The SOCQER-2 Study Surgery in Ovarian Cancer - Quality of Life Evaluation Research

BACKGROUND

The cornerstones of treatment for advanced ovarian cancer are surgical cytoreduction and systemic therapy. The aim of surgery is the maximal cytoreduction of all visible disease, ideally reaching a total macroscopic tumour clearance.

Whilst most gynaecological oncologists accept that surgical procedures needed to achieve complete cytoreduction are on a continuum, a Cochrane review summarises types of ovarian cancer surgery as follows:

• standard surgery
• radical surgery
• supra-[ultra-] radical surgery

Several observational studies and two systematic reviews of non-randomised studies have shown an association between the amount of residual disease following surgery and survival, controlling for various confounding factors. However, the systematic reviews identified no randomised control trials directly comparing standard surgery versus extensive surgery. A National Institute for Health and Care Excellence (NICE) Interventional Procedures Overview found a number of retrospective observational studies comparing extensive with standard surgery, and found that women who underwent extensive surgery had better survival to those undergoing standard surgery, thus suggesting that the utilisation of surgical procedures to achieve complete cytoreduction may overcome disease aggressiveness. Incidence of serious complications, however, was higher in patients undergoing extensive surgery. However, there were several confounding factors e.g patients with good performance status are more likely to be selected to undergo extensive surgery. Equally, patients with lesser disease burden can have complete removal of tumour by standard surgery alone.

A Cochrane review comparing outcomes from ultra-radical surgery versus standard concluded it was unclear whether there were differences in progression free survival, quality of life and morbidity, and that a randomised control trial with sufficient statistical power or well-designed non-randomised studies were needed. Equally important are outcomes other than survival that are valued by and very pertinent to patients and their carers. To date, there have been no methodologically robust studies evaluating Quality of life (Patient Reported Outcomes) in patients undergoing extensive surgery to achieve tumour clearance. Patient reported outcomes are multidimensional health related quality of life measurements.

This study will builds upon a single site pilot study, evaluating short and medium term Patient reported Outcomes (PRO) at the Pan Birmingham gynaecological cancer centre. SOCQER-1 compared PRO Outcomes in 64 women with ovarian cancer, of whom 24 had extensive surgery; SOCQER-1 found no statistically significant difference in PRO in women undergoing standard or extensive surgery for ovarian cancer by 9 months postoperative period - however the limitations of a small sample size and a single centre study should be noted. SOCQER-2 will provide longer term PRO Outcomes and survival on a larger cohort undergoing extensive surgery and will capture variation in practice by centres. SOCQER-2 will provide good quality prospective evidence on the impact of extensive surgery on Quality of life and survival in patients with ovarian cancer in large, multiple UK gynaecological cancer centres.

STUDY DESIGN Patients will be identified at MDT/at clinic and approached during subsequent contact. A team member from the participating centres will provide an information leaflet, explain and obtain informed consent. Sites have the option of sending patient information sheets by post and patients also have the option of consenting by post. This study is intended to be an exploratory study. The underlying hypothesis being tested is whether patients undergoing extensive surgery have a 13% reduction (considered clinically significant difference) in Quality of life as compared to patients with a similar cancer burden at 6 months postoperative timepoint that is balanced by a 4 month survival gain in patients in the extensive surgery group.

The study is a prospective, multicentre observational cohort study. Clinical data will be captured prospectively using CRFs at baseline, surgery, postoperative and 18 months outcome. Patients will complete QoL questionnaires at 6 weeks, 6 months, 12 months, 18 months and 24 months postsurgery. For patients undergoing neoadjuvant chemotherapy, they will also complete a prechemo questionnaire

SAMPLE SIZE As this study is intended to be 'exploratory', the investigators have aimed to calculate a minimal sample size, with an assumption of 2:1 recruitment in extensive surgery: standard surgery groups.

A review of randomised control trials found that even quality of life was a primary endpoint, sample size was not always reported and that there was no consistent basis for these calculations. Analysis of repeated measures quality of life data should take account of missing data using an appropriate method for multiple imputation of missing items and an analysis that does not assume that data are Missing Completely at Random has been recommended. Therefore a mixed model is proposed for analysis of repeated measures data.

