GERiatric Screening in the treatment of elderly patients with Ovarian Carcinoma (GERSOC): study protocol for a pragmatic, cluster randomised controlled trial

Patients aged 70 years or older who are diagnosed with primary ovarian carcinoma (including fallopian tube and peritoneal carcinoma) FIGO stages II, III or IV are eligible to participate in the study. Patients with a high suspicion of advanced-stage ovarian cancer, in whom diagnosis is not yet pathologically confirmed before primary CRS, can also be included. A high suspicion of advanced ovarian cancer should then be based on imaging (i.e. signs of pelvic, abdominal or extra-abdominal metastases). Patients must be able to complete a Dutch questionnaire and written informed consent has to be obtained before inclusion in the trial. Patients who cannot read or write Dutch will be excluded from participation in the study, as they will not be able to complete a Dutch questionnaire.

Patients can refuse to participate or leave the study at any time, without any consequences. In case a patient chooses to withdraw from the study, the collected data will still be used for the study as described in the patient information leaflet and signed informed consent form. Patients with a high suspicion of advanced-stage ovarian cancer who are included prior to surgery are withdrawn from the study when the diagnosis of advanced ovarian cancer is not confirmed by pathology after surgery. Replacement of withdrawn patients will only occur if the aforementioned histopathological criteria for withdrawal are met.

Cluster randomisation was considered the most suitable randomisation method for this study for two reasons. First, we believe that individual randomisation is not appropriate because health care providers then have to alternate CAU and GSC between patients. By performing geriatric screening in patients randomised to GSC, the health care giver will become more aware of possible geriatric risk factors and may subsequently recognise these factors in patients randomised to CAU. It is very likely that such a geriatric screening approach will influence usual care and thus study results. Second, when patients randomised to CAU become aware that other patients were randomised to GSC and receive geriatric screening tests, they may become dissatisfied with their usual care. This could lead to biased results on quality of life.

In our trial, clusters consist of hospitals involved in diagnosis and treatment of ovarian cancer. These hospitals will be randomised by the coordinating study centre after local approval is obtained. Randomisation to either CAU or GSC will be performed in the web-based randomisation software ALEA (version 17.1) using minimisation (with a 75% probability that the underrepresented arm is selected in case of imbalance).

In the Netherlands, CRS for ovarian cancer is centralised in gynaecological oncology centres. It is possible that these centres will include more patients than hospitals that do not perform CRS. To prevent imbalance between the clusters, we will perform a stratified randomisation based on gynaecological oncology centre, defined as a hospital that performs at least 20 CRS procedures for primary ovarian cancer per year.

Following the nature of the intervention, allocation of the hospitals will not be blinded. The Medical Research Ethics Committee (MREC) determined that the patients allocated to GSC must be informed about randomisation on hospital level. Due to these restrictions, the geriatric screening intervention will not be introduced as ‘standard care’ for all patients in hospitals randomised to GSC, but will only be performed in patients who have signed informed consent.

Patients will be recruited by their gynaecologic oncologist, medical oncologist or nurse (practitioner). Allocation is based on the hospital where the patient is treated. Eligible patients receive a short briefing and a leaflet with information about the study. The patient information leaflet in CAU will be generic, stating that we investigate the quality of life of older ovarian carcinoma patients. The patient information leaflet in GSC will contain information about the use of geriatric screening and possible referral to a geriatrician. After obtaining written informed consent, participating patients will be registered in ALEA and subsequently assigned a unique study number for further data handling. Eligible patients who do not want to participate in the study will also be registered, including the reason for refusal.

When the hospital is randomised to CAU, medical care will be provided according to the current guidelines [4]. As the Dutch guidelines for ovarian cancer do not provide any specific advices for care of the elderly patient, decisions of standard or adjusted treatment are based on the treating physicians opinion and multidisciplinary consultation. In general, the final treatment plan will be defined in the multidisciplinary team meeting. If considered necessary, patients may be referred to other specialists for further evaluation at the discretion of the treating physician. Implementing CAU will not alter the current clinical pathways, as usual care in the participating hospital will be continued and no intervention is introduced. All concomitant medication and care within and outside treatment for ovarian cancer is allowed.

