Long-term Prognosis of Nonceliac Enteropathies and a Score to Identify Patients With Poor Outcomes: A 30-year Multicenter Longitudinal Study

Abstract

INTRODUCTION:

Long-term prognosis of nonceliac enteropathies (NCEs) is poorly understood. We aimed to evaluate long-term outcomes and develop a prognostic score for NCEs.

METHODS:

NCEs patients from an international multicenter cohort (4 Italian centers, 1 United Kingdom, 1 French, 1 Norwegian, 1 United States, 1 Indian) followed-up over 30 years were enrolled. Complications and mortality were analyzed with Kaplan-Meier curves, standardized mortality ratios (SMR), and multivariate Cox regression. A clinical score to identify patients at risk of poor outcomes was developed.

RESULTS:

Two hundred sixty-one patients were enrolled (144 female, mean age at diagnosis 49 ± 18 years, median follow-up 70 months, interquartile range 24–109). The most common etiologies were idiopathic villous atrophy (39%), drug related (17%), common variable immune deficiency (15%), infectious (10%), and autoimmune enteropathy (9%). Five-year and 10-year complication-free survival were 89% and 77%, respectively, whereas 5-year and 10-year overall survival were 88% and 74%, respectively. Causes of death included sepsis/major infections (22%), lymphoproliferative disorders (22%), solid-organ malignancies (12%), cardiovascular/metabolic disease (10%), and was unknown in 33%. Mortality was increased in NCEs compared with the general population (SMR 3.17, 95% confidence interval [CI] 2.24–4.34). Older age at diagnosis (P < 0.001), anemia (hazard ratio [HR] 2.53, 95% CI 1.33–4.80, P < 0.01), and lack of clinical (HR 3.21, 95% CI 1.68–6.18, P < 0.01) and histological response (HR 2.14, 95% CI 1.08–4.23, P = 0.04) were independent predictors of mortality at Cox regression. A 5-point score was developed to identify high-risk patients: very low risk (0 pts), low risk (1–2 pts), intermediate risk (3 pts), and high risk (4–5 pts), with 10-year survival rates of 100%, 87%, 62%, and 16%, respectively.

DISCUSSION:

Mortality in NCEs is increased because of complications and lack of response to current therapies. We developed a clinical score to personalize follow-up. Targeted treatments are needed to improve outcomes.

INTRODUCTION

Villous atrophy (VA) of the small bowel is the cardinal histopathological lesion of chronic enteropathies including celiac disease (CD) and non-celiac enteropathies (NCEs) (1–6).

CD is a chronic gluten-dependent enteropathy characterized by both a high prevalence in the general population and an increased mortality due the development of rare but severe complications, such as refractory CD, abdominal lymphomas, and small-bowel carcinoma (7–9). Although positive immunoglobulin A (IgA) tissue transglutaminase and endomysial antibodies are markers of conventional CD (5,6), forms of chronic enteropathies characterized by severe malabsorption, VA, and negative celiac serology exist and they are defined as seronegative enteropathies (SNEs) (1–4,10–17).

The differential diagnosis of SNEs is challenging and should consider 2 major clinical scenarios, which include seronegative celiac disease (SNCD) and NCEs, which are clinically difficult to distinguish from one another (1–4,10–17).

NCEs encompass a group of rare SNEs characterized by extreme etiological heterogeneity (1–4,10–17). The most common underlying etiologies for NCEs include iatrogenic forms such as angiotensin II receptor blockers (ARBs) enteropathy (18,19), autoimmune enteropathy (AIE) (20–22), enteropathy associated with common variable immune deficiency (CVID) (23–26), infectious forms such as giardiasis, tropical sprue, HIV enteropathy, and tuberculosis (27–29), Crohn's disease (30), primary small bowel lymphomas (31–34), and idiopathic villous atrophy (IVA) (14,35).

In the last years, great attention has been dedicated to clarify the nomenclature and the differential diagnosis of SNEs (1–4,10), but there are only a few studies that have specifically evaluated the natural history and prognosis of these conditions (14–16). More precisely, it has been reported that patients with SNEs are characterized by overall worse prognosis when compared with CD (14–16). However, these studies considered the group of patients with SNEs as a whole and they did not specifically investigate long-term prognosis according to the underlying etiology. This point may be relevant because some very recent reports and our clinical experience seem to suggest a more favorable prognosis in some types of NCEs in which the underlying cause can be easily identified and treated, such as ARBs and some infections (10,19,36). Moreover, clinical management and modalities for follow-up of patients affected by NCEs are not standardized, and clinical tools guiding decision-making are currently lacking.

