Predicting patients at risk of developing gastrointestinal symptoms after treatment with pelvic radiotherapy
: a longitudinal observational cohort study

Student thesis: Doctoral ThesisDoctor of Philosophy


Over 350,000 people are diagnosed with cancer in the UK each year of whom over 21,000 women are diagnosed with a new gynaecological malignancy. Pelvicradiotherapy, often used to treat gynaecological malignancies, can result in long term changes in gastrointestinal function that impact severely on quality of life.There is no predictive model to stratify women at risk of gastrointestinal toxicity.Changes in the metabolic signatures of gut microbiome may be a viable way of identifying women at high risk of toxicity. 
The need for this study was supported by a prospective cohort describing the clinical setting of a specialist service assessing and managing people with long term GI toxicity following cancer treatment. A cohort of 220 women who accessed the service is described and highlights the complexity of the gastrointestinal symptoms these women experience (chapter two). A systematic review was undertaken to identify studies examining the gut microbiome in women treated for gynaecological cancer. No studies included analysis of the gut microbiome metabolites and the update of the review (chapter three) highlighted the emerging importance of the vaginal as well as the intestinal microbiome. A prospective cohort study was then used to recruit 41 women newly diagnosed with a gynaecological malignancy who received pelvic radiotherapy as part of their treatment (chapter five). Demographic and treatment variables were collected at six different time points with several samples- saliva, urine, stool, rectal swab-analysed using metabolic signature analysis to find a potential predictive framework. This included analysis by electronic sensing (e-nose) and the gold standard Gas Chromatography/ Ion Mass Spectrometry (CG-IMS). Particular attention was given to faecal calprotectin as a potential biomarker for long-term radiation induced gastrointestinal toxicity, using a generalised linear mixed model (GLMM). In addition, dietary intake, measured by 7- day food diaries, was analysed using a GLMM, to assess whether there were any changes in dietary intake overtime. Gastrointestinal toxicity was measured using the modified Inflammatory Bowel Disease – Bowel subset- questionnaire (IBDQ-B) as primary outcome.Statistical analysis included descriptive statistics and predictive modeling statistics. To test the ability of the olfactory analysis to predict GI toxicity, the samples of the participants were grouped in three groups according to their IBDQB scores and again in according to their stool consistency as measured by the Bristol Stool Chart (BSC) and included area under the curve, sensitivity, specificity,positive predictive value (PPV) and negative predictive value (NPV) analysis.
Two years after completing radiotherapy, 24 women completed the study. The median IBDQ-B score at baseline was 68.5 (range: 38-70). The median score was at its lowest (worst) at the end of treatment (median 58, range 23-66) but this improved by two years after completion of radiotherapy. There was no relationship between the IBDQ-B scores and age, Body Mass Index, the presence of lymphovascular space invasion, chemotherapy and radiotherapy dose.Performance status, tumour type and surgery correlated with the IBDQ-B scores but not two years after radiotherapy. The evaluation of faecal calprotectin levels at the end of treatment showed some evidence of correlation with IBDQ-B scores two years after radiotherapy (r= -0.45, p= 0.04) in this patient cohort but this effect was lost when when applying a bootstrap of 10.000 samples (p= 0.36). Overall, faecal calprotectin was a weak indicator of chronic radiation-induced GI toxicity(Pearson’s R = -0.20, p= 0.03) as measured by the IBDQ-B score but this effect was also lost when accounting for the different time points (p= 0.64). The food diary analysis did not show any statistically significant changes in weight, BMI, dietary fat, fibre, protein or carbohydrate intake. The olfactory signatures of the metabolites of the samples showed that the e-nose analysis was more effective than analysis using GC-IMS in most instances and for all samples in predicting GItoxicity two years after treatment. Stool, urine and saliva samples could be used to distinguish the IBDQ-B score groups at the beginning of radiotherapy using the enose analysis (p<0.05). Stool samples, saliva and rectal swabs were all able to distinguish the three stool consistency groups at the beginning of treatment using the e-nose analysis (p<0.05).
Whilst this study had a small cohort, the results showed that is possible to use a variety of samples at the beginning of treatment to predict GI toxicity, based on IBDQ-B scores and stool consistency, two years after completion of radiotherapy using the e-nose analysis technique. This could be developed further for application in clinical practice and identify women at risk of long-term GI toxicity following treatment for gynaecological cancer.
Date of Award1 Mar 2021
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorChristine Norton (Supervisor) & Ailsa Hart (Supervisor)

Cite this