Effect of Tumor Stage on Surveillance Strategies After Treatment for Ovarian Cancer

Abstract

Background: Given that ovarian cancer has a generally poor prognosis and tends to recur with low survival rates, surveillance of these cancers after treatment is vital. However surveillance strategies tend to vary among gynecologic oncologists. Objective: This survey was conducted to evaluate if initial International Federation of Gynecology and Obstetrics stage accounts for the variation in surveillance strategies used by gynecologic oncologists after initial curative-intent treatment for ovarian carcinoma. Materials and Methods: We created a survey instrument directed to the Society of Gynecologic Oncology (SGO). The survey was conducted by e-mail and included SGO members and candidate members. The 943 SGO members were asked to report how frequently they recommended 11 surveillance modalities in years 1–5 and 10 after primary treatment. Statistical evaluation included generalized linear models with a Poisson distribution. Results: Two hundred and eighty-three of the survey results were evaluable. For several surveillance modalities (office visit, pelvic examination, complete blood count, serum CA-125 level, and comprehensive metabolic panel), the frequency of recommended use was statistically significantly greater for more advanced stage tumors. Conclusions: The intensity of surveillance recommended by practicing gynecologic oncologists varies significantly based on initial patient stage and perceived prognosis. However, the variability related to initial prognosis is clinically small and cannot account for the overall variability previously reported. (J GYNECOL SURG 31:326)

Introduction

Despite advances in detection and treatment, ovarian cancer remains a common cancer and significant cause of death in female patients in the United States, with >20,000 new cases and ∼14,000 deaths from this disease each year.¹ Ovarian carcinoma has a poor prognosis overall. In wealthy nations with modern medical systems, estimates of 5-year survival probability—as defined by the International Federation of Gynecology and Obstetrics (FIGO) guidelines—are 90% for stage IA, 86% for IB, 83% for IC, 71% for IIA, 66% for IIB, 71% for IIC, 47% for IIIA, 42% for IIIB, 33% for IIIC, and 19% for IV.²

The initial tumor stage is an important prognostic factor. For early stage tumors (FIGO stage IA-IIB), treatment is confined to surgical management.³ High-risk patients in these stages may also be given adjuvant platinum-based chemotherapy. Currently, ovarian carcinoma is rarely detected at an early stage; 75% of women present with advanced disease even in wealthy nations.⁴ Advanced-stage disease is treated with cytoreductive surgery and taxane- and platinum-based chemotherapy.^4,5 Although most patients respond well to surgery and chemotherapy, ∼75% of clinically complete responders and 50% of pathologically complete responders experience relapse.³ The prognosis for these patients is poor. Only 30%–50% survive, and the median survival time after relapse is 2 years.^3,5 Recurrent disease is typically not curable so management usually involves palliative care.⁶ The high incidence of ovarian carcinoma and its poor prognosis have led to efforts to detect recurrence as early as possible, with the hope that this would improve survival.

The topic of post-treatment surveillance is underappreciated as a medical and societal problem. This is unfortunate, because surveillance is widely advocated and quite expensive. The National Comprehensive Cancer Network has proposed explicit surveillance strategies (www.nccn.org) for patients treated with surgery alone and surgery plus chemotherapy and radiation. Guidelines from the American Society of Clinical Oncology (ASCO, www.asco.org), the Society of Surgical Oncology (www.surgonc.com), Cancer Care Ontario (www.cancercare.on.ca), the National Institute for Clinical Excellence (www.nice.org.uk), the National Guideline Clearing House (www.guidelines.gov), and the Cochrane Collaboration, (www. Cochrane.org), are also available but are not explicit. The European Society of Medical Oncology (www.esmo.org) provides guidelines based on low-quality evidence; they are not explicit and are not based on randomized control trials.

Edelman et al. evaluated the utility of post-treatment patient surveillance with several common types of cancer (not including ovarian cancer), including economic analysis.⁷ These researchers advocated that the following axioms should apply:

(1) The interval between examinations and duration of testing should be consistent with the time of maximal risk of recurrence and natural history of the tumor.

(2) The tests should be directed at the most likely sites of recurrence and should have high predictive values, both positive and negative.

(3) Therapy should be available that will result in cure, significant prolongation of life, or palliation of symptoms. The initiation of earlier therapy should improve the outcome, compared with therapy given when the patient becomes symptomatic from the tumor.

