The Effect of Chemotherapy and Surgery for Cancer on Exercise Capacity

STUDY SUMMARY

The investigators propose a prospective blinded observational cohort study of patients undergoing NAC prior to elective upper gastrointestinal cancer resection (oesophagectomy and gastrectomy) in three NHS teaching hospitals. Exercise capacity (fitness) before and after NAC will be assessed using CPET and the relationship between the derived variables (AT, VO2 peak) and clinical outcomes will be described. Postoperative outcomes will be objective recorded clinically meaningful and will include mortality 1 year after surgery and short term postoperative harm described by the PostOperative Morbidity Survey (POMS) Quality of lifer using the EQ5D questionnaire and resource use (e.g. hospital bed utilization). The hypotheses will be tested by constructing predictive models incorporating exercise variables and factors known to be related to outcome following surgery (e.g. age, gender, open versus laparoscopic surgery). For example, for the Primary Hypothesis, models will be developed to predict the risk of one year mortality using (i) baseline exercise capacity and (ii) baseline exercise capacity and the relative decrease in exercise capacity. The aim will be to compare the predictive ability of both models to ascertain how prognostic the relative decrease in exercise capacity is, after adjusting for baseline exercise. In addition, Hypothesis 1 will be investigated by comparing AT before and after NAC using a paired t-test and Hypothesis 2 will be investigated by comparing the risk of 1year mortality between patients who stay in the same AT band and those who deteriorate using a chi squared test. Logistic regression will be used to repeat the analysis adjusting for potential confounders.

Study Conduct

This study will be undertaken within Southampton University Hospitals NHS Trust, Aintree University Hospitals NHS Foundation Trust, and University Hospitals, Bristol. All research staff will have evidence of up to date GCP training and investigators with direct patient contact will undergo a formal consenting course.

STUDY PROCESS

CONTACT 1: ELIGIBILITY AND RECRUITMENT (SURGEON \& STUDY NURSE) Consecutive patients will be screened for inclusion and exclusion criteria to establish eligibility for study recruitment. Patients will be approached in surgical clinics following listing for surgery plus NAC. Eligible patients will be approached and the opportunity to discuss the study requested. The study aims and conduct will be explained along with the process of informed consent. A period of 24 hours will be offered for the patient to reflect on their decision.

PROTOCOL

Study mentioned at initial outpatient appointment.

ELIGIBILITY AND RECRUITMENT (SURGEON \& STUDY NURSE

Preliminary eligibility determined MultiDisciplinary Team Meeting(MDT) Patient information and recruitment discussed at time of post-MDT investigation (endoscopic ultrasound). CPET booked if patient is considering proceeding with study. CPET canceled if decision made not to proceed.

CONSENT AND ENROLLMENT (STUDY NURSE) Patients will be given an additional opportunity to raise any questions about the study. Written informed consent will be obtained from patients choosing to take part in the study.

Baseline data will then be collected

• Patient Characteristics
• Quality of Life using EQ5D

BASELINE CPET: (CPET Physiologist/Scientist) Patients consenting for the study will perform symptom limited exercise using a ramped protocol on a cycle ergometer. This breath-by-breath system measures patients gas exchange, heart tracing (12-lead ECG) and oxygen saturation continuously. The protocol will be as follows: A rest period for 3 minutes and then a further 3 minutes of unloaded pedaling followed by the ramped exercise of 10-20watt ramp (depending on patients fitness) for a period of 8-12 minutes approximately.

POST NEOADJUVANT CHEMOTHERAPY CPET (CPET SCIENTIST)protocol as above. This will occur 4 weeks after the end of the last cycle of NAC (2 cycles for oesoghageal cancer and 3 cycles for gastric cancer)

PostNAC Generic Quality of Life EQ5D(Completion of NAC (YES/NO)

PREOPERATIVE HOSPITAL VISITS (STUDY NURSE) On admission to hospital Measure of perioperative risk: Oesophaogeal version of the Physiological and Operative Severity Score for the enumeration of Mortality and Morbidity (OPOSSUM) Score (preoperative components).

