Patient characteristics

In all, 465 patients were enrolled. After screening and eligibility assessment, 392 UC patients were included (Fig. S1). The median observation period from the date of ileostomy takedown was 8.5 years (IQR, 4.8–12.7). The patient characteristics are shown in Table 1.

Table 1 Patient characteristicsPrevalence of endoscopic phenotype of the Chicago Classification

The number of pouchoscopies was 392 at the initial postoperative endoscopy, and 258 at 3 years and 99 at 10 years after the ileostomy takedown. The median time between the date of ileostomy takedown and the date of the initial postoperative pouchoscopy was 1.0 year (IQR, 0.54–2.3). The most common initial phenotype observed at the first postoperative endoscopy was focal inflammation of the pouch body (59%), followed by cuffitis (50%), inlet involvement (34%), AL involvement (20%), normal phenotype (20%), diffuse inflammation (16%), and pouch-related fistula (1.0%) (Table 2).

Table 2 Pouch outcomes and endoscopic phenotypes

Meanwhile, the most common overall endoscopic phenotype, which was determined by postoperative endoscopic examinations at all time points, was cuffitis (58%), followed by focal inflammation (55%), inlet involvement (46%), diffuse inflammation (27%), AL involvement (26%), normal phenotype (15%), and pouch-related fistula (4.8%) (Table 2). The most common type of fistula was perianal fistula, and the most common treatment was antibiotics (Table S2). We also evaluated the postoperative management of patients with CDLPI, defined as a pouch with AL involvement, stenosis at any anatomic location of the pouch, or pouch-related fistula [11]. We identified 110 patients with CDLPI, who were more likely to be treated with antibiotics and tumor necrosis factor inhibitors. Furthermore, the risk of diverting stoma and pouch excision was significantly higher in patients with CDLPI compared to those without CDLPI (Table S3), consistent with previous studies. [12]

Comparison of patient characteristics between our cohort and the University of Chicago cohort [6] showed that the percentage of patients undergoing 3-stage IPAA was significantly higher in the University of Chicago cohort than in our cohort, whereas the rate of colorectal neoplasia and hand-sewn anastomosis was significantly higher in our cohort than in their cohort. Additionally, pre- and post-operative data showed that patients with UC at the University of Chicago were more frequently treated with immunosuppressive therapies and were more likely to experience postoperative complications and pouch excision compared to our cohort (Table S4). Endoscopic phenotype data from the University of Chicago [6] showed that the proportion of normal phenotype was significantly higher in our cohort than in the University of Chicago cohort. Notably, the frequencies of pouch-related fistula and AL involvement were significantly lower in our cohort (4.8% and 22.4%, respectively) compared to the University of Chicago [6] (18.6% and 30.4%, respectively). The frequencies of the other phenotypes were comparable between the two cohorts (Fig. 1, Table S4).

Fig. 1

Prevalence of endoscopic phenotypes according to the Chicago Classification. The prevalence of phenotypes was compared with the published data of 382 UC patients from the University of Chicago. To compare the phenotype data between our multicenter cohort and the University of Chicago cohort, the denominator for phenotype prevalence was the number of overall patients in this analysis

Predictability of initial phenotypes and contributing factors for chronic pouchitis

The frequency of chronic pouchitis was 32% (Table 2). A KM curve showed that the 10-year chronic pouchitis-free survival rate was 71.7% (95% CI 65.3–78.7%) (Fig. 2a).

Fig. 2

Kaplan–Meier curves evaluating chronic pouchitis-free survival for a overall population and b focal inflammation of the pouch body observed at the initial postoperative pouchoscopy. Kaplan–Meier curves evaluating pouch survival for c overall population and d inlet involvement observed at the initial postoperative pouchoscopy

We examined which initial phenotype in the first postoperative scope was associated with subsequent risk of chronic pouchitis (n = 317). This examination showed that patients with an initial phenotype of focal inflammation had a significantly higher risk of chronic pouchitis over time (P = 0.0019), whereas patients with the normal phenotype as an initial phenotype had a significantly lower risk of chronic pouchitis (P = 0.0035) (Fig. 2b, Fig. S2). The same significant results were obtained even in asymptomatic patients (PDAI clinical subscore 0, n = 119) at the initial pouchoscopy (Fig. S3). The Cox proportional hazards model including the 6 factors with the smallest P-values (Table S5) showed that an initial phenotype of focal inflammation was a significant predictor of chronic pouchitis (HR, 2.2; 95% CI 1.1–4.4; P = 0.033) (Table 3). Hand-sewn anastomosis was inversely associated with the risk of chronic pouchitis (HR, 0.55; 95% CI 0.31–0.98; P = 0.042) (Table 3), indicating that the residual rectal cuff resulting from stapled anastomosis may be associated with chronic pouchitis.

Table 3 Cox proportional hazards model to assess factors predicting pouch outcomes

To understand which endoscopic items increase the risk of chronic pouchitis, we performed a subgroup analysis of patients with an initial phenotype of focal inflammation and showed that patients with chronic pouchitis were more likely to have erythema/edema at the distal pouch body than were patients without chronic pouchitis (Table S6 and Fig. S4), although not significant (P = 0.083). Whilst the inflammatory phenotypes and their number in the initial pouchoscopy did not differ, the number of inflammatory phenotypes as well as the percentages of diffuse inflammation, cuffitis, and pouch-related fistula as overall phenotypes were significantly higher in patients with chronic pouchitis than in those without chronic pouchitis. Notably, 30% of patients who later developed chronic pouchitis had a progression from focal to diffuse inflammation of the pouch body (Table S6). Images of a representative case are shown in Fig. S5.

