CPET and Outcome After Oesophagogastric Cancer Surgery

Study overview and background

We hypothesise that reduced physical fitness (measured objectively using Cardiopulmonary Exercise Testing (CPET)) will result in worse post-operative surgical outcome (reduced overall survival, and increased in-hospital morbidity measured by Clavien-Dindo-Demartines Score). We propose multicentre patient recruitment and analyses of routinely collected clinical data, in an attempt interrogate this hypothesis in patients prior to upper gastrointestinal cancer surgery. Patients undergoing cancer therapies pre-operatively (neoadjuvant chemotherapy or chemoradiotherapy) will be included in the analyses. Patients that become ineligible for an operation, mainly due to metastasis or progression of their cancer on neoadjuvant cancer treatments will also be included but will be analysed separately.

Despite advances in anti-cancer therapies, surgery remains the most important treatment for oesophageal and gastric cancer. Major UGI cancer surgery carries substantial morbidity and mortality, particularly in elderly patients and those with co-morbidities. Recent UK cancer audits show 30-day mortality of up to 3% for elective surgery and a 90-day mortality rate of 5.2%. Outcomes after major cancer surgery depend on modifiable factors such as peri-operative medical care and intra-operative technique and what were previously thought to be unmodifiable components, such as the ability of the patient to tolerate surgical trauma. Centralisation of services and optimisation of perioperative pathways have led to improved outcomes, but substantial gains may yet be achieved with accurate and objective preoperative risk prediction.

Current approaches to risk prediction include a plethora of prediction scores and tests, but none have proven efficacy above experienced clinical acumen. More recently reduced physical fitness (objectively measured using CardioPulmonary Exercise Testing - CPET) has been associated with increased postoperative morbidity in a variety of major surgical cohorts. CPET provides an objective method of evaluating physical fitness under stress, partly mimicking the stress of a major surgical event. Accurate risk prediction methods prior to surgery or indeed other forms of 'trauma' like pre-operative cancer therapies might allow better patient selection, permit modification of the patients preoperative status, as well as improve the optimisation of intra- and postoperative management, and enhances shared decision-making. Furthermore CPET allows interrogation of the causes of exercise intolerance, when exercise capacity is reduced, with the exciting opportunity of intolerance modulation for therapeutic gain (Fit for Surgery).

To date, 37 cohort studies (including over 7,800 patients) have reported the relationship between preoperative CPET-derived variables and postoperative outcome. These data have been brought together in several systematic reviews that, in summary, show a remarkably consistent positive correlation between physical fitness and postoperative outcome. The Southampton based Fit-4-Surgery group have clearly demonstrated this in patients undergoing major colorectal surgery, however the relationship between fitness and post-operative outcomes in UGI cancer surgery is not yet well defined.

METHODOLOGY

Patients in all the recruiting centres will undergo a maximal CPET prior to major UGI cancer surgery (defined as a procedure involving oesophageal, junctional or gastric cancer resection) according to the CPET protocol defined in Appendix 1 of protocol. If the CPET deviates from this protocol, the centre would need to clear their local CPET protocol with the Chief Investigator (CI). All variables measured by CPET that will be collected are outlined in Appendix 2 of protocol. If the patient is scheduled to have neoadjuvant cancer therapies, the recruiting site might elect to test patients pre- and post the neoadjuvant therapy to estimate the change in fitness. This is based on changes observed in fitness with neoadjuvant chemotherapy in UGI and chemoradiotherapy in rectal cancer patients.

Following surgery only routine clinically relevant observational data will be collected. These data will relate to hospital length of stay, the level of care required following surgery, post-operative morbidity (Postoperative Morbidity Survey - POMS and/or Clavien-Dindo-Demartins Score) and the recovery process. Most of this information can be accessed from prospectively maintained enhanced recovery databases, however electronic patient records that are maintained and recorded prospectively are also acceptable alternatives.

Each site would also contribute existing data held on the hospital patient database in addition to recruiting patients as they present at the hospital as per methodology above. All data from each individual hospital will be anonymized at source by the PI prior to transfer to the CI and any analyses. CPET variables, post-operative outcome data, and basic patient demographics (outlined in Appendix 2 of the protocol) cannot be linked back to the original patient.

All patients undergoing CPET at each NHS site are given a patient information sheet related to risks and benefits related to CPET and routinely undergo written consent prior to testing, also asking for their permission for these data to be used for research purposes.