(4) The increased risk of a second malignancy, both in the primary site and in other tissues that may have been exposed to the same carcinogens, or in which there is epidemiologic evidence of increased risk, should also guide the ordering of tests. Malignant and nonmalignant complications of therapy must also be considered. Such testing should also be performed with a frequency and duration consistent with the nature of the risk and include only tests with high positive and negative predictive values.

Several comprehensive surveillance strategies have been proposed to detect recurrence, including one offered by a Society of Gynecologic Oncology (SGO) expert panel.⁸ None are based on well-designed, adequately powered randomized controlled trials. To determine the variability in the clinical practice of ovarian carcinoma surveillance, for this survey, SGO members were contacted. The results of this survey quantify the variation in the strategies used. The aim was to determine if this variation in reported surveillance intensity is the result of differences in initial patient prognosis.

Materials and Methods

A survey was conducted with the members of the SGO. The details of this survey have been published.⁹ A cover letter, a three-page survey, and a self-addressed return envelope by conventional mail were sent to all 943 SGO members and candidate members. Only those involved in long-term ovarian cancer–patient surveillance after initial treatment were asked to fill out the survey. No tangible incentive was offered. These experts were asked to describe the surveillance strategies they would use based on four idealized vignettes featuring otherwise healthy women with ovarian carcinoma of different prognoses (FIGO stage I, stage II, stage IIIA (<1 cm of residual disease after maximal debulking), and stage IIIB (>1 cm of residual disease after maximal debulking). The SGO members were asked to state how frequently they would use 11 surveillance modalities in years 1–5 and 10 after curative-intent primary treatment. The modalities offered in the survey were based on a current review of literature and supplemented by an informal review by local SGO members. This confirmed that no other modalities were commonly used.

Statistical analysis was performed to quantify how frequently the modalities were recommended in each year and to determine if the variation in the frequency of their use was attributable to the initial prognosis of the patients. Responses were grouped for all vignettes in all years, and the mean frequency (±one standard deviation [SD]) of recommended utilization was calculated for each diagnostic modality in each post-treatment year. The number of times each modality was recommended in a particular year was fitted into a generalized linear model with a Poisson distribution, followed by multiple post hoc comparisons to assess the effect of prognosis on practice patterns each year. The effect of prognosis on practice patterns was assessed, using generalized estimating equations to account for potential correlation among repeated measures from the same respondent. The statistical analysis was performed with the SAS statistical package (SAS Institutes, Cary, NC). A p-value <0.05 was considered to indicate statistical significance and all tests were two-sided. The data were also expressed as median (minimum, maximum).

Results

There were 323 SGO members who responded to the initial mailings. Two hundred and eighty-three surveys were considered evaluable after application of the preset exclusion criteria. The results, including the mean and SD in frequency of modality use for Stage IIIB, are shown in Table 1. This FIGO stage is a common one in the United States and other wealthy nations. Comparable patients in the Surveillance, Epidemiology, and End Results (SEER) nomenclature are in the “distant” category, which comprises about 62% of newly diagnosed patients in the United States. Table 2 depicts this data as median and range. Table 3 shows the mean and SD in frequency of modality use for each stage and p-value to determine the statistical significance of the differences among the stages.

Table 1.

Frequency of Recommended Use of Surveillance Modalities in Years 1–5 and 10 for Stage IIIB Ovarian Carcinoma (Not Optimally Debulked): Mean±1 Standard Deviation