Therefore, the aim of this study was 3-fold: (i) to investigate the long-term prognosis of a large multicenter cohort of patients affected by NCEs who have been evaluated over 30 years, (ii) to identify clinical predictors of poor outcomes in NCEs, and (iii) to develop a score to stratify NCEs patients based on their risk of poor long-term outcomes.

PATIENTS AND METHODS

Study design and setting

This is an international multicenter longitudinal study aiming to evaluate the long-term prognosis of patients affected by NCEs evaluated over more than 30 years.

The study was proposed during and immediately after the 19th International Society for the Study of Celiac Disease Conference, Sorrento, October 2022, to 9 international referral centers for the study of CD and NCEs. Participating centers included the following: Pavia, Italy; Sheffield, the United Kingdom; Cleveland, the United States; Paris, France; Ferrara, Italy; Bologna, Italy; Padua, Italy; New Delhi, India; and Oslo, Norway.

Study population

Adult patients (age ≥ 18 year old) with a diagnosis of NCEs who had been followed up at each participating center in the period between January 1992 and January 2024 were recruited into the study. Patients with SNCD and CD associated with IgA deficiency were excluded (1,2,11).

Diagnostic workup for seronegative enteropathies and criteria for the differential diagnosis of NCEs

All patients with NCEs enrolled in this study had VA on duodenal biopsies (≥ Marsh 3a/Corazza-Villanacci grade B) taken from the second duodenal portion and were tested for celiac serology while on a gluten-containing diet. Serological tests for CD included IgA endomysial antibodies, IgA tissue transglutaminase antibodies, IgA/immunoglobulin B anti-deamidated gliadin antibodies (DGP), and IgA anti-gliadin antibodies (AGA). However, due to the long timespan of the study across many centers, modalities for serological testing were not standardized across centers.

Diagnostic criteria for identifying the underlying etiology in patients with SNEs adopted throughout 3 decades by different centers all over the world can be difficult to summarize. The first international consensus on the nomenclature and clinical criteria for the differential diagnosis of SNEs was published in 2022 (1), but in this study most enrolled patients received a diagnosis of NCE before that time. In each center, patients with seronegative VA were investigated in accordance with systematic protocols, which aimed to discriminate between either SNCD or other underlying etiologies unrelated to gluten ingestion. More precisely, the following diagnostic criteria were adopted.

1. Diagnosis of AIE was based on persistent severe VA and malabsorption unresponsive to a gluten-free diet (GFD) and after the exclusion of other causes of VA. Positive enterocyte antibodies, a history of autoimmune disorders, and/or positivity for non-organ-specific antibodies and/or a clinical/histological response to immunosuppressive therapies were considered supportive of the diagnosis of AIE. Duodenal histological findings consistent with deep crypt lymphocytosis and/or plasma cells infiltration, neutrophilic cryptitis ± crypt microabscesses, and lack/decrease of Paneth cells were further supportive diagnostic criteria (1,20–22,37).

2. Diagnosis of CVID was made in accordance with international guidelines issued by the European and American societies for immunodeficiency (38). In patients with VA and CVID, Giardia lamblia and other gastrointestinal infections were searched systematically and other causes of VA were excluded. Histological findings such as duodenal intraepithelial lymphocytosis, increased inflammation of the lamina propria, crypt apoptotic bodies, graft versus host disease-like lesions, and mucosal depletion of plasma cells were considered as further supportive criteria (1,23,24,39).

3. Iatrogenic enteropathies were diagnosed based on suggestive pharmacological history after exclusion of other causes of VA. Culprit medications included ARBs (18,19), immune checkpoint inhibitors (40), methotrexate (41), mycophenolate (42), azathioprine (43), colchicine (44) and mefenamic acid (45). Drug history was carefully reassessed particularly in those patients diagnosed before 2012, to exclude a contributing role of olmesartan or other ARBs (18). In addition, the medical history guided the diagnosis of chronic enteropathy due to chemotherapy or radiotherapy.

