Digital patient-reported outcomes and serological response to SARS-COV-2 vaccines in patients with cancer.

Abstract

e13606 Background: SARS-CoV-2 vaccines have changed the course of the current global pandemic. Cancer patients were identified as highrisk of adverse infection outcomes. We have previously characterised the serological response to SARS-CoV-2 vaccines in 220 cancer patients treated at our institution. In addition these patients were given the possibility to report their symptoms (patient-reported outcomes, PROs) weekly using a digital platform (ePROs). We sought to determine if, in cancer patients, the prospectively recorded post-vaccination ePROs could predict the serological response to SARS-CoV-2 vaccines. Methods: We used a pre-existing digital platform that allows monitoring of PROs using weekly questionnaires sent to patients and available on their desktop computers, tablets or smartphones. Serial serologies were performed at 28, 50 and 115 days after vaccination. Results: We observed that at day 50 after the first vaccination dose, coinciding with three weeks after the second dose, patients could be divided into two groups according to their serological response (low – below 1500 U/ml and high – above or equal 1500 U/ml). A peak in symptom burden could be observed after the second dose, as previously described. Omitting ePRO features decreased prediction performance of all models, whereas omitting baseline symptom scores had inconsistent effects. Among all models and feature constructions, the top performance metrics were given by the nearest centroid model7 with baseline symptoms omitted and 20 features chosen with the aforementioned procedure. The model achieved an accuracy of 0.704, an F1-score of 0.759 and an MCC of 0.398. Conclusions: we were able to identify the patients who achieved higher antibody levels against SARS-COV-2 based on the symptom burden reported through ePROs. This represents the first model showing that symptoms, assessed through ePRO can be predictive of response to vaccines. Our results could also be useful information for patients, as they could assuage their fears about adverse -effects, through the knowledge that toxicity could predict better protection against SARS-COV-2. The same toxicity-based prediction of efficacy has been identified with immunotherapy in cancer and is now a routine part of clinical discussions with patients.