Fecal calprotectin as a predictor of gastrointestinal immune-related adverse events (CF-19): A prospective study.

1. Introduction

Immunotherapy has reached oncology to stay, and a new class of immune-mediated adverse effects (irAEs) has emerged, colitis being one of the most frequent.^1,2 It is still unknown if any biomarker may predict irAEs onset, even though they resemble classic autoimmune diseases and some data suggests that some of the parameters that we use in this scenario could be useful in predicting this kind of toxicity.

For example, a phase I study found that adding BCG to ipilimumab in stage III-IV melanoma (n = 15), describes a remarkable increase of auto-antibodies (ImmunomeTM protein array) just before the development of severe irAEs.³ However, there are not prospective studies designed to validate this observation. Also, a retrospective study from a French group describes that CRP (C-reactive protein) increase and hemoglobin and albumin decrease are the main alterations that can be observed in patients diagnosed with immune-mediated colitis during anti-CTLA4 treatment (n = 27).⁴ Furthermore, during ESMO (European Society for Medical Oncology) 2018 Congress some groups showed that the increase in neutrophil/lymphocyte ratio decreases immune-related colitis risk. It was also found that decreasing absolute lymphocyte count and increasing monocytes and eosinophyles correlates with a irAEs risk increase (n = 130).⁵ However, all these parameters could be affected by other illnesses that may be associated with our oncological patients and for that reason, they do not seem sufficiently conclusive to change our daily-clinical practice.

Focusing our attention on gastrointestinal toxicity and on the French study mentioned before,⁴ they concluded that endoscopic lessons from immune-related colitis were very similar in comparison with the ones that appear during inflammatory intestinal diseases like Ulcerative Colitis or Crohn’s disease, some also with extra-intestinal manifestation associated. In eight patients of the study, fecal calprotectin (FC) was analyzed, and it was remarkably high in all of them (1755 μg/g [299–12900]), just like in patients diagnosed with intestinal inflammatory diseases.⁶ Furthermore, their condition is improved if we use the same treatments that are approved for this illness (corticosteroids, infliximab, vedolizumab).^7-9 Nevertheless, Berman et al.¹⁰ cast FC aside as a predictive tool based on the data of Weber’s study,¹² where FC is measured in two groups (ipilimumab + placebo vs ipilimumab + rectal budesonide), but this data could have been affected by the use of corticosteroids in one of the groups. Thus the evidence in the literature is contradictory.

We know that the use of CF as a predictor for intestinal inflammatory disease relapse is globally accepted and may diagnose relapse 3 months before the onset of symptoms (S 78%, E 73%).⁶ However, there are no prospective studies that analyze FC value as gastrointestinal irAEs predictor.

Taking into account that immune-mediated colitis could be endoscopically compared with inflammatory intestinal diseases, we hypothesize that FC may be useful to predict severe immune-related colitis.

2. Methods

3. Results

24 patients were included, 19 of those were male. The primary origin of the tumors included were, from higher to lower frequency: lung (n = 15), head and neck, kidney and bladder (n = 2, each), and colorectal and melanoma (n = 1, each). Almost all of the patients received monotherapy with anti-PD-1/PDL-1 (n = 18) -nivolumab, pembrolizumab, atezolizumab, durvalumab-. Four of them, in combination with chemotherapy (carboplatin or cisplatin with pemetrexed and pembrolizumab). Two patients received anti-CTLA4, ipilimumab, in combination with nivolumab.

For 15 patients, immunotherapy was the first line of treatment for their pathology (2nd line n = 7, and 3rd line n = 2), so, for the majority of subjects included, previous treatments could not have disrupted our results by affecting their microbiome.

Only two patients stopped their treatment due to toxicity (one of them because of immune-mediated colitis). Other reasons to stop immunotherapy were progressive disease (n = 4), death (n = 2), COVID-19 (n = 1), and social issues (n = 1).

Focusing our attention on the main objective of this study, we registered three grade 1 colitis and two ≥ grade 2 (one grade 2 and other, grade 3). All colitis observed appeared during anti-PD1 treatment, always before cycle number six, according to the already published temporality, that suggests immune-related colitis peaks between 8-12 weeks.

No patient was using NSAIDs before being included in this study (according to the exclusion criteria). However, four of them took NSAIDs during immunotherapy treatment, and one of them developed grade 1 colitis. The rest of the patients did not suffer from toxicity, but we registered modifications on their FC values during NSAIDs administration.