4. Infectious etiologies such as Giardia lamblia and other parasitic infections, HIV, and tuberculosis were identified based on positive specific testing. HIV testing, Quantiferon, or stool parasite cultures were routinely performed. Giardiasis was diagnosed on the basis of positive Giardia-specific stool antigens or the direct identification of the trophozoites on formalin-fixed paraffin-embedded hematoxylin and eosin-stained duodenal specimens and/or on the duodenal aspirate, or direct identification of cysts/trophozoites in fresh feces, or positive specific Giardia polymerase chain reaction (1–4,10). Tropical sprue was diagnosed when there was a persistent response in symptoms and reversal of VA with the use of tetracycline given for 12–24 weeks.

5. Diagnosis of lymphomas primarily involving the small bowel was based on a combination of histopathological criteria and typical findings at endoscopy and abdominal imaging. A diffuse infiltration of the epithelium and/or expansion of the lamina propria by small/medium CD3⁺CD4⁺ T-cells evaluated by means of immunohistochemistry or flow cytometry and presence of monoclonal rearrangement for beta-TCR and/or gamma-TCR allowed a diagnosis of indolent CD4⁺ T-cell lymphoma. Diagnosis of immunoproliferative small intestinal disease (IPSID) was based on a high index of suspicion and dense lymphoplasmacytic infiltrate of CD20⁺ B cells and plasma cells expressing the alpha heavy chain but no light chains (1–4,10,31–34).

6. IVA was diagnosed in patients in whom all the possible causes of VA were systematically investigated and excluded (1,14,35).

7. Helicobacter pylori, small-intestinal bacterial overgrowth, and NSAIDs were not included among the causes of NCEs, as recently proposed by an international consensus on the nomenclature and diagnosis of SNEs (1).

8. Finally, patients with seronegative VA who responded histologically to a GFD and in whom other causes of VA were excluded were considered to be affected by SNCD (1–5,11) and were therefore excluded from this study. Patients with CD-associated selective IgA deficiency were also excluded.

Data collection

For each patient enrolled in the study, the following clinical and demographic data were collected from clinical notes or local hospital databases: gender, age at diagnosis of NCEs and type of diagnosis, symptoms and laboratory alterations at baseline, results of upper gastro-intestinal (GI) endoscopy at baseline and during follow-up (when available), results of colonoscopy and video-capsule endoscopy (when available), type of therapy prescribed, clinical and histological response to therapies, necessity for hospital admission, age at onset of complication and type of complication, data of last clinical follow-up/death, and cause of death.

Follow-up and mortality study

For the purpose of this part of the study, data on follow-up and outcomes including clinical and histological response to therapies, development of major life-threatening complications, and mortality (including also causes of death) were retrieved from local hospital multidata sets or from patients' clinical notes.

Clinical response to therapy was defined as improvement or disappearance of symptoms and biochemical abnormalities present at baseline. Histological response to therapy was defined as resolution of VA (Marsh 0–2 or Corazza-Villanacci A) compared with baseline.

Development of lymphoma and other lymphoproliferative disorders, major infections including sepsis, and other life-threatening complications of the gastrointestinal tract, including malignancies, perforations, obstruction, and bleeding were considered major complications. Date of death and causes of death were obtained from local hospital registry and patients' clinical notes.

General population mortality rate data stratified by sex, age group, and calendar year for participating countries were sourced from the World Health Organization (WHO) Mortality Database, which compiles data from individual member states. Mortality data for the United States could be retrieved for the years 1992–2021, whereas for Italy, France, and the United Kingdom, data were available for the years 1992–2020, and for Norway, data were available between 1992 and 2016. No stratified general population mortality data could be retrieved for India.

Ethics

The study was conducted in accordance with the Declaration of Helsinki (6th revision, 2008). Patients underwent clinical tests and assessments as part of their routine care. The protocol of the multicenter study was approved by the Ethics Committee of IRCCS Pavia, Istituti Clinici Scientifici Maugeri, Pavia, Italy (protocol number 2654CE approved on September 14, 2022) and extended to all the participating centers. All results of the study are reported in this article, and additional data can be shared on reasonable request to the corresponding author. The STROBE guidelines were followed for quality assurance. Patients or members of the public were not involved into the design of this study.

