functions; and performance status according to Eastern Cooperative Oncology Group (ECOG) classification.
Data were extracted from a prospectively collected institutional database of the PSM program of the NCI in Milan. In total, 462 CRS and HIPEC procedures performed from August 1995 to February 2012 were included in the study.
CRS and HIPEC
The technique of CRS has been described elsewhere. In brief, the goal of the surgical cytoreduction was to remove all visible tumor with a diameter greater than 2.5 mm by means of diaphragmatic, parietal anterior, and pelvic peritonectomy with greater and lesser omentectomy. Depending on disease involvement, multiorgan resections were performed, including removal of gallbladder, spleen, left/right or total colon, and uterus with annexes.
Peritoneal carcinomatosis was quantified according to the peritoneal cancer index (PCI). The mean PCI was 22 (range 0–39). Completeness of cytoreduction was classified as follows : CC-0, no residual disease (RD); CC-1, RD of 0 to 2.5 mm; CC-2, RD of 2.5 mm to 2.5 cm; and CC-3, RD greater than 2.5 cm. Incomplete cytoreduction was defined as RD greater than 2.5 mm (CC-2/3).
HIPEC was performed with a closed-abdomen technique using an extracorporeal device (Performer LRT; RAND, Medolla [MO], Italy), which maintained the intra-abdominal temperature at 42° to 43°C. Chemotherapy schedules were as follows: cisplatin (CDDP; 25 mg/m 2 /L) and mitomycin C (MMC; 3.3 mg/m 2 /L) for colorectal cancer, gastric cancers, and pseudomyxoma peritonei; and CDDP (43 mg/L of perfusate) and doxorubicin (Dx; 15.25 mg/L of perfusate) for ovarian cancer and peritoneal mesothelioma. Patients older than 70 years and those who had undergone previous chemotherapy received a 30% dose reduction of both drugs. The same surgical team guided by the same principal operator (D.M.) operated on all of the patients.
In the postoperative period, patients were assisted in an intensive care unit for at least 48 hours. In the successive in-hospital stay, patients were cared for by a multidisciplinary team comprising internists with a nutritional science specialization, anesthetists, rehabilitation professionals, and psychologists.
Study Parameters
Adverse events (morbidity, anastomotic leak, systemic toxicity) were graded according to the NCI CTCAE v3 criteria. When more than 2 complications occurred during the same period in 1 patient, the complication was considered as more serious for analysis. Anastomotic leak was defined according to the criteria established elsewhere. The procedure-related mortality (PRM) was defined as death occurring during the in-hospital stay after CRS and HIPEC. G3-5 morbidity was defined as the combination of surgical and medical G3-5 morbidities, PRM, and G3-5 systemic toxicity. To assess surgical outcomes according to case sequence, they were divided into 6 subgroups of 70 for cases 1 to 420, while cases 421 to 462 were designated as another subgroup.
The outcomes of interest were rates of incomplete cytoreduction, G3-5 morbidity, and short-term oncologic failure. Potential confounders to the outcomes of interest were identified among the following clinicopathological parameters: age, sex, ECOG performance status, Charlson comorbidity index (combined condition and age-related score), previous systemic chemotherapy, previous surgical score (PSS), body mass index (BMI), preoperative serum albumin, tumor histotype, PMP histologic subtype (disseminated peritoneal adenomucinosis [DPAM], peritoneal mucinous carcinomatosis [PMCA]), PM subtype (biphasic/sarcomatoid vs others), PCI, bowel anastomosis, number of peritonectomy procedures, operating time, CDDP dose for HIPEC, subgroups of case sequence, incomplete cytoreduction rate, and HIPEC drug schedule.
The risk modeling and elaboration of SPRT curves for oncologic outcomes were done separately for PMP and PM. Short-term oncologic failure was defined as emergence of recurrence and/or death within 1 year from the combined procedure for PMP patients and within 6 months for PM patients. Therefore, PMP cases with less than 1 year of follow-up and PM cases with less than 6 months of follow-up that did not present recurrence and/or death were excluded.
Statistical Analysis
One-way analysis of variance (ANOVA) and the χ 2 test were used to analyze changes of continuous and discrete variables, respectively, according to subgroups of case sequence.
The risk-adjusted SPRT plot was used to chart, across the case sequence, changes in the rates of incomplete cytoreduction, rates of G3-5 morbidity, and short-term oncologic failure. To elaborate the SPRT, 4 parameters were defined: estimated probabilities of outcomes of interest for each case, a prespecified odds ratio (OR) for the outcomes of interest, and type I and II error rates. Probability of type I and type II (α and β) error were both set at 0.05. From these, 2 control limits (h0 and h1) and the cumulative sum of log-likelihood ratio with risk adjustment were calculated according to equations outlined by Rogers and colleagues.