Modality	Year 1	Year 2	Year 3	Year 4	Year 5	Year 10
Office visit	4.4±1.5	3.7±1.0	2.7±1.0	2.2±0.6	2.0±0.7	1.1±0.6
Pelvic examination	4.4±1.5	3.7±1.0	2.6±1.0	2.2±0.7	2.0±0.7	1.1±0.6
Papanicolaou smear	1.2±1.3	1.1±1.2	1.0±0.8	0.9±0.7	0.9±0.6	0.8±0.4
Complete blood count	1.6±2.4	1.0±1.5	0.7±0.9	0.6±0.8	0.5±0.8	0.4±0.5
Metabolic panel	1.2±1.9	0.8±1.1	0.5±0.8	0.5±0.7	0.5±0.7	0.3±0.5
Serum CA-125 level	4.3±1.8	3.7±1.4	2.7±1.4	2.2±0.9	2.0±0.9	1.1±0.7
Chest X-ray	0.5±0.8	0.4±0.7	0.3±0.6	0.3±0.5	0.3±0.5	0.2±0.4
Abdominal-pelvic CT	0.6±0.8	0.4±0.7	0.3±0.5	0.2±0.5	0.2±0.4	0.1±0.3
Chest CT	0.1±0.4	0.1±0.3	0.0±0.2	0.0±0.2	0.0±0.2	0.0±0.1
Abdominal-pelvic MRI	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.6	0.0±0.0	0.0±0.0
Transvaginal ultrasonogram	0.2±1.1	0.2±1.0	0.1±0.6	0.1±0.6	0.1±0.5	0.1±0.2

The number in each cell is the number of times a particular surveillance modality was recommended in a particular post-treatment year. These data present all values from all evaluable responses for one particular vignette as mean±one standard deviation. This depiction of the data gives a conservative impression of the variability in practice.

CT, computed tomography; MRI, magnetic resonance imaging.

Table 2.

Frequency of Recommended Use of Surveillance Modalities in Years 1–5 and 10 for Stage IIIB Ovarian Carcinoma (Not Optimally Debulked)

Modality	Year 1 (min, max)	Year 2 (min, max)	Year 3 (min, max)	Year 4 (min, max)	Year 5 (min, max)	Year 10 (min, max)
Office visit	4 (2, 12)	4 (2, 12)	2 (0, 6)	2 (0, 6)	2 (0, 6)	1 (0, 6)
Pelvic examination	4 (1, 12)	4 (1, 12)	2 (0, 6)	2 (0, 6)	2 (0, 6)	1 (0, 6)
Papanicolaou smear	1 (0, 12)	1 (0, 12)	1 (0, 6)	1 (0, 4)	1 (0, 3)	1 (0, 2)
Complete blood count	1 (0, 14)	0 (0, 12)	0 (0, 4)	0 (0, 4)	0 (0, 4)	0 (0, 2)
Metabolic panel	0 (0, 14)	0 (0, 4)	0 (0, 4)	0 (0, 4)	0 (0, 4)	0 (0, 2)
Serum CA-125 level	4 (0, 12)	4 (0, 12)	2 (0, 12)	2 (0, 6)	2 (0, 6)	1 (0, 6)
Chest X-ray	0 (0, 6)	0 (0, 4)	0 (0, 4)	0 (0, 3)	0 (0, 2)	0 (0, 1)
Abdominal-pelvic CT	0 (0, 4)	0 (0, 3)	0 (0, 2)	0 (0, 2)	0 (0, 2)	0 (0, 1)
Chest CT	0 (0, 4)	0 (0, 3)	0 (0, 2)	0 (0, 2)	0 (0, 2)	0 (0, 1)
Abdominal-pelvic MRI	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)
Transvaginal ultrasonogram	0 (0, 12)	0 (0, 12)	0 (0, 6)	0 (0, 4)	0 (0, 4)	0 (0, 2)

The number in each cell is the number of times a particular surveillance modality is recommended in a particular post-treatment year. This data present all values from all evaluable responses for one particular vignette as median and range (minimum, maximum). This depiction of the data emphasizes the variability in practice.

min, minimum; max, maximum; CT, computed tomography; MRI, magnetic resonance imaging.

Table 3.

Frequency of Recommended Surveillance Modalities in Years 1–5 and 10: According to Stages I to IIIB