Statistical analysis

Statistical analysis was performed using R version 4.3.1 (R Core Team [2022]. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/). Categorical variables were summarized as total counts and percentages and compared among groups using the Fisher exact test. Continuous variables were summarized as mean and SD or as median and interquartile range (IQR), whereas skewed variables were compared among groups using one-way ANOVA, Kruskal-Wallis test, unequal variances t test, or Mann-Whitney U-test, as appropriate. Post hoc multiple comparisons were adjusted for multiplicity using the Benjamini-Hochberg procedure.

Overall, incidence rates for development of complications and mortality were calculated per 100,000 person-years of follow-up with 95% confidence intervals (95% CI). Complication-free survival and overall survival was compared among groups with Kaplan-Meier curves and the log-rank test. Multivariate Cox regression was used to identify predictors of mortality, and the proportional hazards assumption was verified with the cox.zph R function and by visually inspecting Schoenfeld residuals. Variables for Cox regression analysis were selected among those showing an association with mortality and those deemed clinically relevant. To address missing values in the data set for Cox regression, we performed multiple imputations using the R mice package for variables with a small number of missing values. The imputation process was repeated 10 times, and Cox regression models were fitted to each of the 10 imputed data sets and then combined using Rubin rules. Hazard ratios (HR) and 95% CI were calculated, and model discriminatory ability was assessed by calculating the pooled C-index with 95% CI.

To assess the robustness of this approach, a sensitivity analysis was conducted repeating Cox regression on only complete records, with records missing values excluded and then compared with the results obtained after multiple imputation and no significant deviations were observed. Based on the predictors for mortality identified by Cox regression, we developed a 5-point score to stratify patients according to mortality risk. The area under the receiver operating characteristic curve was calculated to evaluate the score's discriminatory ability for predicting survival at 5 years and 10 years. Imputation of data was performed only for Cox regression, and for all other analyses, records with missing data were excluded.

Standardized mortality ratios (SMR) were calculated to compare mortality in the study cohort against general population data from the WHO mortality database. Patient follow-up time was split into 1-year intervals to account for changing age and calendar year during follow-up and matched by 5-year age-band, sex, country, and calendar year to the corresponding WHO mortality rate data. Expected deaths were calculated by multiplying person-years of follow-up in each band by the matched general population mortality rate. The overall SMR was calculated as the ratio of observed-to-expected deaths, with 95% CI computed using the exact Poisson method. Additional stratified analyses were performed to calculate SMRs for different age groups (< 40 years, 40–59 years, ≥ 60 years), sex, and type of enteropathy. No data imputation was performed for SMR calculations; records without available matching WHO mortality rate data were excluded from the analysis or censored at the last calendar year for which corresponding WHO mortality rate data were available. Two-tailed P values < 0.05 were considered statistically significant for all analyses.

RESULTS

Baseline clinical and demographic features and underlying etiologies

Overall, 261 patients with NCEs (144 female, mean age at diagnosis 49 ± 18 years; median follow-up 70 months, IQR 24–109) were included in the study. The 5 most common underlying etiologies for NCEs were idiopathic (39%), drug-related (17%), common variable immune deficiency (15%), infectious (10%), and AIE (9%). Other less common etiologies included Crohn's disease (5%), primary intestinal T-cell lymphomas (2%, including monomorphic epitheliotropic intestinal T-cell lymphoma and CD4⁺ T-cell lymphoma), IPSID (2%), eosinophilic enteritis (1%), and radiation enteritis (1%). Drug-related enteropathies were due to ARBs in 30 patients (68%), immune checkpoint inhibitors in 7 patients (16%), mycophenolate in 3 patients (7%), and finally due to methotrexate, tacrolimus, 6-mercaptopurine, and colchicine in 1 patient each. Infectious enteropathies included 9 cases of giardiasis (36%), 5 cases of HIV enteropathy (20%), 5 cases (20%) of intestinal tuberculosis (with Giardia coinfection in one of them), 4 cases of tropical sprue (16%), and 1 case each of hookworms and small bowel microsporidiosis. Table 1 presents the clinical and demographic features at baseline of enrolled patients and according to the underlying etiology.

Table 1.

Baseline clinical and demographic features of patients according to the type of underlying enteropathy

Overall, diarrhea (78.6%) and weight loss (69.6%) suggestive of severe malabsorption were the most common presenting symptoms, followed by anemia (45.4%) and low albumin (33.6%). Associated autoimmune disorders were found in 31.7% of NCEs patients and osteoporosis in 26.8%. Associated autoimmune disorders were significantly more common (P < 0.001) in patients with AIE (50%), CVID (50%), and in the group of other enteropathies (65.2%) than in patients with drug-related (15.6%), infectious (14.3%), and idiopathic (20.7%) enteropathies. Osteoporosis was significantly more common (P < 0.001) among patients with AIE (52.4%) and CVID (50%) than in the remaining NCEs subtypes.

