Relation of Ultrasound Findings and Abdominal Symptoms obtained with the CFAbd-Score in Cystic Fibrosis Patients

Abdominal symptoms are a hallmark of Cystic fibrosis (CF). Yet, their association with morphological abnormalities of different abdominal organs is still poorly understood. Aim was therefore to relate these symptoms, assessed with a questionnaire, to findings in abdominal ultrasound (US). In 114 CF patients of all ages, findings in US considering seventeen specific parameters were related to abdominal symptoms compiled with our novel CF-specific 26-modal symptom score (CFAbd-Score). US abnormalities were detected in 95% of the patients. Most frequent findings were pancreatic lipomatosis (88%), liver steatosis (37%), hepatomegaly (31%), and thickened bowel walls (23%). Highest burden of GI-symptoms was clearly associated with pancreatic lipomatosis (p = 0.036). In detail, patients revealing this pathology reported higher rates of abdominal pain (p = 0.018), flatulence (p = 0.006), heartburn (p = 0.04), and reflux of stomach content (p = 0.006). Patients with pancreatic sufficiency had less US-findings (p = 0.033), which in turn was associated with lower rates of abdominal symptoms. The majority of them were carriers of class IV-VI or G551D mutations. Our approach gives new insights regarding the underestimated multi-organ abdominal involvement in CF. The new score can be of high interest e.g. as a complementary tool to assess the gastrointestinal effects of promising novel CF therapeutics.

Ethical Statement. The study was approved by the Jena University ethics committee (registration number 4458-06/15) and done in adherence to ethical principles. All methods were performed in accordance with the relevant guidelines and regulations. All patients and parents of minors provided written informed consent.
Evaluation of abdominal symptoms. The CFAbd-Score was further developed from our recently presented JenaAbdomen-CF Score 1.0 4 with additional 9 items concerning GI symptoms-related quality of life (QoL), thus now consisting of 26 items. QoL items were: embarrassed, physical activity limitation, reduced productivity, fatigue, reduced concentration, frustrated/restless/irritable, sad, difficulty falling asleep, and waking up at night, measured with a 6-point Likert scale from 'not at all' (0 pts) to 'always' (5 pts).
In contrast to the first version, each item was reversely scored and linearly transformed to a 100-points scale (0 = 100, 1 = 80, 2 = 60, 3 = 40, 4 = 20, 5 = 0) with lower rates for increasing severity of GI symptoms. Only the domains with more than 50% of the items answered were included in the calculation. CFAbd-Score (26 items) was calculated as the sum of the items over the number of items answered in each domain.
Measurement of ultrasound abnormalities. The US examinations were performed in a single center using an ultrasound scanner (Philips; U22; Philips Healthcare, Best, the Netherlands) under standardized conditions. The findings were confirmed by and discussed with a second experienced radiologist in order to enhance the quality of the data. The abdominal US findings included evaluation of 17 parameters: (1) Bowel wall thickness (BWT) was measured in a longitudinal and transverse section. The measurement was taken from the central hyperechoic line of the lumen (representing the interface between content of the lumen and the mucosa) to the outer hyperechoic margin of the wall (representing the serosa). BWT was considered 'normal' up to 4 mm 5 . (2) Enlarged mesenteric lymph nodes were defined as greater than 5 mm in the shorter axis and larger than 10 mm in the long axis 5 . (3) Intussusception was defined as the invagination of a proximal segment of the intestine into an adjacent distal segment. In US, it appears as a mass with multiple concentric rings or doughnut signs on the short axis 5 . (4) Appendiceal thickening was defined as a diameter of more than 6 mm and/or an appendiceal wall thicker than 2 mm 6 . (5) Free fluid within the peritoneal cavity. (6) Pancreatic cystosis was defined as anechoic structures which are usually round and oval (cystic lesions) with sizes greater than 1 cm (macroscopic cysts) 7 . (7) Pancreatic lipomatosis was given when pancreatic echogenicity was partially or completely higher than liver (or the kidney in case of liver hyperechogenity) 8 . (8) Cholecystolithiasis was demonstrated by typical acoustic shadow in the gallbladder. (9) Micro-gallbladder was defined as less than 2-3 cm long and 0.5-1.5 cm wide 9 . (10) Coarseness of the hepatic parenchyma, (11) nodularity of the liver edge, and (12) periportal fibrosis (increased periportal echoes) were documented according to a scoring system established by Williams et al. 10 . (13) Liver steatosis criteria were increased echogenicity compared to renal parenchyma, vascular blurring, and deep attenuation of the US signal 11 . (14) The maximum velocity of flow in the portal vein (PV) was measured in cm/s and was considered decreased below 15 cm/s 12 . (15) Transient elastography (TE) was measured in kPa and was considered increased with a value above 7.1 kPa 13 . (16) Hepatomegaly was indicated when the liver span at the mid-clavicular line exceeded the upper limits according to height 14 . (17) Splenomegaly was indicated when spleen length (measured as the larger diameter through the hilum in a cranio-caudal axis) exceeded the upper limits according to age 15 .