Modality	Year	Stage I	Stage II	Stage IIIA	Stage IIIB	p-Value
Office visit	1	4.0±1.1	4.3±1.3	4.4±1.5	4.4±1.5	<0.002
	2	3.5±0.9	3.6±0.8	3.7±1.0	3.7±1.0	<0.0001
	3	2.4±0.8	2.5±0.9	2.6±0.9	2.7±1.0	<0.0001
	4	2.1±0.6	2.2±0.6	2.2±0.6	2.2±0.6	<0.003
	5	1.9±0.7	2.0±0.6	2.0±0.6	2.0±0.7	<0.02
	10	1.1±0.5	1.1±0.5	1.1±0.5	1.1±0.6	NS
Pelvic examination	1	4.0±1.1	4.2±1.4	4.3±1.6	4.4±1.5	<0.001
	2	3.5±0.9	3.6±0.9	3.7±1.0	3.7±1.0	<0.0001
	3	2.4±0.8	2.5±0.9	2.6±0.9	2.6±1.0	<0.0001
	4	2.1±0.6	2.2±0.6	2.2±0.6	2.2±0.7	<0.01
	5	1.9±0.6	1.9±0.6	2.0±0.6	2.0±0.7	<0.05
	10	1.1±0.6	1.1±0.5	1.1±0.5	1.1±0.6	NS
Papanicolaou smear	1	1.2±1.3	1.2±1.3	1.2±1.3	1.2±1.3	NS
	2	1.1±1.1	1.1±1.2	1.1±1.2	1.1±1.2	NS
	3	1.0±0.7	1.0±0.8	1.0±0.8	1.0±0.8	NS
	4	0.9±0.6	0.9±0.6	0.9±0.7	0.9±0.7	NS
	5	0.9±0.5	0.9±0.6	0.9±0.6	0.9±0.6	NS
	10	0.8±0.4	0.8±0.4	0.8±0.4	0.8±0.4	NS
Complete blood count	1	1.2±2.0	1.5±2.4	1.6±2.5	1.6±2.4	<0.002
	2	0.7±1.1	0.8±1.2	0.9±1.4	1.0±1.5	<0.0001
	3	0.5±0.7	0.6±0.8	0.6±0.9	0.7±0.9	<0.002
	4	0.5±0.7	0.5±0.7	0.5±0.8	0.6±0.8	<0.01
	5	0.4±0.6	0.5±0.7	0.5±0.7	0.5±0.8	<0.005
	10	0.3±0.5	0.4±0.5	0.4±0.5	0.4±0.5	NS
Metabolic panel	1	0.9±1.6	1.1±1.9	1.1±1.9	1.2±1.9	<0.02
	2	0.6±1.0	0.7±1.1	0.7±1.1	0.8±1.1	<0.005
	3	0.4±0.6	0.5±0.7	0.5±0.7	0.5±0.8	<0.03
	4	0.4±0.6	0.4±0.6	0.4±0.6	0.5±0.7	NS
	5	0.4±0.6	0.4±0.6	0.4±0.6	0.5±0.7	NS
	10	0.3±0.5	0.3±0.5	0.3±0.5	0.3±0.5	NS
Serum CA-125 level	1	4.0±1.5	4.2±1.6	4.3±1.8	4.3±1.8	<0.005
	2	3.4±1.1	3.5±1.1	3.6±1.3	3.7±1.4	<0.02
	3	2.4±1.0	2.5±1.0	2.6±1.2	2.7±1.4	<0.002
	4	2.1±0.8	2.1±0.8	2.2±0.8	2.2±0.9	<0.005
	5	1.9±0.8	1.9±0.8	2.0±0.9	2.0±0.9	<0.01
	10	1.0±0.5	1.0±0.6	1.0±0.6	1.1±0.7	NS
Chest X-ray	1	0.5±0.7	0.5±0.8	0.5±0.8	0.5±0.8	NS
	2	0.4±0.7	0.4±0.7	0.4±0.7	0.4±0.7	NS
	3	0.3±0.8	0.3±0.6	0.3±0.6	0.3±0.6	NS
	4	0.3±0.5	0.3±0.5	0.3±0.5	0.3±0.5	NS
	5	0.3±0.5	0.3±0.5	0.3±0.5	0.3±0.5	NS
	10	0.2±0.4	0.2±0.4	0.2±0.4	0.2±0.4	NS
Abdominal-pelvic CT	1	0.5±0.7	0.6±0.8	0.6±0.8	0.6±0.8	<0.0001
	2	0.3±0.6	0.4±0.7	0.4±0.7	0.4±0.7	<0.003
	3	0.3±0.7	0.3±0.5	0.3±0.6	0.3±0.5	NS
	4	0.2±0.4	0.2±0.4	0.2±0.5	0.2±0.5	NS
	5	0.2±0.7	0.2±0.4	0.2±0.4	0.2±0.4	NS
	10	0.1±0.2	0.1±0.2	0.1±0.3	0.1±0.3	NS
Chest CT^a	1	0.1±0.4	0.1±0.4	0.1±0.4	0.1±0.4	—
	2	0.0±0.3	0.1±0.3	0.1±0.3	0.1±0.3	—
	3	0.0±0.2	0.0±0.2	0.0±0.2	0.0±0.2	—
	4	0.0±0.2	0.0±0.2	0.0±0.2	0.0±0.2	—
	5	0.0±0.2	0.0±0.2	0.0±0.2	0.0±0.2	—
	10	0.0±0.1	0.0±0.1	0.0±0.1	0.0±0.1	—
Abdominal-pelvic MRI^a	1	0.0±0.0	0.0±0.1	0.0±0.1	0.0±0.0	—
	2	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	—
	3	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	—
	4	0.0±0.6	0.0±0.6	0.0±0.6	0.0±0.6	—
	5	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	—
	10	0.0±0.0	0.0±0.0	0.0±0.0	0.0±0.0	—
Transvaginal ultrasonogram	1	0.2±1.1	0.2±1.0	0.2±1.0	0.2±1.1	NS
	2	0.2±1.0	0.2±1.0	0.2±1.0	0.2±1.0	NS
	3	0.1±0.6	0.1±0.6	0.1±0.6	0.1±0.6	NS
	4	0.1±0.5	0.1±0.5	0.1±0.5	0.1±0.6	NS
	5	0.1±0.4	0.1±0.4	0.1±0.4	0.1±0.5	NS
	10	0.1±0.2	0.0±0.2	0.0±0.2	0.1±0.2	NS

