The optimal organization of gynecologic oncology services : a systematic review

Background A system-level organizational guideline for gynecologic oncology was identified by a provincial cancer agency as a key priority based on input from stakeholders, data showing more limited availability of multidisciplinary or specialist care in lower-volume than in higher-volume hospitals in the relevant jurisdiction, and variable rates of staging for ovarian and endometrial cancer patients. Methods A systematic review assessed the relationship of the organization of gynecologic oncology services with patient survival and surgical outcomes. The electronic databases medline and embase (ovid: 1996 through 9 January 2015) were searched using terms related to gynecologic malignancies combined with organization of services, patterns of care, and various facility and physician characteristics. Outcomes of interest included overall or disease-specific survival, short-term survival, adequate staging, and degree of cytoreduction or optimal cytoreduction (or both) for ovarian cancer patients by hospital or physician type, and rate of discrepancy in initial diagnoses and intraoperative consultation between non-specialist pathologists and gyne-oncology–specialist pathologists. Results One systematic review and sixteen additional primary studies met the inclusion criteria. The evidence base as a whole was judged to be of lower quality; however, a trend toward improved outcomes with centralization of gynecologic oncology was found, particularly with respect to the gynecologic oncology care of patients with advanced-stage ovarian cancer. Conclusions Improvements in outcomes with centralization of gynecologic oncology services can be attributed to a number of factors, including access to specialist care and multidisciplinary team management. Findings of this systematic review should be used with caution because of the limitations of the evidence base; however, an expert consensus process made it possible to create recommendations for implementation.


INTRODUCTION
The annual incidence of gynecologic cancers exceeds a million cases worldwide, with half a million deaths annually 1 .In some jurisdictions, gynecologic oncology services have been centralized 2 in one or more centres with higher patient volumes and interdisciplinary collaboration 3 .Those centres receive referrals from less-specialized hospitals within a network, region, or defined catchment area.The effectiveness of that model of service delivery has been shown for other cancer disease sites 4 , but its benefits have historically not been as clear for gynecologic oncology [5][6][7] .
The present systematic review was designed to assess the relationship of the organization of gynecologic oncology services with patient survival, surgical outcomes, the delivery of chemotherapy, and some aspects of the role of specialist pathologists in gynecologic oncology.This choice of topic was based on perceived issues, including the absence of a health care system-wide network of care and a lack of collaboration, even in a setting in which most gynecologic oncologists are practicing in major centres with teaching hospitals and using a multidisciplinary team approach.Specific areas of potential concern for patient care included low rates of appropriate disease staging 8,9 , geographic variations in treatment 8 , and variations in Outcomes of interest included overall or disease-specific survival, short-term survival, adequate staging, degree of cytoreduction or optimal cytoreduction (or both) for ovarian cancer patients by hospital or physician type, and rate of discrepancy in initial diagnoses and intraoperative consultation between non-specialist pathologists and gyne-oncology-specialist pathologists.

METHODS
The electronic databases medline and embase (ovid: 1996 through 12 December 2011) were searched using terms related to gynecologic malignancies combined with organization of services, patterns of care, and various facility and physician characteristics (Table i).The Cochrane Database of Systematic Reviews was searched for topic-specific reviews up to Issue 12 (December), 2011.Reference lists of included articles were scanned for additional citations.The search engine Google Scholar was used to identify articles using the terms "gynecological cancer" or "gynaecological cancer" and "centralisation" or "centralization" or "volumes."Additional search terms were used to locate studies that addressed some aspect of quality in pathology diagnosis.
Articles were selected by the project methodologist and were reviewed by all authors.To identify existing relevant guidelines published between 1995 and 2011, a search was conducted of the Web sites of major guideline developers a .a Standards and Guidelines Evidence directory (http://www.ca ncer v iew.ca/c v/por ta l/Home/Treat ment A ndSuppor t/ TSProfessionals/ClinicalGuidelines/GRCMain/GRCSAGE/ GRCSAGESearch), the U.S. National Guidelines Clearinghouse (http://www.guideline.gov), the U.K. National Institute for Health and Clinical Excellence (http://www.nice.org.uk/Guidance), the Scottish Intercollegiate Guidelines Network (http://www.sign.ac.uk/guidelines/index.html), the American Society of Clinical Oncology (http://jco.ascopubs.org/site/misc/specialarticles.xhtml), the U.S. National Comprehensive Cancer Network (http:// www.nccn.org/professionals/physician_gls/f_guidelines.asp),Cancer Australia Guidelines (http://canceraustralia.gov.au/publications-and-resources/clinical-practice-guidelines), and the New Zealand Guidelines Group (http://www.health.govt.nz/about-ministry/ministry-health-websites/new-zealand-guide lines-group) The search was repeated on 9 January 2015 to improve the currency of the evidence base.