Hospital admission was required in 95 NCEs patients (59%) overall, with significant difference across subtypes (P < 0.001), being patients with AIE (91.3%), CVID (73.7%), and drug-related enteropathies (65.6%) who more commonly required hospital admission.

Colonoscopy was performed in 148 NCEs patients (67.9%) overall, and mucosal changes at colonoscopy were found in 41 (27.7%), with significant difference in the prevalence of colonic involvement among NCEs subtypes (CVID 46.4%, AIE 38.9%, drug-related enteropathies 26.9%, other enteropathies 38.9%, idiopathic and infectious <15%, P = 0.01). Major findings during colonoscopy included microscopic colitis in 15 (10%), inflammatory bowel disease-like colitis in 8 (5%), non-specific colon inflammation with erosions/ulcers in 6 patients (4%), ileitis due to Crohn's disease in 5 patients (3%), indeterminate colitis in 3 patients (2%), ulcerative colitis in 2 patients (1%), graft-versus-host disease-like lesions in 1 patient, and eosinophilic colitis in 1 patient. Finally, in one patient a B-cell lymphoma was diagnosed at colonic biopsies.

Outcomes during follow-up

Table 2 summarizes the clinical outcomes during follow-up of the entire cohort of NCEs patients according to the underlying type of enteropathy.

Table 2.

Outcomes during follow-up in the entire cohort and according to the type of enteropathy

Clinical and histological response to treatment.

The most commonly prescribed treatments were budesonide (18%), systemic steroids (16%), GFD (15%), drug discontinuation in drug-related enteropathies (14%), antibiotics (8%), biological drugs (7%), azathioprine/6-mercaptopurine (6%), and intravenous immunoglobulins (6%). In 53 patients (21%) affected by an idiopathic form of VA, a watch-and-wait approach was adopted and no specific therapy was prescribed. More details on treatments in our cohort of patients and according to underlying etiology are provided in Supplementary Digital Content (see Supplementary Table 1, http://links.lww.com/AJG/D550). We specify that although 15% of patients were started on a GFD for suspicion of seronegative CD, none of the patients included in this study were diagnosed with seronegative CD, as there was no histological response to a GFD, which is the main criterion for diagnosing seronegative CD, along with the exclusion of other causes of VA.

Overall, clinical response occurred in 175 patients (78.1%), with significant difference according to the type of enteropathy (P < 0.01), being rates of clinical response higher in patients with a drug-related enteropathy (97.2%) and in patients with infectious enteropathies (85.7%) compared with patients with CVID (61.5%), AIE (75%), idiopathic (79.1%), and other etiologies (70%).

A follow-up esophagogastroduodenoscopy with duodenal biopsy was performed in 220 patients (84.3%) after a median of 16 months (IQR 7–54) from diagnosis and histological response occurred in 145 patients (65.6%) overall. Significant differences (P < 0.001) were found in histological response to treatment according to the type of enteropathy because most patients with drug-related enteropathies (93.3%) and infectious enteropathies (93.8%) had complete histological recovery after treatment, whereas histological recovery was less common in patients with CVID (25.6%), AIE (52.2%), other etiologies (63.2%), and idiopathic VA (72.3%).

Incidence of complications, types of complications, and predictors

Figure 1 shows complication-free survival during follow-up overall (Figure 1a) and subdivided according to clinical/histological response (Figure 1b) and underlying etiology (Figure 1c). Overall, 42 of 261 patients (16.1%) developed complications after a median of 38 months (IQR 5–81) from diagnosis of NCEs, corresponding to an incidence of complications of 25.5 per 1,000 person-years (95% CI 18.4–34.5). 5-year and 10-year complication-free survival was 89% and 77%, respectively.

Figure 1.

Complication-free overall survival (a) and stratified according to clinical/histological response (b), and according to underlying type of enteropathy (c).