In hospitals randomised to GSC, all participating patients will receive a geriatric screening after (suspected) diagnosis of advanced ovarian cancer. The geriatric screening consists of the G8 questionnaire and the TUG [19, 21]. Patients determined unfit on either or both of the tests will be referred to a geriatrician for CGA. The geriatric screening as well as the CGA will be performed before a treatment strategy is defined. At the start of the study, the involved health care professionals will receive a brief instruction on the execution of the geriatric screening. Deliberate deviation from the treatment advice is permitted if the treating physician considers this necessary. Implementing GSC will alter the current clinical pathway of patients who participate in the study, as they will receive a geriatric screening and will possibly be referred to a geriatrician for additional treatment advice. All concomitant medication and care within and outside treatment for ovarian cancer is allowed.

Clinical, demographic and socio-economic data will be collected from medical charts. In addition to basic clinical data, we will focus on comorbidity, geriatric risk factors and geriatric assessment. These data will be collected in a web-based electronic case report form (ALEA, version 17.1) and will be stored in coded form using the unique study number. Clinical data will be collected up to two years of follow-up.

Cancer-specific HRQL will be measured using the EORTC QLQ-C30 questionnaire [24], designed and validated specifically to measure quality of life in cancer patients. It comprises functional and symptom scales, a scale on global health and quality of life and several single item symptoms. This core instrument will be complemented by the ovarian cancer specific module EORTC QLQ-OV28 [25]. This questionnaire focusses on symptoms and side effects specific for ovarian cancer.

For the cost-effectiveness analysis, utilities will be retrieved to derive QALY’s by means of the EuroQol-5D-5L (EQ-5D-5L) questionnaire [26]. Utilities reflect the preferences of society for length of life corrected for the quality of those life years. Direct healthcare cost data will be collected from the hospitals administrative database. Indirect costs data will be derived bottom-up from participating health care providers and using an abbreviated version of the institute for Medical Technology Assessment’s Medical Consumption Questionnaire for costs outside the hospital (iMCQ) [27]. Patients are also asked to report additional demographic and socio-economic data.

All patient-reported outcomes will be collected before start of treatment and 6, 12 and 24 months after diagnosis. Patients will receive the first questionnaire from the nurse or treating physician. Patients will return the completed questionnaire in a pre-stamped envelope to the coordinating study centre. The subsequent questionnaires will be sent to the patient by the central study coordinator and can be completed on paper or online after secured login (www.​profielstudie.​nl). Non-respondents will be sent a reminder letter and questionnaire by mail or e-mail after four weeks.

The primary outcome ‘percentage of completed standard and adapted treatment’ will be compared between the intervention and control group. Comparisons of the percentages will be analysed using logistic regression, adjusting for differences in patient and hospital characteristics at baseline. Secondary outcomes, such as treatment adjustments during follow-up, will be compared between the two groups using logistic regression analyses. PFS, OS and DSS will be estimated using the Kaplan-Meier method. Conventional survival analyses do not allow us to account for the loss of independence that potentially follows from clustering of patients within one centre. Therefore, differences between the groups will be compared using Cox regression models with mixed effects: this approach modifies a Cox proportional hazards regression model by incorporating cluster-specific random effects that modify the baseline hazard function [29].

We will use a multilevel modelling approach [30] to investigate differences between the intervention and control group. This approach is appropriate to account for the clustering at hospital level [31]. Repeated measures analysis using generalized estimating equations will be carried out to account for the intra-patient dependency of the repeated measures [32]. In addition, differences in effect of the geriatric screening and usual care at the different time points will be evaluated [32]. Missing outcomes will be assumed missing at random. An advantage of the use of a multilevel modelling approach is that all patients can be included in the analyses, regardless any missing follow-up measurements. Clinically meaningful differences in HRQL scores will be based on previously published medium/large minimally important deteriorations in the EORTC QLQ-C30 scales [33]. Data on patient and tumour characteristics requested from the Netherlands Cancer Registry will be used to compare the group of respondents, non-respondents and patients with unverifiable addresses, using t-tests for continuous variables and Chi-square analyses for categorical variables.