Study Selection
Study designs eligible for inclusion were observational studies with a retrospective or prospective assessment of a cohort of patients and systematic reviews of such study designs.Case-control studies and case series were excluded to develop an evidence base of the highest possible quality.
Multidisciplinary care and supportive care delivery were outside the scope of the present review.Studies published in a language other than English and studies published before 1996 were excluded.

Data Extraction and Quality Assessment
Guidelines based on a systematic review of the relevant literature were assessed for quality using the agree ii instrument 10 , a 23-item tool that rates the quality of a guideline in various domains, including scope and purpose, stakeholder involvement, rigour of development, clarity of presentation, applicability, and editorial independence.Systematic reviews were assessed for methodologic quality using the checklist of items provided in amstar 11 .
For individual studies, the determination of study quality was based on an assessment of study design, balance of baseline patient characteristics, completeness of follow-up, and whether the analysis was adjusted for comorbidities, age, and stage and grade of disease 12 .Funding source and outcomes of interest were also extracted.Data extraction was verified by a project research assistant.All authors reviewed and discussed a draft of the evidence summary.

Systematic Reviews
The evidence search identified a guideline published in 1999 by the U.K. National Health Service, Guidance on Commissioning Cancer Services: Improving Outcomes in Gynaecological Cancers -The Manual 13 , and a summary of the associated systematic review 14 that closely aligned with the objectives of our guideline development group.That publication was based on a systematic review of the literature and scored well on most of the agree ii domains.Its evidence base was therefore adopted, and a search for evidence published between 1999 and 2011 was undertaken.
Three additional systematic reviews 12,15,16 that addressed the organization of services for ovarian cancer were identified.The most up-to-date and comprehensive review 15 aligned with the research questions.The working group considered using that systematic review as a source of primary studies; however, it was decided that, because the objectives of that review were so well aligned with the objectives of the present systematic review, the found review 15 would be adopted in its entirety into the evidence base.An additional systematic review 17 found in the January 2015 search was not retained for inclusion in the evidence base because it contained no studies beyond those already captured.

Individual Studies
The search of electronic databases encompassing 1996 to December 2011 identified 3350 unique English-language citations for screening.Ninety of those articles were retrieved for full-text review.Two additional articles for fulltext review were identified from a scan of the reference lists of included articles, Google key word searching, and files of working group members.After review, fifteen articles that met the inclusion criteria were retained.Of 3832 non-duplicate records identified in the update search in January 2015, one additional article 18 met the inclusion criteria and was added to the evidence base.

Study Characteristics and Quality Assessment
Improving Outcomes in Gynaecological Cancers -The Manual 13 included recommendations for the clinical management and organization of services for all gynecologic oncology disease sites.The document scored well on the agree ii domains of scope and purpose, stakeholder involvement, applicability, and clarity of presentation.
A comprehensive ovarian cancer systematic review 15 addressed whether any institutional or physician characteristics are associated with the outcomes of survival, tumour debulking, completeness of staging, and compliance with guidelines for the selection of chemotherapy.This systematic review rated highly on the amstar tool, indicating high-quality methods and reporting, but the body of evidence included in the review was judged to be of lower quality because of the risk of bias inherent in the retrospective nonrandomized design of the studies and other limitations.
Our review also found sixteen primary studies that addressed the centralization of gynecologic oncology services (Table ii).Study locations included the United Kingdom (three studies), elsewhere in Europe (six studies), the United States (five studies), and Canada (two studies).The research design of most studies was retrospective.Data sources included hospitals, regional or national administrative databases and registries, and patient charts.Time periods ranged from the 1990s to about 2005, with the most recent study including data up to 2006.Follow-up times also varied considerably, ranging up to 5 years.Two studies corrected for clustering of outcomes within facilities.The funding source was not stated for most studies; the rest had government funding or a combination of government and private funding.The numbers of patients ranged from 124 to more than 155,000.
Outcomes included various survival measures and the quality of surgery.By consensus, the guideline development group judged that the evidence base was of lower quality because of a lack of consistency in independent variables, outcome measures, study populations, and geographic locations, and because of limited control for clustering of outcomes.A meta-analysis was not considered feasible for the same reasons.