Complications in these 42 patients, some of whom developed multiple types of complications, included lymphoproliferative disorders in 14 patients (12 lymphomas and 2 hemophagocytic syndromes), major infections in 27 (15 sepsis, 5 pneumonia, 2 systemic candidiasis, 1 leishmaniosis, 1 lymphocytic encephalitis, 1 recurrent cytomegalovirus infection, and finally 2 developed multiple types of infections), and other major abdominal complications in 9 (2 gastric adenocarcinomas, 2 esophageal adenocarcinomas, 2 small bowel obstructions with a contained bowel perforation in one of them, 1 small bowel ischemia, 1 small bowel pneumatosis, and 1 biliary pancreatitis).

As shown in Figure 1, development of complications during follow-up differed significantly (P < 0.001) according to underlying etiology, with CVID being at the highest risk of developing major complications, mainly of an infectious nature (14 patients), but also lymphomas (4 patients) and solid organ malignancies (2 patients developed gastric adenocarcinomas). On the contrary, none of the patients with an infectious cause of enteropathy developed complications during follow-up.

Mortality, causes of death, and predictors

Figure 2 shows overall survival during follow-up (Figure 2a) and subdivided according to clinical/histological response (Figure 2b) and underlying etiology (Figure 2c). Overall, 49 of 261 patients (18.8%) died after a median of 56 months (IQR 22–95) since diagnosis of NCEs, with an overall mortality rate of 28.0 per 1,000 person-years (95% CI 20.7–37.0). The 5-year and 10-year overall survival rates were 88% and 74%, respectively. Compared with general population mortality data retrieved from the WHO mortality database, the age-band, sex, country, and calendar-year matched mortality of patients with NCEs was markedly increased (SMR 3.17, 95% CI 2.24–4.34). Stratified results according to age group, sex, and type of enteropathy are reported in Supplementary Digital Content (see Supplementary Table 2, http://links.lww.com/AJG/D551). Notably, compared with the general population, mortality was increased for all types of NCEs except drug-related enteropathies (SMR 0.73, 95% CI 0.15–2.13).

Figure 2.

Overall survival in the entire cohort (a) and stratified according to clinical/histological response (b), and according to underlying type of enteropathy (c).

Causes of death in these patients included lymphoproliferative diseases (11 patients, 22.4%), infections (11 patients, 22.4%), solid organ malignancies (6 patients, 12.2%), cardiovascular disease (4 patients, 8.2%), and vascular complications of diabetes (1 patient, 2.0%), whereas cause of death could not be retrieved in 16 patients (32.7%). Cause of death did not differ significantly according to underlying type of enteropathy (P = 0.6).

There was no statistically significant difference in overall survival according to the underlying type of enteropathy at survival analysis (Figure 2c, P = 0.57), although the overall percentage of patients who died was higher in AIE (33.3%), CVID (27.5%), and in the group of other enteropathies (22.2%) and lower in patients with drug-related enteropathies (14.9%) and infectious enteropathy (8%). A post hoc sensitivity analysis (see Supplementary Figure 1, http://links.lww.com/AJG/D549) revealed that among drug-related enteropathies, mortality differed very significantly (log-rank P < 0.001) according to the type of drug involved, with ARB-related enteropathy having very low mortality, but significant mortality being present in enteropathy because of immunosuppressants, and especially in enteropathy because of immune checkpoint inhibitors.

Factors associated with increased mortality included older age at diagnosis (P < 0.001), anemia at diagnosis (P < 0.01), low albumin (P < 0.001), hospitalization (P < 0.01), increased erythrocyte sedimentation rate (P = 0.02), C-reactive protein (P = 0.04), and treatment with systemic steroids (P < 0.01). On the contrary, as shown in Figure 2c, both clinical and histological response were associated with significantly lower mortality (both P < 0.001).

Multivariate Cox regression analysis (Table 3) confirmed that older age at diagnosis (P < 0.001), anemia at time of diagnosis (P < 0.01), lack of clinical response (P < 0.001), and lack of histological response (P = 0.04) were independent predictors of mortality. Model discriminatory ability was good with a C-index of 0.80 (95% CI 0.75–0.86).

Table 3.

Predictors of mortality at Cox regression analysis

Clinical score for risk stratification

Based on the results on the predictors of mortality identified at Cox regression, we developed a simple clinical score (shown in Figure 3) for stratifying patients into 4 risk groups: very low risk (score 0; 5-year survival 100%, 10-year survival 100%), low risk (score 1–2; 5-year survival 96%, 10-year survival 87%), intermediate risk (score 3; 5-year survival 87%, 10-year survival 62%), and high risk (score 4–5; 5-year survival 51%, 10-year survival 16%). The area under the receiver operating characteristic curve for predicting 5-year survival was 0.84 and for predicting 10-year survival was 0.83.