The ROC curve (Figure 1) presents an AUC 0,559 (IC95%:0,32-0,798) and colitis RR (Relative Risk) taking into account that FC value is 1,001 for each 10 units (p = 0,493). Furthermore, Figure 2 shows the FC value distribution comparing patients who have developed colitis with patients that have not. We did not find a statistical difference. In group 1 (the ones that did not develop the condition), FC median value was 63,5 mg/kg (31,9-172). In contrast, in the group of patients that developed colitis, FC median value was 124 mg/kg (5-1119). We could not find an explanation for the extreme value that was registered for one of the patients who, in addition, did not develop severe immune-related colitis.

Figure 1. ROC curve of FC as predictive value for immune-related colitis (AUC 0.559).

Figure 2. FC distribution value comparing group 1 (no immune-related colitis) with group 0 (immune-related colitis developed).

4. Discussion

Taking into account the already published data about FC and the urgent need to find biomarkers that may predict immune-mediated toxicity, this study seemed promising. Also, it is the first prospective study that analyzes this issue, and we expected to find a biomarker that allowed us to not stop potentially effective treatments^14–16 for patients with limited options. However, some factors may have contributed to not having reached our main objective.

First of all, it is true that the small sample size recruited may reduce the power of the study, and having completed the estimated size and having included patients in all the centers that we planned, would have been extremely interesting to observe the number of immune-related colitis and to show stronger conclusions. However, if we analyze the figures above we cannot find any trend towards statistical significance, and we are afraid that increasing the number of patients included would not produce more significant results. Furthermore, if we analyze the only patient that was diagnosed with severe immune-related colitis we can see that FC was not only not increased before the development of the toxicity, but also, it was even decreasing until very close to the event (Figure 3).

Figure 3. FC evolution in patient diagnosed from grade 3 colitis on 01/02/2020.

In addition, and after analyzing Figure 3 we find one of the other big problems that we faced during the development of this study was lost values. Oncological patients, due to either their illness, related symptoms, or and palliative treatment as opioids, suffer from constipation. Usually, patients were not able to comply with the timeline, and sometimes, they decided not to bring us the sample because they found it excessively disgusting even though we tried to transmit its importance. These lost values could also have affected to the final results.

Secondly, the exclusion criteria of “not being under NSAIDs” was very problematic during recruitment. The vast majority of potential candidates were under this kind of treatment and taking into account the complexity of oncological patients and that pain is usually their main symptom, NSAIDs are a useful treatment that we cannot easily avoid. FC value is easily altered in this context as we showed before and because of that reason, it was considered to be an exclusion criterion. However, and despite remembering its importance in every visit, some patients started NSAIDs during the course of this study. We registered and analyzed this fact, but luckily, it did not significantly affect our results.

Another fact that may be related to the high variability of FC in our study is the microbiome as well as inflammatory activity related with tumors and with immunotherapy,¹⁷ and its uncertain evolution. There are different ongoing studies that are trying to analyze if the microbiome can modulate or even improve the efficacy of some treatments.^18,19

To sum up, and taking into account the huge complexity that we had to confront to include patients, the extensive variability of FC, especially in oncological patients, and also, the high number of lost values, we consider that extending this study would not contribute to clarifying results in favor of FC as a predictor biomarker for immune-related colitis.

5. Conclusion

Even though we must take into consideration our study limitations (unicentric, small sample size), we cannot conclude that FC would be useful as predictive biomarker for immune-related colitis development. Also, we consider that increasing our sample size and extending this study would not clarify results in favor of FC. Furthermore, and thanks to the expertise that this study brought to us, we consider that FC is not a good parameter to use in daily clinical practice in this context because of its enormous variability, unpredictability, and also the effort that it entails to measure in oncological patients. We will continue looking for predictive biomarkers able to predict immune-related toxicity in order to try to design studies that may offer early treatment to avoid stopping potentially effective treatments.

Author contributions

Ana Cardeña was the beneficiary of the grant that covered this project and consequently, its main coordinator. She was the protocol designer, the informed consent writer, and an active participant during recruitment, data entry and final analysis. Also, she is the author of the complete manuscript. Xabier Mielgo contributed to the protocol design and was second supervisor of the study activity and during the writing of the manuscript. Manuel Ruiz helped to focus the initial idea and also, participated during designing of the protocol that was sent to apply for the grant and finally, contributor for designing of future studies where we could use FC as predictor. Ruth Martinez contributed during the data entry and designing of CRF. Elia Pérez helped with the statistical design of the protocol and final data interpretation. Juan Carlos Cámara is the head of our investigator team and supervised all the process and also, participated actively in the patient recruitment. The rest of the authors contributed equally during recruitment and management of our patients in consultations and oncology ward.

Data availability