Systematic Reviews Addressing Centralization of Gynecologic Oncology Services
Improving Outcomes in Gynaecological Cancers (United Kingdom): The major findings of Improving Outcomes in Gynaecological Cancers -The Manual 13 15 included seventeen studies that reported survival outcomes and eight studies that reported surgical outcomes, all by hospital type.Eighteen of the studies reported survival outcomes, and thirteen reported surgical outcomes-all by physician type.Overall, the review found these trends, while cautioning that the limited evidence showed considerable heterogeneity: n A positive impact on survival with higher degree of physician specialization in six of eleven studies of patients with advanced disease (figo stages iii-iv) n A significant positive effect of physician discipline and subspecialization on surgical outcomes, including staging and extent of debulking n No relationship between institutional characteristics and survival outcome (This association was not tested in a multivariate analysis, however; and could be clustered with other factors such as hospital volume.)

Primary Studies
Our review found sixteen additional studies that met the inclusion criteria and that were not included in the systematic reviews previously described.

Survival Outcomes by Hospital Type:
One study compared outcomes in the Anglia region of England, where centralization of gynecologic oncology services had been implemented by 2000 (Table iii).More than 97% of cases in the study had been histologically confirmed, and only the 60% of cases with stage and grade data were included in the survival analysis.Survival for patients with gynecologic cancer improved significantly post-centralization compared with the pre-centralization study period [hazard ratio (hr): 0.71; 95% confidence interval (ci): 0.64 to 0.79].
Another study 25 looked at outcomes before and after the implementation of the U.K. National Health Service cancer plan for England in 2000 and found significant improvements in mean annual survival for cervical (p < 0.05), endometrial (p < 0.01), and ovarian cancer (p < 0.01) between 2001 and 2006.
In Denmark, the treatment of ovarian cancer was slowly centralized over a decade, and subsequently, a significant reduction in mortality for patients with stage iiic-iv ovarian cancer treated in tertiary centres, compared with other facilities, was found (hr: 0.03; 95% ci: 0.70 to 0.98) 30 , although data on the involvement of gynecologic oncologists was unavailable.
In one of the few studies to control for the tendency of outcomes to be clustered within facilities 23 , no significant differences in overall survival by hospital type over a 10year period were found.A study of ovarian cancer found no difference in survival for centres with a gynecologic oncologist compared with other centres 21 .Another found no difference in the hr for overall survival of endometrial cancer patients between community and teaching hospitals 22 .One study 26 found a significant improvement in overall survival for patients in facilities with two or more medical oncologists (hr: 0.8; 95% ci: 0.7 to 1.0; data not shown in Table iv), that variable being used as a proxy for centralization.

Survival Outcomes by Physician Type:
Six studies reported survival outcomes by physician type (Table v), including three studies of patients with ovarian cancer 21,24,28 and three studies of patients with endometrial cancer 20,22,29 .Four of those studies included patients with figo stage i-iv disease 21,22,24,29 .One study included only patients with stage ia-iia endometrial cancer 20 .The study that reported outcomes for advanced-stage patients (figo iiic or iv ovarian cancer) was the only one to find a significant effect of treatment by the combined reference group of gynecologic oncologists and gynecologists compared with general surgeons (hr: 1.63; 95% ci: 1.56 to 1.71) or with other physicians (hr: 1.56; 95% ci: 1.52 to 1.61) 28 .The rest of the studies did not find a significant difference between subspecialists and other physician types.