Contraindications to CPET based on ACCP/ATS Guidelines. (American Journal of Respiratory and Critical Care Medicine 2003; 167: 211-77)

Please see also for further information:

http://www.thoracic.org/statements/resources/pfet/cardioexercise.pdf http://www.pcpet.co.uk/Links

HYPOTHESIS

This study tests the hypothesis that selected CPET variables are related to overall survival (1 and 3-year) and to in-hospital morbidity in patients undergoing major UGI cancer surgery.

Our primary aim is to establish a reliable relationship between 1-year overall survival (present or absent) and oxygen uptake (Vo2) at peak exercise (Vo2 Peak); a secondary aim is to explore the multivariable relationship between Vo2 at the estimated lactate threshold (AT), other selected CPET variables (weight, BMI and Body Surface Area (BSA) adjusted) and other important prognostic variables with post-operative complications in an attempt to risk stratify patients before major surgery.

3. DATA ANALYSES

Cardiopulmonary Exercise Tests (CPET)

The CPET involves cycling on an exercise bike for 8-12 minutes. Starting with a very low resistance on the pedals the patient will pedal at 60 revolutions per minute. After 3 minutes of cycling, the resistance will gradually increase until the patient can no longer turn the pedals at the required speed. The test will be thoroughly explained to the patient beforehand and trained staff will make the experience as comfortable as possible. CPETs will be stopped early by the study researcher/doctor in the presence of any adverse events. Each CPET appointment will last approximately one hour.

CPETs carry a small risk (1:10,000) of exercise induced myocardial infarction or dysrhythmia. Patients are continuously monitored using a 12 lead electrocardiograph (ECG) during all tests and recovery. The CPET test will entail the patient becoming slightly hot or maybe sweaty, and appropriate clothing should be worn. The patient's heart will be monitored by an electrocardiogram. The patient will wear a soft rubber mask in order to continuously sample expired air with an online breath-by-breath gas analyzer. This may cause some mild claustrophobia. Full detail of the CPET procedure is given in appendix 1. Accredited doctors or allied health professionals with experience in analysis of exercise tests will analyse all of the exercise data. Patients will be anonymised and the person performing the data analysis will be blinded to the outcome data. All data will be stored, processed and analysed on secure NHS computers on NHS encrypted and password-protected servers.

4. STATISTICAL METHODOLOGY

Non-parametric receiver operator characteristic (ROC) curves will be constructed for Vo2 at AT, Vo2 at Peak, O2 Pulse at AT and ventilatory equivalents for carbon dioxide (VE/Vco2 at AT) in order to assess their independent ability to discriminate between patients with and without post-operative morbidity. All selected CPET variables will be adjusted for weight, BMI and BSA. Optimal cut-points will be obtained by minimising the distance between points on the ROC curve and the upper left corner. Variables (to satisfy the 10 events per variable rule) will be identified as candidates for a multivariable logistic regression model including but not limited to: Vo2 at AT and at Peak, gender, operation type (laparoscopic/open), and O2 Pulse at AT and VE/VCo2 at AT. A final multivariable model will be obtained using forward stepwise selection (minimising Akaike Information Criteria (AIC)). Its sensitivity to variable exclusion and re-inclusion will also be assessed using AIC. Model fit will be assessed using the Hosmer-Lemeshow goodness-of-fit test. In order to explore the univariate relationship between CPET variables and length of stay, continuous CPET variables will be dichotomised at the optimal cut-point for the ROC curve and Kaplan-Meier curves will be constructed. The log rank test will be used to compare survival curves; patients who died before discharge will be treated as right-censored. All analyses will be conducted using Stata (StataCorp. 2011 Stata Statistical Software: Release 12. College Station, TX: StataCorp LP.). Continuous variables will be reported as mean and standard deviation (SD) or median and inter-quartile range (IQR) depending on the distribution. Categorical variables will be presented as frequency (percentage). P-values will be obtained using univariate logistic regression (continuous) and Chi-squared or Fisher's Exact tests (categorical). Statistical significance will be taken at 5%.

We anticipate an overall death rate of 33.5% in 12 months post-surgery. We assumed that the death event rate would increase in the unfit (control) patient group up to 50% while the fit group would show the same average of 33.5% rate thus showing a difference of at least 16.5%. We consider that this size of the effect (50% reduced to 33.5%) is clinically relevant. Assuming 5% (2-sided) significance and 90% power, 150 events are needed to detect this size of difference in a log-rank survival comparison of the two trial groups. This requires 376 patients in total in the analysis. We will inflate this total by approximately 20% to allow for attrition, giving a target recruitment of 470 patients (with complete data on CPET variables and outcome).