These 17 parameters were scored either as absent (0 pts) or present (1 pt) with a maximal score of 17 points (US-17). In addition to US-17, the Williams score 10 and TE were correlated separately to the CFAbd-Score.
In order to link the pancreatic status (pancreas sufficiency = PS/pancreas insufficiency = PI) to US abnormalities, doses of substituted enzymes quantified as intake of international units of pancreatic lipase per kg of body weight and day (IU/kg/d) were additionally factored in.
Measures of clinical data. The recently established pancreatic insufficiency prevalence (PIP) scores adapted from Ooi were used to measure the severity of specific CFTR mutations in regard to pancreatic function 16 . Patients carrying mutations which had not been included into the study from Ooi could not be attributed to a specific PIP score and thus were excluded from the PIP-genotype analysis (12/114 patients). CFTR-mutations were classified as I-III (severe) and IV-V (mild) 17 . An adequate visualization of the pancreas by US was present in 92% of the patients (105/114). TE could be performed in 99 of the 114 patients. CF-liver disease (CFLD) was defined according to Debray et al. 18 .
Data Analysis. Statistical analyses were performed using SPSS v.23.0. Median, lower (Q 1 ) and upper (Q 3 ) interquartiles of the score in relation to the respective US parameter are given as median [Q 1 ;Q 3 ]. To detect statistical differences between the US findings-associated symptom scores, Mann-Whitney-U test was chosen following Kolmogorov-Smirnov (K-S) testing of normal distribution. To evaluate the difference between two patient groups (e.g. PS vs. PI), the Hodges-Lehmann (HL) estimator (median of all pairwise differences) with 95% confidence interval (CI) was reported. Nominal data was compared with the Chi-square test or Fisher's exact tests, as appropriate. Correlations between variables were examined using the Pearson's correlation coefficient. A p-value ≤ 0.05 indicates a significant difference or correlation.

Relation of US-findings to symptoms assessed with the CFAbd-Score. Ultrasound abnormalities detected in the included CF patients are listed in
Furthermore, we did not find differences in US-17 between sexes. Altogether, only 5.3% of the included CF patients (6/114) did not reveal any pathological US-finding, which in part also could have been caused by superposition of the pancreas or other organs with bowel gases (minimal abdominal US). Moreover, enzyme dose was not related to US abnormalities (Supplementary Table S2).

Discussion
To our knowledge, this is the first study correlating the wide range of different findings in abdominal ultrasound in CF patients with the range of typical abdominal symptoms assessed with a second, revised version of our CF-specific score (CFAbd-Score). Thereby, we queried frequency, duration, and intensity of symptoms together with the related burden for the patient's daily life. By using US measures we found that particularly pancreatic lipomatosis is associated with an increased abdominal symptom load. Altogether, pancreatic lipomatosis was detected in 88% (92/105) of the included CF patients. Overall, this accords well to the pancreatic status, as the majority of the included CF patients with pancreatic lipomatosis were pancreas insufficient (PI) (91%), and these US changes are a common finding in this subgroup 13,19 . By means of serial radiology, four states have been revealed in the course of the progression from PS to PI: normal pancreas, an atrophic state subsequent to recurrent pancreatitis, a lipomatous state in which pancreatic tissue is successively replaced by fat-dense tissue, and a macro-or microcystic pancreas with parietal calcifications 20 . Finally, many patients who initially were classified as PS will become PI. The finding of lipomatosis in our PS patients indicates an advanced stage in the progression from PS to PI. On the basis of a first and preliminary version of our score (JenAbdomen-CF Score 1.0) we demonstrated that PI-CF patients report significantly more GI symptoms compared to PS-CF patients 4 . A recent study using a symptom questionnaire developed as PI-specific patient-reported outcome measure (PROM) revealed that 84% of the patients reported experiencing abdominal pain 21 . This is quite similar to our finding that 79% of our CF patients (92% PI) recurrently and regularly suffer from abdominal pain. The most frequent locations of abdominal pain were the umbilical (83%) and epigastric regions (11%) 4 . Our finding on the impact of a pancreatic lipomatosis fits very well with the patients´ statements. The elevated rate of GI symptoms in PI patients may be caused by the combination of exocrine pancreatic deficiency to liberate enzymes, the failure to neutralize the acidic gastric contents, and intestinal inflammation 22 . Nevertheless, this finding should be taken carefully, as a smaller number of patients with elevated pancreatic echogenicity quoted as pancreatic lipomatosis still maintain residual pancreatic function according to a PS phenotype and vice versa. At the same time, US revealing a hypoechoic pancreas, which does not accord to pancreatic lipomatosis, could represent a fibrotic pancreas, as found in patients with PI 19 . However, these cases accounted for a minority of our cohort. Interestingly, prominent US findings such as pancreatic cystosis, which was detected in 6/105 patients (6%), with cyst lesions measuring up to 40 mm did not significantly contribute to the burden of abdominal symptoms in these patients. However, this specific pathology might cause symptoms later, e.g. when further growth of cysts compress surrounding structures, as described in a single case report by deGruchy et al.: the 9 year old patient with pancreatic cystosis suffered from intense radiating abdominal pain 23 . Previously, Dietrich et al. reported that abdominal US reveals small pancreatic cysts with a mean diameter of 18 mm (7-27 mm) in up to 18% of CF patients (n = 12/67) 9 .