Surgical Outcomes by Hospital Type:
A study 31 found that before centralization in the West Midlands, England, 15 different surgical procedures had been described for the treatment of vulvar cancer 34 .After centralization, only 4 types of surgery were performed, and 84% of patients who required lymph node dissection underwent the procedure, although heterogeneity of surgical technique remained evident.Implementation of centralization also coincided with improved histology reporting and achievement of adequate excision margins 31 .
In a population of patients for whom lymphadenectomy was recommended and feasible, the procedure was significantly more likely to be performed in central hospitals than in primary, secondary, and tertiary hospitals (endometrial cancer: p = 0.001; ovarian cancer: p = 0.046); however, even in the central sites, adherence to appropriate guidelines was found to be lacking 19      for whom that procedure was recommended found that performance of the procedure was significantly more likely at semi-specialized [odds ratio (or): 4.6; 95% ci: 2.3 to 9.2] or specialized hospitals (or: 3.9; 95% ci: 2.0 to 7.6) 12,31 than at general hospitals, and more likely after than before centralization (76% vs. 46%, p = 0.003).One study that assessed optimal debulking also found that procedure to be significantly more likely in specialized than in general hospitals (or: 1.7; 95% ci: 1.1 to 2.7) 12 .
Surgical Outcomes by Physician Type: Two studies reported rates of lymphadenectomy performance by physician type (Table vi).One of them, which included endometrial cancer patients with figo stage ia-iia disease, reported a significant difference between gynecologic oncologists and gynecologists (79% vs. 23%, p = 0.006) 20 .The other, which included ovarian cancer patients with figo stage i-iia disease, reported a significantly greater odds of lymphadenectomy being performed by a gynecologic oncologist (or: 9.5; 95% ci: 4.7 to 19.0) or semi-gynecologic oncologist (or: 8.5; 5% ci: 3.8 to 19.3) 12 than by a gynecologist.In another study 21 , the risk of repeat surgery was significantly higher with gynecologists (or: 6.54; 95% ci: 2.5 to 16.9) or general surgeons (or: 16.97; 95% ci: 6.4 to 45.3) than with gynecologic oncologists.

Chemotherapy-Related Outcomes:
One study found that, in addition to being more likely to perform complete surgical staging, gynecologic oncologists were more likely than other physicians to deliver chemotherapy 29 .

Organization of Gynecologic Pathology:
Four articles that addressed the rates of discrepancy in initial diagnoses and intraoperative consultation between non-specialist pathologists and gyne-oncology-specialist pathologists were identified [35][36][37][38] .In one hospital, intraoperative consultation (that is, analysis of frozen sections) for gynecologic surgery by gynecologists was provided by surgical pathologists; subspecialized gynecologic pathologists provided that support for gynecologic oncologists.In that setting, the rate of major discrepancies between intraoperative consultation and final pathology was significantly increased for the non-specialist pathologists compared with the subspecialist pathologists (p = 0.0266) 35 .
Another study that assessed the use of frozen sections in ovarian cancer found that non-specialist pathologists were significantly more likely than gynecologic oncologist pathologists to "misdiagnose" a tumour.The study found that accurate intraoperative diagnosis of borderline ovarian tumours depended mainly on the level of experience of the pathologist 36 .Similarly, Bige et al. 37 looked at the accuracy of frozen sections in ovarian cancer with the aim of identifying the role of the gynecologic pathologist based on level of experience.Sensitivity and specificity were higher in the subspecialist group (p value not reported), and more malignant tumours diagnosed by frozen section were found to be discordant with the final diagnosis in the group of non-gynecologic oncology pathologists.
Another assessment of intraoperative consultation compared the diagnostic error rates in a specialist gynecologic pathology unit and in a general pathology laboratory.It found that errors attributable to the quality of technical preparation and pathologist misinterpretation were more common in the general pathology laboratory; however, because several pathologists worked in both laboratories, it was not possible to conclude that the observed discrepancy was a result of experience or expertise 38 .