POSTOPERATIVE HOSPITAL VISITS (STUDY NURSE) Postoperative days 3,5,8 and 15. Postoperative components of OPOSSUM Score. Short-term postoperative harm: Post- Operative Morbidity Survey (POMS)(16) on postoperative days 3, 5, 8 and 15)

FOLLOW-UP DATA COLLECTION AT FINAL HOSPITAL DISCHARGE No Patient contact

FOLLOW-UP AT 30DAYS AND 1YEAR FOLLOWING SURGERY 12 month follow-up appointment Quality of Life (QoL)EQ5D

FOLLOW-UP ONGOING No patient contact

Date of death via National Medical Information Service (ongoing, see below)

DATA COLLECTION

1. BASELINE DATA

Patient characteristics: Study ID, Hospital number, Gender, Date of birth (calculated age), Height and Weight(calculated BMI and calculated ideal weight), Smoking (never, previous, now) Alcohol (never, minimal, moderate, heavy), Postcode (to obtain index of socioeconomic deprivation), blood sample (5 mls) for genetic analysis. (+/biomarkers + Haemaglobin).

•Quality of Life (QoL) EuroQol (EQ5D) (19)

2•CPET

VO2 at AT and VO2 peak. Additional CPET derived and Spirometry variables.

1. PREOPERATIVE HOSPITAL VISITS

Risk measures:

Variables and factors required to calculate Physiologic Score of OPOSSUM Score. Factors required to calculate RCRI

1. POSTOPERATIVE HOSPITAL VISITS Elements to calculate Operative Score of OPOSSUM (Post-Operative Morbidity Survey (POMS) on postoperative days 3,5,8, and 15 (see below) describing the presence of absence of morbidity in 9 domains

In-hospital death (date) FOLLOW-UP DATA COLLECTION AT FINAL HOSPITAL DISCHARGE AND BEYOND

Patient location throughout hospital stay (general ward, critical care) in order to calculate bed usage. Reoperation, readmission to critical care, readmission to hospital QoL Measures: EQ5D, at 30days and 1year after surgery.

Date of death from the National Medical Information Service (updated continuously).

JUSTIFICATION OF MEASURES Justification of predictor variables (AT and VO2 peak) The VO2 (oxygen consumption) at AT and VO2 peak are reliably measurable (using CPET) and widely accepted measures of exercise capacity (or fitness). There is no simpler, or cheaper method of reliably measuring exercise capacity (fitness) without using CPET. Both LT and VO2 Peak are known to be associated with outcome following major surgery.

Justification of other CPET variables Other CPET derived variables will also be recorded in order to explore the reasons for limitations in exercise capacity(fitness) in order to inform on what types of interventions might be effective to mitigate these changes.

Justification of risk measures O-POSSUM is a reliable and valid measure of perioperative risk which will provide a standard against which to compare the predictive performance of the CPET derived variables.

JUSTIFICATION OF POSTOPERATIVE OUTCOME MEASURES Justification of primary and secondary outcome variables

• Mortality

The primary outcome variable will be all cause mortality 1 year after date of surgery. The investigators chose this variable for several reasons, including that it:

• Is meaningful to patients, carers, administrators and policy makers.
• Is readily verifiable, and therefore reliable.
• Occurs with sufficient frequency to allow construction of a predictive model with a sample size that can be collected within a manageable timeframe.

Limitations of this variable are that the cause of death may not be directly related to chemotherapy or surgery but may arise due to disease progression or recurrence. However, all-cause mortality remains an important outcome to patients.

• POMS In order to explore the relationship between NAC, surgery and short-term harm the investigators propose to monitor short-term postoperative harm using the PostOperative Morbidity Survey (POMS). POMS is the only validated measure of short term postoperative harm16 and has been used in outcomes research and effectiveness research. It is in current use as a primary outcome variable in MRC and NIHR funded studies.
• Quality of Life (QoL) Measures using EQ5D The impact of NAC and surgery on patient quality of life will be assessed using A validated questionnaire. The investigators will assess subjective changes in patient's state using a a questionnaires. The questionnaires will be self-completed by the patient, taking about 15-20 minutes in total. The questionnaires to be used will be EQ5D which has been recommended for use as a generic PROM following major surgery.
• Resource Use In order to obtain a full picture of the clinical course and resource use of the study patients, additional outcome variables will include: Reoperation, Readmission to hospital, length of post-operative hospital stay, readmission to hospital, total postoperative hospital bed usage, length of postoperative critical care stay, readmission to critical care, total postoperative critical care bed usage.