Figure 3.

Clinical score for risk stratification of patients with nonceliac enteropathies.

DISCUSSION

This is an international multicenter clinical study evaluating long-term prognosis of patients with NCEs in a large cohort spanning 3 decades. We have compared the clinical characteristics of different types of NCEs according to underlying etiology, and we have delineated the main risk factors for poor outcomes and developed a clinical score to identify patients at higher risk of poor outcomes.

The main findings of our study based on real-world data show that mortality of NCEs was increased more than 3-fold compared with the general population. 5-year and 10-year overall survival rates were 88% and 74%, respectively, with a high burden of major complications which contributed significantly to morbidity and mortality. Complications included development of abdominal lymphomas and other lymphoproliferative disorders, major infections including sepsis and other major gastrointestinal events such as solid organ malignancies, small bowel obstruction, perforation, and ischemia. These findings corroborate the previous descriptions mainly as case reports/small case series of lymphoproliferative disorders, malignancy, and major infections particularly in patients with AIE and CVID enteropathy (26,31,46–48).

We have identified major clinical risk factors of poor outcomes that can be applied in clinical practice, regardless of the underlying type of NCEs. These include older age at diagnosis, with those diagnosed after the age of 60 years at highest risk, anemia at baseline, and lack of clinical and histological response to therapy. Based on these factors, we developed a simple clinical score stratifying NCEs patients according to short-term and long-term mortality risk. Taking into consideration that management and follow-up modalities of NCEs patients are not standardized, this score can be a valuable tool to help optimize follow-up strategies for NCEs patients.

Another relevant point is that the differential diagnosis of NCEs is still extremely challenging even at referral centers, as in almost two-fifths of patients, an underlying etiology could not be identified and a diagnosis of idiopathic VA was made (1,14,35). This highlights the unmet need for moving toward precision medicine to identify underlying pathogenic mechanisms in these patients, which may be helpful for developing targeted effective therapies and biomarkers for differential diagnosis (46,49).

In this regard, only a small minority of patients received a diagnosis of NCEs based on both clinical and molecular findings, with most patients diagnosed based on clinical criteria alone. This aspect is even more relevant considering that more than one-fifth of patients lacked clinical response and more than one-third had unsatisfactory histological response to available treatments. However, they also reflect underlying real-world clinical practice and the difficulty of managing these rare and complex disorders even at referral centers.

The difficulty in effectively managing these patients is reflected by the fact that almost one-sixth of patients were maintained on a GFD, despite CD having in some cases been excluded. In most cases, a GFD was started in the suspicion of seronegative CD before the correct diagnosis was reached. Patients were started on a GFD because of the severity of the clinical picture, and some of them even after reaching a different diagnosis preferred to continue a GFD. Some of them had some symptomatic benefit to a GFD, which has already been reported in the literature (14). This reflects the difficult differential diagnosis in these patients based on real-world data spanning 3 decades. Systemic steroids were among the most commonly and broadly used treatments in our cohort across almost all etiologies, despite the well-known side effects associated with their use. It is also worth noting that among all treatments used in our cohort, systemic steroids were the only one associated with increased mortality, although it is unclear if this might be due to the long-term adverse events of this treatment or if it instead reflects the severity of the underlying enteropathy, leading to the use of systemic steroids.

When an underlying etiology could be found, the leading causes included medications (ARBs, immune checkpoint inhibitors and immunosuppressants), CVID, infections, and AIE. Other less common etiologies were extremely heterogeneous but also included rare conditions with a poor prognosis, such as primary intestinal T-cell lymphomas.