DISCUSSION
The present systematic review analyzes the evidence for organizational factors that affect patient outcomes (surgical and survival) in gynecologic oncology.The research method used was systematic and comprehensive, but  the resulting evidence base was of limited quality and largely comprised nonrandomized retrospective studies with inconsistent definitions of comparison groups and outcomes.A significant limitation of the literature in this area is the failure account for the "clustering" phenomenon: the fact that patients treated within the same facility are not entirely independent, and that the outcomes of patients treated in a single facility tend to be similar 23 .Most of the studies included in the evidence base did not account for clustering, and the effect of hospital type might therefore have been exaggerated.For that reason, it is difficult to draw strong conclusions from the available literature on this topic.In addition, although our review attempted to assess the effect of centralization on outcomes in gynecologic oncology, the systematic review and most of the included primary studies considered patients with ovarian cancer; fewer data were available on other gynecologic cancers.However, despite the lack of evidence from randomized studies, some trends were observed.
Studies found a positive relationship between completeness of cytoreduction and survival in advanced disease 24 .In patients with early-stage disease confined to the ovary, treatment involves extensive surgical staging to rule out occult metastatic disease 39 .A previously published high-quality systematic review that assessed the relationship between organizational factors and outcomes in ovarian cancer concluded that subspecialty care (by a gynecologic oncologist) is associated with better surgical outcomes than is care delivered by non-subspecialists and has a positive effect on survival for patients with advanced disease; institutional characteristics were found to have a smaller effect on outcomes 15 .Our review concurs with the finding that, for advanced-stage ovarian cancer patients, care delivery by subspecialist gynecologic oncologists is associated with better outcomes.
With respect to pathology, our review found that intraoperative consultation should be rendered by gynecologic pathologists when feasible, or that gynecologic pathologists should be available to provide consultation 32 .Such consultation would require centralization to ensure a sufficient caseload.A significant improvement in overall survival for ovarian cancer patients was also found in facilities with a higher number of medical oncologists, suggesting that, where geographic distance allows, delivery of chemotherapy in addition to surgery could benefit from concentration in a specialized centre 26 .However, the latter finding (from a single study) might in fact reflect a benefit of hospital type.
Three studies found a significant improvement in survival for gynecologic oncology patients after centralization of services (compared with an earlier decentralized model).Studies also found that, in more centralized models of care, surgery was more likely to be standardized and lymphadenectomy to be performed when indicated.In one study, the improvements were attributed to multiple features of centrally organized care, including central pathology review for high-risk endometrial cancer, stateof-the-art imaging, more extensive specialist surgery resulting in better staging and more tailored adjuvant therapy, use of the Risk of Malignancy Index to make appropriate referrals, multidisciplinary review for patients having non-specialist surgery, and early discussion of cases in the "unit" hospital.Overall, the major differences were summarized as "access to specialized surgery" and "management within a multidisciplinary team" 35 .At the same time, one study found no difference in survival for ovarian cancer patients treated in centres with subspecialist gynecologic oncologists than in other centres.Compared with studies that assessed survival, studies of surgical outcomes were more likely to find significant differences.The reasons for that discrepancy are not clear and could be the subject of future research.

CONCLUSIONS
The evidence found in the present review is consistent with previous research showing a likely benefit from the delivery of gynecologic oncology care in specialized centres with subspecialists working as part of a multidisciplinary team-particularly for patients with more advanced ovarian cancer.It should be cautioned, however, that unlike other disease sites such as pancreas and esophagus 4 , the evidence for this finding is far from strong.
Because of the lower-quality nature of the evidence base, an expert consensus process was used to create an organizational guideline, the results of which are published separately 40 .Briefly, the consensus process led to recommendations for the performance of definitive surgery for most invasive cancers by subspecialist gynecologic oncologists within designated gynecologic oncology centres.The recommendations indicate that some services, such as radiation therapy, could be provided in other affiliated centres.Multidisciplinary team management is also endorsed, and recommendations are provided for human and physical resource needs and some aspects of pathology.The recommendations are intended to minimize provincial variations in practice and to make best use of limited resources.
Our review also found that, even when practice and outcomes appeared to have improved under a centralized model, heterogeneity of technique remained, even among groups of subspecialists.That finding suggests that a plan to implement care in centralized or specialized facilities featuring subspecialists and multidisciplinary care should also include the adoption of appropriate guidelines across the care continuum and of the necessary system processes and evaluation plans.It will be a challenge to implement those changes across a health care system spanning a large and variable geographic area with a range of independent hospitals.

