Introduction

Cystic fibrosis (CF) is a genetic life-threating disease resulting from absence or dysfunction of the cystic fibrosis transmembrane regulator protein (CFTR), reducing the chloride transport and leading to production of thick dehydrated secretions.1 The disease particularly impacts the respiratory and digestive system. In the digestive system, obstruction of the pancreatic ducts leads to reduced pancreatic enzyme secretion into the duodenum, thus compromising nutrient digestion and absorption, most importantly fat.2

Pancreatic enzyme replacement therapy (PERT) is the indicated treatment of exocrine pancreatic insufficiency (PI), but there is clear evidence of variable efficacy, both between and within the same patient, in terms of fat absorption and gastrointestinal (GI) symptoms due to overdosing or underdosing.3–5

In recent years, a large European project was conducted to explore the mechanisms underpinning fat digestion in paediatric patients with CF, by addressing a combined in vitro–in vivo digestion approach. The main product of the study was a personalised medical device (mobile app) supported by a predictive algorithm of the optimal dose of PERT.6 The system was evaluated in a multicentre clinical trial, which indicated improved gastrointestinal-related quality of life (GI-QoL)7 and improved fat absorption after using the app in patients with the poorest baseline coefficient of fat absorption (CFA).8 However, CF is a multisystemic complex disease and other factors beyond adequate PERT may possibly affect fat absorption and GI symptoms.

Intestinal inflammation is potentially one of these factors, and faecal calprotectin (fCP), a commonly used biomarker of inflammation, has been reported to be increased in these patients; however, there is not enough evidence on the relationship between elevated levels of fCP and intestinal dysfunction.9 Moreover, some studies have reported that calprotectin is also produced by neutrophils in the lung in the context of pulmonary inflammation.10

The present study aims to assess if there is any relationship between fCP and a series of clinical variables including fat malabsorption, GI symptoms and pulmonary function over a follow-up period of 1 year.

Material and methodsSubjects and study design

This observational study was conducted in the frame of a large, multistage, European multicentre project (MyCyFAPP H2020-ref.643806) involving paediatric CF centres, to assess the evolution of fCP over time and its possible associations with clinical variables (forced expiratory volume in 1 s (FEV1), CFA and GI-QoL). In MyCyFAPP Project, a prediction model to estimate optimal PERT dose was created based on in vitro data and integrated into a mobile app.11 GI-QoL and CFA values were obtained as either primary or secondary outcomes in two studies involving faecal sample collection within the MyCyFAPP Project, so the power calculation was performed to assess those outcomes7 8 (figure 1).

Figure 1

Overview of the study design. *Considered as the primary outcome. 1Calvo-Lerma et al. 12 2Calvo-Lerma et al.11 CFA, coefficient of fat absorption; fCP, faecal calprotectin; PERT, pancreatic enzyme replacement therapy.

The first was a pilot study (carried out in 2016–2017) assessing the effect of the predictive model of enzyme dose for fat digestion. Patients were instructed to use a fixed diet, with a PERT dose instructed by the prediction model. This was performed twice (v1, v2). Faecal samples were collected for 24 hours and CFA considered as the primary outcome.12 The second study (carried out in 2018) was a 6-month prospective clinical trial assessing changes in QoL (measured by the modified PedsQL-GI) after the use of the app to adjust the dose of PERT as primary outcome.7 Faecal collection for analysis of CFA and fCP was performed at two visits, 1 month apart (v3 and v4). The first 24-hour faecal collection (v3) was conducted on a free diet and usual dose of PERT, the second 24-hour faecal collection (v4) was done when patients followed the same diet but using the PERT dose recommended by the prediction model in the app.8 In total, four 24-hour faecal samples per patient were obtained over a period of 18 months: that is, sample 2 was collected 5 months after sample 1, sample 3 was collected 1 year after sample 2 and sample 4 was collected 1 month after sample 3.

GI-related symptoms

The modified PedsQL-GI symptoms questionnaire was used.13 It is composed of 10 items that are rated by the patients from 0 (bad) to 100 (good), which include the overall score of GI-QoL and nine symptoms. The questionnaire was specifically adapted and validated for the purpose of v3 and v4, so in the previous stages (v1 and v2), this outcome could not be measured.

Faeces analysis

Faecal samples were collected between colorimetric markers and processed according to a previously published protocol.12 fCP analysis was performed using the Faecal Sample Preparation Kit (Roche Diagnostics) and quantified using the EliA Calprotectin 2 assay (Phadia AB).14 Unabsorbed fat was analysed by infrared spectroscopy15 and CFA was calculated as previously published.12

Statistical analysis

Data were summarised using mean (SD) and median (first and third quartiles) for numerical variables and absolute frequency (%) for continuous variables.