Health Technology Assessment (HTA) consists of two parts: First, a scenario analysis to anticipate on barriers and facilitators for implementation will be effectuated. This analysis consists consecutively of the identification of (dynamic) aspects having impact on adoption; a brainstorm on possible scenarios by informal interviews with experts; a scenario construction; validation of the scenarios by means of semi-structured interviews with experts and finally, a quantification into parameters for cost-effectiveness modelling. A cost-effectiveness analysis (CEA) will be performed, with incremental costs per QALY gained as most important outcome.

For the CEA, a Markov model will be constructed with four mutually exclusive health states: disease free survival, recurrence, distant metastasis and death. Using a monthly cycle length, the model will simulate the course of events in a hypothetical cohort of 1000 patients with an average age of 75 years and stage III ovarian cancer. A societal and hospital perspective, plus lifelong time horizon will be adopted. The gathered direct and indirect costs from the trial will be included. Subsequently, costs for treatment and treatment of recurrences, follow-up and palliative care will also preferably be based on the trial, but in case of insufficient numbers, these will be based on literature.

Mostly, trial data on costs are not normally distributed, therefore the non-parametric Mann-Whitney U-test will be applied, with a two-sided significance level α = 0.05. Differences in treatment arms will be tested using the Student's t-test; the paired t-test will be used for differences in time.

Incremental cost-effectiveness ratios will be calculated and stochastic parameter uncertainty handled probabilistically (10,000 Monte Carlo simulations). Distributions will be assigned to the parameters. According to the Dutch guideline [34], future costs and effects will be discounted to their present value by a rate of 4% and 1.5% per year respectively. The cost-effectiveness model will be tested using (deterministic) sensitivity analyses. The results of the probabilistic sensitivity analysis will be illustrated in a cost-effectiveness plane, and decision uncertainty will be shown by cost-effectiveness acceptability curves [35]. The latter will show the probability that a pathway has the highest net monetary benefit, and thus is assumed cost-effective, for a range of Willingness to Pay values for one additional QALY (the ceiling ratio). In this analysis, we will use the Dutch (informal) ceiling ratio of € 80,000 per QALY [36]. A budget impact analysis will be performed following the ISPOR guidelines [37]. We will use hospital perspective and a 5-year time horizon. For all other parameters, the input parameters of the CEA will be used.

The project management group from the coordinating centre (The Netherlands Cancer Institute) will include and randomise participating hospitals, assure (local) ethical approval, initiate the study at all sites, provide instruction on geriatric screening, address questions concerning the trial, collect the data, complete follow-up and analyse the data. A list of study sites can be obtained from the coordinating centre as well. The steering committee will oversee inclusion rates, data collection and data analysis and propose interventions if necessary. The project management group will meet every month to review trial conduct and the steering committee will meet once a year. A report on the progress of the trial will be submitted to the accredited MREC once a year.

Concerning the need for a Data Monitoring Committee (DMC), the Dutch Central Committee on Research Involving Human Subjects (CCMO) refers to the European Medicines Agency (EMA) guideline on data monitoring committees [38]. Although advanced stage ovarian cancer is a life threatening disease, the intervention in this trial does not concern a pharmaceutical product or device, is well characterised and not considered being potentially harmful to patients. The G8 questionnaire and the TUG are validated and commonly used for the identification of frail elderly patients. The intervention is not blinded and is performed only once per patient. Consequently, this trial does not include a DMC. The trial protocol does not describe standardised reporting on (Serious) Adverse Events ((S)AEs), as we do not expect to that participation in the study or receiving a geriatric screening will result in any (S)AEs. However, if unforeseen and unintended adverse effects occur and are spontaneously reported to the study team, these will be assessed and reported to the MREC if necessary. In the absence of a DMC and considering the trial is unblinded and that no safety issues are expected, no interim analyses or formal stopping rules are established. The absence of interim analysis, standardized (S)AE reporting and a DMC in this trial is approved by the responsible MREC.