ACKNOWLEDGMENTS
The work of the Program in Evidence-Based Care (pebc) is supported by the Ontario Ministry of Health and Long-Term Care through Cancer Care Ontario (cco).The pebc is editorially independent of its funding source.
The Gynecologic Oncology Organizational Guideline Working Group thanks the following participants in the guideline development process for their review and comments on the manuscript: Hans Messersmith and Sheila McNair, assistant directors of cco's pebc; Harkanwal Randhawa, student project assistant; Nadia Coakley and Bryan Rumble, pebc health research methodologists; and Wei Cao, project coordinator, Surgical Oncology Program, cco.

b
American Joint Committee on Cancer codes.c Disease-specific survival.FIGO = International Federation of Gynecology and Obstetrics; Pts = patients; HR = hazard ratio; CI = confidence interval; GYO = gynecologic oncologist; GYN = gynecologist; HIR = high-intermediate risk of local failure (stage IB; grade III, IC; grade II or III or stage IIA); GS = general surgeon.Values in boldface type are statistically significant.e290 Current Oncology, Vol.22, No. 4, August 2015 © 2015 Multimed Inc.

TABLE II Continued
33proved for ovarian cancer patients who are treated by expert multidisciplinary teams and who undergo surgery by a specialist gynecologic oncologist.The decision to recommend multidisciplinary care was based mainly on a study of ovarian cancer conducted in Scotland, which found that follow-up at a multidisciplinary clinic was an independent predictor of survival at 5 years.The recommendation for treatment by gynecologic oncologists for most patients was based on findings in a prospective study of all patients in Scotland, which found that women with stage iii cancer who were treated by gynecologic oncologists, compared with gynecologists, had a 25% lower death rate at 3 years.Based on those findings (and other lesser-quality evidence), centralization of services was advised, with all patients recommended for treatment by specialist gynecologic oncologists at designated cancer centres, except for those with stage ia or ib disease [stage ia according to the 2009 revised International Federation of Gynecology and Obstetrics (figo) staging system33], grade 1 or 2 endometrial cancer, and pelvic masses for which the risk of malignancy is low.Cancer units serving smaller populations would refer patients to the designated centres while retaining responsibility for initial diagnostic procedures and surgery for lower-risk cases.
Pts = patients; R = retrospective; NS = not stated; NCIC = National Cancer Institute of Canada; P = prospective; BA = before and after; NA = not available; AMA = American Medical Association.Current Oncology, Vol.22, No. 4, August 2015 © 2015 Multimed Inc. is Du Bois et al., 2009: A comprehensive systematic review of ovarian cancer studies published up to 2007

TABLE III
Hazard ratio for survival by hospital type (specialization or centralization of services, or both, compared with decentralization) a Mean annual change in survival 2001-2006 (restricted to patients who survived at least 1 year).b Five-year survival, death from a gynecologic malignancy during the 10-year study period.c Disease-specific survival.FIGO = International Federation of Gynecology and Obstetrics; Pts = patients; HR = hazard ratio; CI = confidence interval; GYO = gynecologic oncologist; NS = nonsignificant; SEER = Surveillance, Epidemiology, and End Results; ASA = American Society of Anesthesiologists; ECOG = Eastern Cooperative Oncology Group.Values in boldface type are statistically significant.e289 Current Oncology, Vol.22, No. 4, August 2015 © 2015 Multimed Inc.

TABLE IV
Surgical outcomes, by hospital type (specialization and/or centralization of services vs. decentralization) FIGO = International Federation of Gynecology and Obstetrics; Pts = patients; OR = odds ratio; CI = confidence interval; ASA = American Society of Anesthesiologists physical status score.Values in boldface type are statistically significant.

TABLE V
Hazard ratio for survival, by physician type a Survival time from initial surgery to date of death from any cause.Follow-up to latest available health insurance data.

TABLE VI
Surgical outcomes, by physician type (specialist or non-specialist)