Interestingly, the analysis comparing mortality of NCEs with that of the general population revealed that drug-related enteropathies were the only type of NCEs showing no evidence of increased mortality compared with the general population. However, it should also be noted that prognosis was heterogeneous among drug-related enteropathies, with enteropathy due to ARBs having a benign prognosis, but enteropathy due to immune checkpoint inhibitors and immunosuppressants having a more unfavorable prognosis, likely because of the underlying condition for which these medications were prescribed. Another possible explanation may be that while ARBs can easily be substituted with other antihypertensive medications, this is not always the case for immune checkpoint inhibitors and immunosuppressants. Moreover, data suggest that mortality due to Common Terminology Criteria for Adverse Events grade ≥3 GI adverse events of immune checkpoint inhibitors is low overall (<10%) (50). On the other hand, small bowel involvement because of immune checkpoint inhibitors is still poorly studied (40) and development of severe malabsorption can worsen the prognosis of these patients. Finally, the issue is also complicated by immune checkpoint inhibitors often being the last hope for prolonging survival in these patients with cancer, as it can lead to discontinuation of treatment with immune checkpoint inhibitors.

Our study has many strengths, which include being the largest study to date on NCEs, a very long follow-up period reflecting real-world clinical practice at many referral centers across the world, and the development of a clinical score to predict risk of poor outcomes in these patients. However, we must also acknowledge several limitations. These were mainly linked to the rarity of these conditions and the retrospective study design spanning more than 3 decades across many international referral centers. These included the presence of some missing data, which required imputation of data to address, and most patients being diagnosed and managed according to the clinical expertise of each center, as a consensus on clinical criteria for diagnosing and managing these conditions was not available. Given that some variables under study could only be evaluated at different points during follow-up, we cannot exclude influence of potential biases because of this. Moreover, our clinical risk score still would require a validation on a separate prospective cohort. However, to be realistic, reaching the sample size needed in a prospective study would be a very difficult task; therefore, this retrospective study would be as close to the reality as possible. Considering the long timespan of the study across 9 centers in different countries, we cannot exclude the possibility of remaining unmeasured confounders because of the diverse data sources used in this study. It was also not possible to adjust our survival analysis for immortal time bias, although the impact of this is likely limited as time-dependent variables (e.g., clinical/histological response) were measured early in follow-up, and very few deaths occurred before clinical/histological reassessment were completed. Potential differences in the diagnostic approach and management across participating centers over time should also be considered a potential limitation. Finally, regarding the analysis comparing mortality of NCEs compared with general population data, we had to exclude Indian patients from this analysis because we could not retrieve reference data for the Indian general population.

In conclusion, NCEs are a group of heterogeneous conditions with a poor long-term prognosis overall and a significantly increased mortality compared with the general population. We have identified risk factors of poor outcomes and developed a clinical score to help optimize management and follow-up strategies in patients with NCEs, although this is only a first step in personalizing the care of these patients and molecular approaches based on precision medicine are needed in the future to aid differential diagnosis and development of targeted therapies.

CONFLICTS OF INTEREST

Guarantor of the article: Federico Biagi, MD.

Specific author contributions: A.S. and F.B.: planned the study. A.S., S.M., D.S., S.R., K.E.I., C.M.J., A.R., C.M., G.M., M.V.L., A.D.S., G.C., U.V., F.Z., G.M., G.B., G.M., L.M., P.D., K.L., S.S.C., D.S.S., and F.B.: took care of the patients and collected the data. S.M.: performed the statistical analysis. A.S.: and S.M.: interpreted the data and drafted the manuscript. All authors critically revised and accepted the final version of the manuscript.

Financial support: None to report.

Potential competing interests: None to report.

Study Highlights.

WHAT IS KNOWN

• ✓ Nonceliac enteropathies (NCEs) are rare conditions, and data on their follow-up and long-term prognosis are scarce.

WHAT IS NEW HERE

• ✓ NCEs patients are characterized by a significantly increased mortality compared with the general population.

• ✓ Lack of clinical and histological response to therapies and major complications (lymphoproliferative disorders, major infections/sepsis, and others, such as abdominal malignancy, bowel obstruction, and perforation) are common and significantly affect the prognosis of NCEs.

• ✓ A clinical score based on age at diagnosis, anemia at time of diagnosis, and lack of clinical and histological response to therapies was developed to stratify NCEs patients according to their risk of poor outcomes.

• ✓ Strategies for clinical and endoscopic follow-up of patients with NCEs can be guided by the results of the score.

Supplementary Material

acg-120-2660-s001.docx ^{(24KB, docx)}

acg-120-2660-s002.docx ^{(16.3KB, docx)}

acg-120-2660-s003.docx ^{(14.5KB, docx)}

acg-120-2660-s004.pdf ^{(289.4KB, pdf)}