The risk predictions for incomplete cytoreduction, G3-5 morbidity, and short-term oncologic failures were elaborated by multivariate analysis, using the logistic regression model. Clinical factors were selected as covariates when P values were less than .15 on univariate analysis. Using backward stepwise selection, independent factors were identified. The probability of event was calculated, which was adjusted with independent risk factors of incomplete cytoreduction, G3-5 morbidity, or short-term oncologic failures from the logistic regression model. Discrimination was measured by the area under the receiver-operating characteristic curve (AUC), with values of 0.5 representing no discrimination and 1.0 representing perfect discrimination. Model fitness was assessed by the Hosmer-Lemeshow goodness-of-fit test, with P values of greater than .05 indicating acceptable fit.
When creating the RA-SPRT curve, each case was plotted in sequence along the x-axis. When a success occurred (no G3-5 morbidity, optimal cytoreduction, or short-term oncologic success) a log-likelihood ratio with risk adjustment ( s ) was subtracted from the cumulative score. When a failure occurred, the constant 1 − s was added to the cumulative score. Thus, an ascending slope in the RA-SPRT line indicated deterioration of performance, whereas a descending slope indicated improvement of performance.
An unacceptable OR was set for increase in rates of incomplete cytoreduction and G3-5 morbidity, respectively, at 1.8 and 1.4, adopting the same boundary lines of the authors’ previous experience of LC analysis. In brief, these figures were obtained considering the highest and lowest rates of these outcomes available in the literature. The upper and lower control limits were respectively defined as h1 (reject line) and h0 (accept line). OR was prespecified as 2 for short-term oncologic failure, for both PMP and PM.
If the RA-SPRT curve crossed the upper decision limit (h1) from below, this meant that the actual OR for outcome was equal to or higher than the prespecified OR with the probability of type I error of 0.05. If the line crossed the lower decision limit (h0 or accept line) from above, this indicated that the actual OR for the outcome being studied is less than the unacceptable OR with the probability of type II error of 0.05. When the line was between h0 and h1, no statistical inference could be made. Expertise was estimated to be achieved at point(s) in which the curve of RA-SPRT crossed the accept line (h0) for outcomes of interest. At the time of crossing to the lower boundary, the graph was reset to 0 to start the surveillance of surgical performance with audit intent.
Once the expertise is acquired, the surveillance continues to monitor any eventual deterioration of the performance. The RA-SPRT curve could assume a positive slope in the case of enrollment of new surgeons in the surgical team, launch of new protocols using more toxic or more effective treatments, and the introduction of changes or innovations in surgical strategy based on new data obtained from other centers.
Evolution of predicted risk of incomplete cytoreduction and G3-5 morbidity along the case sequence was assessed with the moving-average (MA) method. An MA order of 10 was used.
The applied statistical software was SPSS 18.0 (SPSS Inc, Chicago, IL), while the MA method and RA-SPRT model were calculated by Excel version 2003 (Microsoft Corporation, Redmond, WA). Statistical significance was set at P <.05.
Results
Patients’ Characteristics in the Entire Series
The mean age was 53.7 (SD ± 12.6). The male/female ratio was 172:290. Four hundred fifty-two (97.8%) cases had ECOG performance status of 0 or 1. The mean Charlson comorbidity index was 3.6 (±0.9). The mean BMI was 25.3 (±14.4) kg/m 2 . The mean preoperative serum albumin level was 4.1 (±0.6) g/dL. Three hundred nineteen (69%) cases had a PSS of 0 or 1. Two hundred sixty-one cases (56.6%) had previous systemic chemotherapy. The histologic distribution was as follows: colon cancer 23 (5.0%), gastric cancer 12 (2.6%), ovarian cancer 56 (12.1%), PM 151 (32.7%), PMP 168 (36.4%), peritoneal sarcomatosis 34 (7.4%), and other tumors 18 (3.9%). The mean PCI was 18.1 (±9.9).
The mean number of peritonectomy procedures performed for each case was 8.0 (±4.4). The mean operating time was 557 (±164) minutes. The mean number of red cell units transfused intraoperatively was 2.8 (±3.5). Rate of incomplete cytoreduction was 10.2%. The HIPEC drug schedules were distributed as follows: 56.5% CDDP + Dx and 43.5% CDDP + MMC. The mean CDDP dose was 186 (±46) mg. The mean length of stay in the intensive care unit was 3 days (range 0–10) and the mean in-hospital stay was 23.7 (±15.3) days.
