Aims Biological therapy in IBD is increasing however, response rates remain modest. Raman Spectroscopy describes the scattering of inelastic light giving spectra that is highly specific for individual molecules revealing tissue biochemistry. Our aim was to establish spectral changes in IBD following biological therapy and whether Raman Spectroscopy can predict response.
Tab. 1
Intensity of Raman Shifts seen in pre- vs. post-biological therapy in UC and CD
|
Raman Shift (cm-1)
|
Pre biologic UC (Intensity)
|
Post biologic UC (Intensity) Cross-validation 0.95
|
Pre biologic CD (Intensity)
|
Pre biologic CD (Intensity) Cross-validation 0.94
|
|
1001
|
2.99
|
2.74
|
3.57
|
2.79
|
|
1302
|
1.34
|
1.52
|
1.11
|
1.53
|
|
1449
|
3.51
|
3.52
|
3.69
|
3.29
|
|
1656
|
1.99
|
1.72
|
1.82
|
1.30
|
Methods IBD patients undergoing endoscopic assessment pre- and 12 weeks post-biological therapy were recruited. Biopsies were taken for ex vivo Raman Spectroscopy analysis alongside biopsies for histological analysis. Response to treatment was defined as: endoscopic score (UCEIS ≤ 1/SES-CD ≤ 4) and histological healing (Nancy (0–1) in UC/modified Riley (0) in CD). We used artificial neural networks and supervised learning model to demonstrate spectral differences and build predictive modelling.
Results A total of 18 patients-7 UC/11 CD were included. Using data projection, there is clear separation between responder (3 UC/3 CD) and non-responders (4 UC/8 CD). Key spectral differences between pre- vs. post-biologic in responders are demonstrated using feature extraction (table). An increase at 1302 cm−1 may indicate a potential biomarker of healing.
A machine learning algorithm is able to differentiate between responders from non-responders with a sensitivity, specificity, NPV and accuracy of 100.0% (95%CI 93.5–100.0), 92.3% (95%CI 83.0–97.5), 100.0% and 95.8% (95%CI 90.5–98.6) respectively in UC and CD.
Conclusions We demonstrated changes in response to biological therapy and a potential biomarker for mucosal healing using Raman Spectroscopy. Using this modelling there is a potential to predict response to biological therapy, however prospective validation will not need to take place before clinical application.