To evaluate time-related changes of the clinical parameters (fCP, CFA and FEV1%) along the study period, linear mixed models were performed including time period as monotonic effects. The model was extended with individuals as a random effect with a random intercept to correct for the non-independence of the data. The precision of the estimates was assessed by its 95% CIs. fCP was log-transformed to minimise the effect of skewed data. Additionally, associations between fCP and GI symptoms were carried out using linear mixed models (one per GI symptom score), including as covariables of the model, the score for the symptom (ie, diarrhoea, constipation and total PedsQL-GI score), age, time point and the interaction score:time point.

All the analyses were performed using R software (V.4.0.3). A p value below 0.05 was considered statistically significant.

ResultsDemographic and clinical data

Of all the participants in the two studies within MyCyFAPP Project (42 in the pilot study and 56 in the clinical trial substudy), 29 subjects (14 female) complied with the criterion of having contributed the 24-hour faecal samples at the four scheduled time points. The demographic and clinical characteristics of the patients are shown in table 1. At baseline (v1), the mean age was 8.3 years.

Table 1

Clinical and demographic data in the study cohort along the four study visits

Time-related changes of the clinical parameters along the study period

Levels of fCP, CFA and pulmonary function (FEV1) were monitored along the four study visits. With respect to the fCP levels, an overall significant increase was detected (p<0.001, estimate 0.373, 95% CI (0.205, 0.542)), starting with median levels of 62 µg/g, which raised to 256 µg/g at the end of the follow-up (figure 2A). Concerning fat absorption, the CFA was similar in the two first time points and below the 90% clinical target, but in the last two control points (v3 and v4), this parameter significantly increased to 94% and 95%, respectively (p<0.001, estimate 4.03, 95% CI (2.44, 5.65)) (figure 2B). Finally, the median values of FEV1 decreased significantly along the four study visits, being the initial values in v1 and v2 approximately 97%, 90% in v3 and 87% at the end of the study (figure 2C).

Figure 2

Time-related changes in levels of faecal calprotectin (fCP, µg/g) (A), the coefficient of fat absorption (CFA, %) (B) and pulmonary function (FEV1, %) (C) along the four time points conforming the studies. FEV1, forced expiratory volume in 1 s.

An inverse, but not significant, relationship between CFA and fCP (p=0.073, estimate −1.427, 95% CI (−2.756, 0.101)) was documented; in addition, no significant associations were found between fCP and both nutritional parameters and type of mutations. PERT dose was neither associated with the change in the fCP. However, the decrease in pulmonary function was significantly associated with the increase in the fCP levels (p<0.001, estimate −4.012, 95% CI (−5.88 to –2.016)).

Association between fCP and GI symptoms

The association between fCP levels and the GI symptoms included in the PedsQL-GI score was evaluated in the last two study time points, as the PedsQL-GI Questionnaire was only validated in the second half of the project (v3 and v4). Throughout these two study time points, the associations with fCP were statistically significant only for diarrhoea (estimate −0.015, 95% CI (−0.03; −0.001), p=0.04) and for the total score of the modified PedsQL-GI Questionnaire (estimate −0.033, 95% CI (−0.062; −0.004), p=0.03) (figure 3). In addition, fCP was not associated with other study variables, such as age, gender, centre or nutritional status parameters.

Figure 3

Association between fCP levels and gastrointestinal-related quality of life included in the PedsQL-GI Questionnaire: diarrhoea score (A) and total score (B). fCP, faecal calprotectin.

Discussion

This study, carried out in the pre-CFTR modulator era, involved 29 patients with CF in the paediatric age. Time-related changes of three relevant clinical indicators during 1 year of follow-up were documented: intestinal inflammation (fCP), which increased over time, fat absorption (CFA%), which also increased, and pulmonary function (FEV1%), showing decreasing values. The reduction in pulmonary function was statistically associated with increased fCP levels, but the increase in CFA was not related with the other parameters. In addition, the fCP levels were associated with the GI-QoL.