Regarding sample size for repeated measures quality of life data, common practice is to pick a single time point and calculate sample based on a single difference in group means, inflating sample size for dropout. This gives a conservative estimate of sample size in most cases. This is the approach adopted here, as although sample size calculations based on repeated measures models can be more accurate, in this case this approach is unlikely to provide more accuracy, given uncertainties as to the final group sizes for this observational study. The investigators assume that there will be maximal quality of life difference at 6 months and expect on the basis of clinical estimates that the ratio of group sizes for very extensive v standard surgery of 2:1. The investigators have assumed p<.05 and power of 80%. It has been suggested that an evidence based medium difference in the EORTC QLC 30 is 13 points. The mean score in stage II-IV ovarian cancer patients at baseline in the EORTC scoring manual is 56.3 (SD 24.5), while six month follow-up scores in a trial of neo-adjuvant versus primary de-bulking surgery was 73.1 (SD3.0) at six months in the primary surgery arm (n=201). The difference in Standard deviations may reflect the tight inclusion criteria for a randomised trial, and it might be assumed that in this observational study of surgical candidates would fall somewhere between the two figures. Assuming that QLC 30 at six months may be lower than this average those undergoing very extensive surgery at 66, but that the standard deviation is as reported in the scoring manual at 24 (likely to be a conservative assumption) and that a 13 point difference would be of clinical importance, a sample size of 123 (group 1=41 and group 2=82) would be required, with additional allowance for dropout or a comparable SD to that reported in the scoring manual (10 to 20 patients) (calculations made in Stata 13.1).

In the single centre SOCQER study, baseline EORTC QLC30 mean scores were 58.33 (SD 21.6) in the standard surgery group (n=32) and 63.1 (SD 25.6) in the extensive surgery group, figures broadly in line with the EORTC manual. The maximum difference between groups was 6 points in favour of the extensive surgery group 6 weeks postsurgery with smaller differences in favour of the standard surgery group at 3, 6 and 9 months. Although these differences did not at any point reach the suggested level for an evidence based medium sized difference in quality of life, given this is a single centre study which may incur selection bias between groups and there were some differences in stage and residual disease between groups, a study powered detection of a medium sized, rather than a small difference remains appropriate. Standard deviations are higher in this study than in the Greimal trial which in part may result from sample variation. Symptom scores however did show some differences which did not reach statistical significance in a mixed model, with a score (higher =poorer outcome) of 35.0 (SD 17.0) in the extensive surgery group and 25.0 (SD 17) in the standard surgery group. While symptom score should be considered secondary to quality of life, a sample size of 129 on the same assumptions as above would be adequate to demonstrate this difference plus additional allowance for dropout.

Although one approach to sample size calculations for subscales in RCTs involving quality of life outcomes is to adjust the alpha value to allow for multiple testing, in this observational study no further sample size calculations have been made as the investigators acknowledge that the purpose of analysis of subscale outcomes will be hypothesis generating.

However, this sample size does not allow for subgroup analysis, known confounders, loss to follow-up, the additional heterogeneity from different patient populations and treatment policies in the multiple centres in the study and, in the case of propensity score analysis, loss of observations that cannot be matched during the propensity score matching stage.

It therefore represents a minimum target not a maximum ceiling, and recruitment will continue for the full 12 month period. Recruitment of patients will take place over the course of 12 months. The high volume centres (>40 ovarian cancer surgeries per year) conduct an average of 62 surgeries per year.

STATISTICS All statistical analysis will be conducted in Stata 13. Differences between the "extensive" and "standard" groups for baseline variables will be compared using descriptive statistics and statistical tests appropriate for the level of measurement and sample distribution will be used. Outcome variables (composite quality of life and survival) for each of the groups will also be summarised using means or percentages as appropriate.

For outcome- As this is an observational study, residual confounding cannot be ruled out, and reporting of results will acknowledge this. For the outcome variables measured at multiple time points (specifically patient-reported quality of life), measures at each time point will be described, and repeated measures models will be used to reduce the probability of type I errors. Imputation techniques will be used where data on quality of life or other outcome variables is missing. Propensity score analyses will also be considered if informative scores and adequate matching can be achieved.

This observational study will provide information on current practice and outcomes that are likely to be used in national guidance, to inform future research and to provide information about previous patients experience to women undergoing ovarian cancer surgery; descriptive and exploratory analyses will be carried out to provide maximal information to inform policy and design future research. The investigators anticipate comparing QoL and other outcomes in three groups - women treated with standard surgery with no residual disease (less cancer load), those treated with standard surgery with residual disease (greater cancer load) and those with extensive surgery and no residual disease with greater cancer load and use of surgery with the aim of overcoming this).

Descriptive analyses may include:

• identify centres performing extensive surgery and the extent of surgery performed
• associations of specific components of ovarian cancer surgery with different aspects of quality of life;
• characteristics of patients undergoing "extensive" and "standard" therapy including performance status;
• descriptions of postoperative quality of life which might indicate opportunities for intervention around specific morbidities or in supportive care with the potential to improve patient's quality of life.
• Associations with PROMs.