SAMPLE SIZE The hypotheses listed in (Aims of the Project) will be tested by constructing predictive models incorporating exercise variables and factors known to be related to outcome following surgery (e.g. age, gender, open versus laparascopic surgery). For the primary hypothesis, models will be developed to predict the risk of one year mortality using (i) baseline exercise capacity and (ii) baseline exercise capacity and the relative decrease in exercise capacity.

The aim will be to compare the predictive ability of both models to ascertain how prognostic the relative decrease in exercise capacity is, after adjusting for baseline exercise. To develop reliable prediction models approximately 252 patients are required. This is using the "Rule of 10" and assumes that the one year mortality rate is 20% (conservative estimate) and that models contain 5 factors (e.g. age, gender, centre, location of tumour, laparoscopic versus open).

Given a 20% non-completion of NAC and 2 CPET tests (based on Liverpool data), this will be achievable since approximately 525 patients will attend the centres over the 25month recruitment period (based on average of last three years patient throughput, with an anticipated recruitment rate of 60%.

Sample size for Hypothesis 1: A sample of 152 patients would be required to detect a difference in 1 [unit] of VO2@LT using a paired t-test at the 5% significance level with 90% power. This is assuming the standard deviation of the difference in VO2@LT values is 3.8 [units].

Sample size for hypothesis 2B: A sample size of 242 is required to detect a difference in one year mortality rates of 15% (30% versus 15%) between the two LT change groups (no change / deteriorate) using a chi squared test at the 5% significance level with 80% power. This is assuming that there are equal numbers of patients in both groups.

ANALYSIS PLAN

1. Relationship between exercise capacity changes and clinical outcome.
2. Effect of chemo on exercise capacity and other CPET variables. Patients who do not complete NAC and 2 CPET tests will be eligible to enter the study and their data collected. The sample size calculation and primary analysis will be of patients who complete CPET before and after NAC. Secondary analyses will include patients who were unable to complete CPET before and after NAC, and will compare patients who underwent surgery and those in whom surgery was subsequently canceled.

Descriptive analyses (means and proportions) will be carried out to describe the patients' characteristics including baseline and changes in exercise capacity, and their clinical outcomes including 1year mortality, morbidity and PROMS.

The calculation for hypothesis 1 is based on a comparison of fitness (using VO2@AT) before and after NAC. As this is a within-patient comparison the investigators use the paired t-test. The standard deviation for the within-patient difference is assumed to be 3.8. This was estimated from the pilot data in the Background and assumes a correlation of 0.5 between before and after values. The clinically important difference of 1 was considered conservative and feasible since the pilot data had a difference of 2.1. The calculation for hypothesis 2B is based on a dichotomisation of patients in two groups: a) those who deteriorate after NAC b) those who don't. To compare mortality rates between the two groups the investigators would use a chi squared test(although a logrank test may be appropriate if dropout is an issue). It's assumed that the two groups will be approximately equal (this is discussed elsewhere) and that the mortality rates for the two groups are 30% and 15% respectively.

The sample size calculation for hypothesis 2/2A is not based on power but instead is based on the amount of data required to develop a robust and reliable risk model. The idea is to develop a risk prediction model based on preNAC factors and quantify its predictive ability (e.g. using the ROC area). McCulloch (BMJ, 2003) does something similar. Next the investigators will add postNAC fitness to the model and establish whether this extra data provides extra predictive ability.

The usual approach to calculating the sample size for a risk model is based on the rule of 10, that is the investigators require 10 times as many (mortality) events as there are predictors in the model. Our models will have 5 factors so the investigators require 50 events. Assuming a mortality rate of 20% suggests that the investigators require 250 patients in total. Note that this does not give us specific power to detect some predefined clinical effect (this would be a very difficult calculation since pre- and post-NAC fitness are likely to be correlated) but rather gives us enough data to fit risk prediction models to quantify the incremental benefit of post-NAC fitness.