Note: This table shows the differences in recommended use of surveillance modalities for the different International Federation of Gynecology and Obstetrics stages and the associated statistical significance of the differences based on these stages.

The data for chest CT and abdominal-pelvic MRI are presented without statistical analysis.

CT, computed tomography; MRI, magnetic resonance imaging; NS, nonsignificant.

The frequency of recommended use decreased with each post-treatment year, as expected. Significant effects attributable to initial tumor prognosis (p<0.005) were detected during at least the first 3 years for the following modalities: office visit; pelvic examination; complete blood count; comprehensive metabolic panel; and serum CA-125 level. The frequency of recommended use of Papanicolaou smear, chest X-ray, and abdominal-pelvic computed tomography (CT) after the second postoperative year, and the frequency of recommended use of transvaginal ultrasound for all postoperative years did not vary statistically significantly, based on initial prognosis. Chest CT and abdominal-pelvic magnetic resonance imaging were used too infrequently to yield any clinically meaningful results and thus were not analyzed statistically. Initial stage for most diagnostic modalities had a statistically significant influence on the frequency of recommended use for most diagnostic modalities. However, the absolute differences in surveillance intensity were quite small.

Discussion

The current survey researchers determined that SGO members appeared to utilize surveillance modalities at statistically significantly different rates. Many tests are used by gynecologic oncologists for surveillance of the conditions of ovarian cancer patients. The data from this survey provide direct quantitative empirical data about the surveillance strategies used by expert clinicians. The main modalities used by these gynecologic oncologists were office visit, pelvic examination, and serum CA-125 level. Other modalities were used less often.

Wennberg and Gittelsohn are credited with demonstrating that quantitative analysis of geographic variation in medical practice can be an effective tool for detecting unwarranted inequalities in resource utilization.¹⁰ This approach has been widely applied and generalized to include other independent variables, such as physician age, patient race, insurance status, etc. The four idealized vignettes used in the survey eliminated all potential variables except stage, but other sources of variation are likely to have affected the data. Recall bias (inaccurate recollection of how often each modality was recommended) is likely to have occurred. Practice location (inner city versus wealthy suburbs), insurance status, and other difficult-to-measure parameters may explain the variations that were observed.

Long-term surveillance is believed to be valuable for most patients but Harmandayan et al. have shown previously that there is considerable variation in the strategies used by practicing experts.⁹ Current guidelines concerning the surveillance of ovarian cancer patients after initial treatment are almost all consensus-based because of the paucity of well-controlled clinical trials. Such trials can clearly be influential, as shown in the 2010 practice-changing report of Rustin et al., which indicated that serial serum measurement of serum CA-125 concentration confers no survival benefit to ovarian cancer patients.¹¹ The main finding of the current analysis is that surveillance strategies vary according to initial patient prognosis, although this variation is quite small. These results also provide data on which to base future controlled trials to determine which post-treatment surveillance modalities, if any, are effective for improving overall survival or enhancing quality of life (QoL).