The fCP levels detected in our cohort were at all time points above the upper limit for normal adults (50 µg/g), higher than in previous series of healthy children16 and comparable with values previously reported in patients with CF. In the study by Garg et al, 17 children with CF and PI aged 4–10 years old had median fCP values around 56 µg/g faeces.17 Similarly, Dhaliwal et al 18 reported mean values of 94.3 µg/g in 28 children with CF, most of whom had PI; in their study, there was no correlation between fCP and FEV1, PERT dosage or GI symptoms, but an association was found between fCP and poor growth.18 In contrast, another study on 19 children with CF (median age 4 years old) reported median fCP concentrations below the limit of normal in 19; 9 of them were on antibiotic therapy.19 However, others reported fCP values more similar to ours with a median value of 184 µg/g in 19 children with CF (age 2–9 years old).20 Other authors found in 59 patients with CF (median age 8 years old), 43 with PI and 16 with pancreatic sufficiency (PS), a median fCP level of 94 µg/g, but no significant associations were found between elevated fCP and gender, growth parameters, genotype, GI symptoms, or being on PERT and/or probiotics.10

Notably, most of the commented studies were cross-sectional, whereas our series was prospective showing increasing fCP levels over time. This finding may be related to lung disease, as indicated by the significant and inverse association of lung function with fCP. It is known that calprotectin is also produced in the context of lung inflammation, a condition characterised by decline in lung function and increased sputum production. If swallowed, sputum calprotectin reaches the intestine and is then excreted in faeces.21 Therefore, the fCP values might relate, at least in part, to calprotectin of pulmonary origin. We must however consider that sputum-derived calprotectin may be denaturated to some extent during gastric digestion because of the acidic conditions.10 A similar inverse association between FEV1 and fCP levels was previously described by Adriaanse et al, 21 in adult patients with CF and not in children.21 Moreover, Rumman et al 10 found that patients with lower FEV1 were more likely to have higher fCP levels, but this association was not statistically significant.10

In this respect, another long-term longitudinal retrospective study including 171 patients (0–61 years of age) reported median fCP concentrations of 61 µg/g, and a surprising decline of median fCP values in the whole patient cohort along 12 years of follow-up.22 The authors related this finding to therapeutic interventions with Lactobacillus GG in the patients with higher levels of fCP. In that study, there was no clinically relevant association between fCP and sex, age, FEV1, body mass index or GI symptoms, but patients with PI had higher fCP levels compared with children with PS.

In addition, in our study, a significant association was found between fCP levels and GI symptoms, particularly, total score for GI-QoL and diarrhoea. This suggests that GI symptoms are not only related to fat malabsorption, but also to intestinal inflammation, in contrast with other studies in children with CF that did not find this association.10 18 Recently, it was reported that the fCP levels were associated with GI symptom scores, and that high fCP levels were associated with a worse QoL score.23

With regard to fat absorption data, the increase of the CFA detected in v3 is probably attributable to the training of the patients achieved during the previous stages of MyCyFAPP Project in which they were informed about the importance of adjusting the dose of enzymes to the type of foods.12 However, we could not detect any association between increasing levels of CFA over time with the levels of fCP.

Overall, the results of different studies are heterogeneous and controversial when addressing associations between fCP and clinical parameters such as age, nutritional status or pulmonary function, but it should be considered that study designs are also heterogeneous in terms of type of patients (PI/PS), age range of the cohorts (some studies gather both adults and children), methodology and analytical tools. For these reasons, the large variability of fCP in CF precludes this parameter to be used as a valid marker to monitor inflammation and clinical interventions.

The main finding of the present study is the negative association between fCP and lung function over time and the association between fCP and diarrhoea. We are aware that our study has several limitations: the cohort belonged to a clinical trial that did not include patients with PS, some had pulmonary exacerbation during the study period (despite it was an exclusion criterion only at baseline) and were treated with antibiotics (which could explain the overall decrease in lung function through the follow-up period), and there was no control group. Of note, the study was performed before the regular implementation of the CFTR modulator treatment. In addition, our study originally aimed to assess interventions to improve fat absorption by means of accurate PERT dose, so the impact on both fCP and lung function was not specifically assessed. Furthermore, we did not collect simultaneous blood and airway samples to analyse calprotectin or more direct markers of inflammation in the lungs, to investigate whether the calprotectin measured in faeces could be of pulmonary origin from ingested sputum. On the other hand, we cannot exclude that the significant association between fCP and FEV1 might be causal.

In future studies, the origin of fCP should be assessed also in blood and sputum samples. In parallel, and in view of the high variability in this parameter, other biomarkers of intestinal inflammation in CF should be explored, given the relevance of achieving fat absorption in these patients.

In conclusion, this study assessed for the first time fCP prospectively along several control points over a 1-year period, providing new evidence that in patients with CF, fCP is inversely associated with pulmonary function in paediatric patients. The origin of fCP possibly linked to pulmonary inflammation should be further evaluated by new specific biomarkers. On the other hand, fCP levels were also associated with GI symptoms. Therefore, it remains to be established whether fCP could be used to test intestinal inflammation in CF, due to the confounding effect of pulmonary inflammation.