Thickened bowel walls, as detected in 26 of 114 patients (23%) of our CF cohort, were not significantly correlated with an increased burden of GI symptoms. Thereby, until now, etiology of bowel wall thickening in CF is not fully understood. A relation to intestinal inflammation, and possibly to dysbiosis has been discussed and also that in the long run it could lead to segmental submucosal fibrosis 24 . However, this assumption needs further evaluation in future studies. As previously reported, the dose of applied pancreatic enzymes did not correlate with the presence of thickened bowel walls 24 , even though such a correlation was discussed 25 .
The presence of an enlarged appendix was observed in 7 of the included CF patients (6%). This finding was not related to an increased burden of GI symptoms. Apparently, patients remain asymptomatic until occurrence of an appendiceal perforation or abscess formation 6 . Altogether, the appendix in CF has been reported to be usually filled with mucoid contents, which maintain the appendiceal lumen distended and less prone to total luminal occlusion and acute inflammation. Consequently, progress to acute appendicitis in CF appears to be rare, occurring in only 1-2% of patients 26 .
As suspected, hepatic abnormalities including coarse parenchyma, nodular edges, periportal fibrosis, organomegaly, increased TE, and decreased PV velocity was not correlated with an increased burden of GI symptoms. However, symptoms like upper GI hemorrhage and tarry stools may manifest with end-stage liver disease 27 . Such an acute hepatic decompensation, which is often associated with metabolic disorders, is rare and did not occur in our observational period.
In the present study, 42 patients (37%) revealed sonographic signs of liver steatosis, which is in accordance with previous ultrasound studies reporting steatosis in 23-75% of CF patients 28 . The most common abdominal symptom found in that subgroup was 'fatty stools´, which was significantly more frequent in this group (p = 0.031). Interestingly, 41 out of these 42 patients were PI. As recently published, patients who underwent pancreatectomy are likely to develop hepatosteatosis due to maldigestion. This largely depended on the endo-and exocrine function of the remnant pancreas and a well-adjusted enzyme supplementation 29 . In non-CF patients, hepatosteatosis is closely linked to an impaired glucose metabolism 30 . It is conceivable that also in CF the development of a liver steatosis is promoted by insulin deficiency due to pancreatic insufficiency. Once in progress, the situation is worsened by an inadequate synthesis of hepatic phosphatidylcholine and apolipoprotein B, which results in impaired VLDL assembly and, subsequently, disturbed removal of triacylgycerols from the liver 1,31,32 . Moreover, deficiency of essential fatty acids and carnitine are factors discussed in the context of hepatosteatosis in CF 33 .
Enlarged abdominal lymph nodes, consistently with a reactive morphology, were detected in 10 of our patients (9%). This finding was not associated with elevated GI symptomatology. Likewise, detection of gallstones (found in 3% of patients) was not related to elevated symptoms. However, it is not unlikely to give rise to symptoms in the course of time.
Interestingly, PS patients carrying a mild genotype (class IV-VI) as well as patients with a G551D mutation had significantly lower rates of US abnormalities (US-17) compared to patients with class I-II mutations and PI, respectively. As a limitation, 8 of the 15 patients carrying a G551D mutation were treated with ivacaftor (IVA) at the time of inclusion. With higher patient numbers, it would be most interesting to assess changes of abdominal symptoms due to introduction of the potent CFTR modulator. According to recent publications, IVA improved many of the abdominal pathologies such as a hyperacidic gastric content and liver steatosis, and even restored exocrine pancreatic function to some extent in some patients [34][35][36] . Even more impressive, improvement of GI symptoms was observed in CF patients treated with orkambi, the combination of the CFTR corrector lumacaftor and the potentiator IVA, which was recently approved for CF patients homozygous for F508del in many countries. Here is where a differential score assessing the multimodal abdominal involvement in CF is of outstanding interest to sufficiently document and quantify such interesting effects. This would be one of many interesting fields for the application of a score like the present CFAbd-Score, which could bring substantial new insights as a PROM. As previously mentioned, the validation of this PROM is still in progress and, just like the JenAbdomen-CF Score 1.0, the second version presented herein is preliminary 4 . Nevertheless, we show that especially pancreatic lipomatosis was well reflected by self-reported elevated GI-symptoms. Our data may contribute to the still little understanding of the abdominal involvement in CF.