In view of Edelman's axioms,⁷ the value of surveillance for ovarian cancer patients as measured by increased length of life or QoL is essentially unknown. There are no randomized trials involving modalities other than serum CA-125 levels. Such trials are expensive and time-consuming. A major reason such randomized controlled trials have not been carried out is the fact that relevant funding agencies have not provided the required resources. The 2010 U.S. Patient Protection and Affordable Care Act made provisions for comparative effectiveness research. Cancer patient surveillance after initial curative-intent treatment is an excellent target for such research. As more patients are cured, the number of long-term survivors is projected to increase. The testing modalities in current use are quite costly, uncomfortable, and often inconvenient. If well-controlled large trials indicate that a surveillance modality is not valuable, changes in surveillance practices should result, with subsequent cost savings.

Conclusions

Given that variation in the use of medical interventions is now considered prima facie evidence of overuse, underuse, and/or misuse of scarce medical resources, such variations should prompt research to identify the most beneficial treatment modalities and schedules of post-treatment surveillance. The results of this survey indicate that surveillance intensity is comparable among SGO members.

Most of the existing literature presents low-quality data and, clearly, has failed to define best medical practice. The need for well-designed clinical trials to compare the utility of a high-intensity surveillance strategy versus a low-intensity surveillance strategy is also clear. Trials will need to be repeated periodically as new diagnostic tests and new treatments for ovarian cancer become available. Given that there currently are >14 million cancer survivors in the United States alone, the need for such trials for most types of cancer is great. Professional societies such as ASCO, advocacy groups, political organizations, and other entities will presumably have to demand that funding be provided in order to carry out these trials. Anything less than robust evidence is not likely to reduce the variation in clinical practice.

Footnotes

Acknowledgments

The authors wish to acknowledge the support of the Biostatistics Core, the Siteman Cancer Center, and the National Cancer Institute's Cancer Center Support Grant P30 CA091842.

Disclosure Statement

None of the authors have a conflict of interest to report.

References

Siegel

, Ma

, Zou

, Jemal

. Cancer statistics, 2014. CA Cancer J Clin, 2014; 64:9.

Heintz

, Odicino

, Maisonneuve

, et al. Carcinoma of the ovary: FIGO 27th Annual Report on the Results of Treatment in Gynecological Cancer. Int J Gynaecol Obstet, 2006; 95(suppl1):S161.

Gadducci

, Cosio

. Surveillance of patients after initial treatment of ovarian cancer. Crit Rev Oncol Hematol, 2009; 71:43.

Fader

, Rose

. Role of surgery in ovarian carcinoma. J Clin Oncol, 2007; 25:2873.

Berkenblit

, Cannistra

. Advances in the management of epithelial ovarian cancer. J Reprod Med, 2005; 50:426.

Prat

, Parera

, Adamo

, et al. Risk of recurrence during follow-up for optimally treated advanced epithelial ovarian cancer (EOC) with a low-level increase of serum CA-125 levels. Ann Oncol, 2009; 20:294.

Edelman

, Meyers

, Siegel

. The utility of follow-up testing after curative cancer therapy: A critical review and economic analysis. J Gen Intern Med, 1997; 12:318.

Salani

, Backes

, Fung

, Holschneider

, Parker

, Bristow

, Goff

. Posttreatment surveillance and diagnosis of recurrence in women with gynecologic malignancies: Society of Gynecologic Oncologists recommendations. Am J Obstet Gynecol, 2011; 204:466.

Harmandayan

, Gao

, Mutch

, Virgo

, Gibb

, Johnson

. Ovarian cancer patient surveillance after curative-intent initial treatment. Gynecol Oncol, 2011; 120:205.

10.

Wennberg

, Gittelsohn . Small area variations in health care delivery. Science, 1973; 182:1102.

11.

Rustin

, van der Burg

, Griffin

, et al. Early versus delayed treatment of relapsed ovarian cancer (MRC OV05/EORTC 55955): A randomized trial. Lancet, 2010; 376:1155.