Rates of G3-5 morbidity, G3-5 hematologic toxicity, and PRM were 29.2%, 6.2%, and 2.4%, respectively. The most common G3-5 surgical complications were gastrointestinal anastomotic leak/perforation (9.7%), infection (10.4%), and postoperative bleeding (3.2%).
The independent risk factors for the outcomes of interest (incomplete cytoreduction, G3-5 morbidity, short-term oncologic failures for PMP and PM) as well as goodness of fit and discriminant capacity of each risk modeling are outlined in Table 1 . The MA of predicted risk of outcomes of interest along the case sequence are outlined in Fig. 1 .
| Dependent Variables | Independent Risk Factors | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Incomplete Cytoreduction | G3-5 Morbidity | Six-Month Recurrence/Death in 119 PM | One-Year Recurrence/Death in 136 PMP | ||||||||||
| OR (adj) | 95% CI | P | OR (adj) | 95% CI | P | OR (adj) | 95% CI | P | OR (adj) | 95% CI | P | ||
| Age (>52 y) | 11.09 | 1.37 | 90.12 | .02 | |||||||||
| Gender (male) | 2.25 | 0.92–5.50 | .08 | ||||||||||
| ECOG PS (0 vs 1/2) | 0.37 | 0.19–0.73 | .04 | 0.52 | 0.31–0.87 | .013 | |||||||
| Charlson comorbidity index (>4) | |||||||||||||
| Preoperative serum albumin <3.2 g/dL | 6.36 | 1.21 | 33.51 | .03 | |||||||||
| BMI <20 kg/m 2 | 4.54 | 1.19 | 17.30 | .03 | |||||||||
| PM histologic subtype | — | — | — | v | 5.14 | 1.13 | 23.29 | .03 | — | — | — | ||
| PMP histologic subtype | — | — | — | — | — | — | — | — | — | 3.27 | 1.22–8.76 | .02 | |
| PSS (0/1 vs 2/3) | |||||||||||||
| Previous sCT | 2.16 | 1.12–4.19 | .022 | ||||||||||
| PCI (>20) | 4.13 | 2.04–8.38 | <.001 | 2.44 | 1.57–3.80 | <.001 | |||||||
| Anastomosis | 0.38 | 0.19–0.76 | .006 | 2.85 | 0.94–8.58 | .06 | |||||||
| Mean no. of peritonectomy procedures (>8) | |||||||||||||
| CC | — | — | — | ||||||||||
| Ostomy | |||||||||||||
| HIPEC drug schedule (CDDP + Dx vs CDDP+MMC) | — | — | — | 1.72 | 1.10–2.69 | .018 | — | — | — | — | — | — | |
| CDDP dose (>240 mg) | — | — | — | 2.75 | 1.38–5.47 | .004 | |||||||
| Duration of the surgery (>570 min) | |||||||||||||
| Subgroups of case sequence a | 0.77 | 0.65–0.92 | .004 | 1.15 | 1.01–1.31 | .032 | — | — | — | — | — | — | |
| G3-5 morbidity | — | — | — | — | — | — | 2.43 | 1.01–5.86 | .05 | ||||
| Expertise for CC and morbidity | — | — | — | — | — | — | 0.16 | 0.04–0.62 | .01 | ||||
| Model goodness of fit b | χ 2 = 5.69; P = .68 | χ 2 = 4.73; P = .786 | χ 2 = 2.78; P = .427 | χ 2 = 2.01; P = .98 | |||||||||
| Discriminant capacity (AUC) | 0.78; 95% CI: 0.71–0.85 | 0.69; 95% CI: 0.63–0.74 | 0.81; 95% CI: 0.69–0.93 | 0.75; 95% CI: 0.66–0.85 | |||||||||
Related posts:
Hyperthermic Intraperitoneal Chemotherapy
Current Status and Future Directions in Appendiceal Cancer with Peritoneal Dissemination
Pharmacology of Perioperative Intraperitoneal and Intravenous Chemotherapy in Patients with Peritoneal Surface Malignancy
Early Intervention for Treatment and Prevention of Colorectal Carcinomatosis
Current Status and Future Directions of Cytoreductive Surgery and Hyperthermic Intraperitoneal Chemotherapy in the Treatment of Ovarian Cancer
Randomized Clinical Trials for Colorectal Cancer Peritoneal Surface Malignancy
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