Next, we analyzed the contributing factors to chronic pouchitis. To evaluate the phenotype that developed during the entire clinical course, the overall phenotype was included in this analysis. The rates of AL involvement, inlet involvement, diffuse inflammation of the pouch body, cuffitis, and pouch-related fistula as overall phenotypes were significantly higher in patients with chronic pouchitis than in those without it (Table S7). On logistic regression analysis, chronic pouchitis was significantly associated with diffuse inflammation of the pouch body (OR, 3.8; 95% CI 1.4–10.5; P = 0.009) and cuffitis (OR, 2.9; 95% CI 1.3–7.3; P = 0.015) as the overall endoscopic phenotype. Conversely, age at colectomy was inversely associated with risk of chronic pouchitis (OR, 0.97; 95% CI 0.943–0.996; P = 0.028) (Table 4).

Table 4 Logistic regression analysis to assess factors contributing to pouch outcomes

All these findings suggest that patients with an initial phenotype of focal inflammation were more likely to experience chronic pouchitis when the focal inflammation subsequently progressed to diffuse inflammation of the pouch body.

Predictability of initial phenotypes and contributing factors for pouch failure

The rate of pouch failure was 4.9%, and DLI and pouch excision were conducted in 4.9% and 0.8% of cases, respectively (Table 2). A KM curve showed that the 10-year pouch failure-free survival rate was 95.3% (95% CI 92.9–97.8%) (Fig. 2c).

We found that patients with an initial phenotype of inlet involvement had an increased risk of pouch failure over time (P < 0.001, n = 376) (Fig. 2d, Fig. S6). Similar trends were observed in asymptomatic patients at the initial pouchoscopy (n = 121) (Fig. S7). The Cox proportional hazards model including the 4 factors with the smallest P-values (Table S8) showed that an initial phenotype of inlet involvement was a significant predictor of pouch failure (HR, 6.3; 95% CI 1.3–30.8; P = 0.023) (Table 3).

Our subgroup analysis of patients with an initial phenotype of inlet involvement showed that the rate of inlet ulcers was 2 times higher in patients with pouch failure than in those without it (P = 0.055) (Table S9). KM curves showed that patients with inlet ulcers had a significantly increased risk of pouch failure over time (P = 0.034) (Fig. S8). Whilst the number of inflammatory phenotypes in the initial endoscopy did not differ, the number as well as the rate of pouch-related fistula in the overall phenotypes was significantly higher in patients with pouch failure than in those without it. Among 9 patients with an initial phenotype of inlet involvement who later developed pouch failure, 5 patients (56%) and 1 patient (11%) developed pouch-related fistula and inlet stenosis, respectively (Table S9). Images of a representative case are shown in Fig. S9.

In terms of contributing factors to pouch failure, the rates of AL involvement, inlet involvement, and pouch-related fistula as overall phenotypes were significantly higher in patients who experienced pouch failure than in those who did not (Table S10). Logistic regression analysis showed that the risk of pouch failure was significantly associated with an overall phenotype of pouch-related fistula (OR, 13.8; 95% CI 3.0–67.9; P < 0.001) (Table 4).

This result demonstrated that inlet involvement, especially inlet ulcer, is a predictor for pouch failure and is likely to be complicated by pouch-related fistula. Furthermore, our cohort had a lower risk of pouch excision than that of the Chicago Classification study (10.7%) [6], which may be attributable to the low frequency of pouch-related fistula.

Contributing factors for acute pouchitis

The frequency of acute pouchitis was 31% (Table 2). A KM curve showed the 10-year acute pouchitis-free survival rate to be 85.5% (95% CI 80.5–90.8%) (Fig. S10a).

Our initial phenotype analysis showed no specific initial phenotype associated with the risk of acute pouchitis (Fig. S10). The frequencies of AL involvement, inlet involvement, and diffuse inflammation of the pouch body as overall phenotypes were significantly higher in patients with acute pouchitis than in those without it (Table S11). Logistic regression analysis showed that the risk of acute pouchitis was significantly associated with hand-sewn anastomosis (OR, 4.2; 95% CI 2.3–8.1; P < 0.001) and an overall phenotype of inlet involvement (OR, 4.1; 95% CI 2.0–8.6; P < 0.001) (Table S12).

Correlation between endoscopic phenotype and clinical subscore of PDAI

To evaluate the correlation between clinical symptoms and endoscopic phenotypes, we combined 672 scopes with available clinical subscores of the PDAI regardless of the postoperative timing. Our analysis showed that patients with a subscore of 4–6 had a significantly higher rate of multiple inflammatory phenotypes (82%) than did patients with a subcore of 0 or 1–3 (52% and 52%, respectively). The proportions of patients with AL involvement, inlet involvement, and diffuse inflammation were significantly higher, whereas the rates of normal phenotype and focal inflammation were significantly lower in patients with a subscore of 4–6 than in those with a subscore of 0 or 1–3 (Table 5).

Table 5 Clinical subscore of pouchitis disease activity index and endoscopic phenotypes