Epidemiology, treatment patterns, and associated risk factors in perianal fistulizing Crohn’s disease: A systematic literature review

Abstract

BACKGROUND

Data regarding complex Crohn’s perianal fistulas (CPF) epidemiology are limited, and optimal treatment strategies are elusive. An improved understanding of how CPF treatment options are used in the real-world setting and factors associated with CPF development, treatment failure, and reasons for undergoing multiple surgeries may help to inform optimal patient management strategies, reduce treatment burden, and improve outcomes in patients with CPF.

AIM

To describe the epidemiology, treatments, outcomes, and associated risk/protective factors for complex CPF.

METHODS

Electronic databases (MEDLINE, EMBASE, EBM Reviews, EconLit) were searched. Two reviewers independently used populations, interventions, comparators, outcomes, study designs, and time criteria to identify relevant studies. Observational studies published in English from January 1, 2015 to February 17, 2022 with > 50 patients were included, even if complex CPF was not defined. Items of interest included complex CPF definitions, epidemiology, treatment patterns, morbidity, mortality, and risk factors associated with complex CPF development, treatment failure, and undergoing multiple surgeries. Data were reported using descriptive statistics.

RESULTS

Overall, 140 studies were included. Complex CPF definitions were heterogeneous and rarely reported (24 studies). Hence, data mostly related to CPF in general. CPF prevalence was variable (range: 1.5%-81.0%). Incidence was wide-ranging and mostly reported cumulatively at 1-year post-Crohn’s disease diagnosis (range: 3.5%-50.1%). Overall healing and failure rates after treatment were in the range of 10.5%-80.2% and 3.6%-83.0%, respectively. Abscesses were the most frequently reported morbidity (n = 18). No CPF-related deaths were reported. No consistent risk or protective factors were identified.

CONCLUSION

Epidemiology, treatment patterns, and risk factors for CPF vary, likely due to inconsistent CPF and clinical outcome definitions. Standardization would facilitate comparability, which may inform optimal complex CPF treatment strategies.

Key Words: Complex perianal fistula; Crohn’s disease; Epidemiology; Treatment patterns; Morbidity; Risk factors

Core Tip: This systematic literature review summarized data from recent studies reporting the epidemiology of Crohn’s perianal fistulas (CPF), treatment patterns, therapeutic response, morbidity, and mortality alongside risk/protective factors associated with CPF development, treatment failure, and reasons for undergoing multiple CPF-related surgeries. Substantial variability in outcomes was observed, indicating the need for standardized CPF study designs and consistent definitions for CPF and CPF-related clinical outcomes to achieve optimal treatment strategies and improved patient outcomes.

INTRODUCTION

Crohn’s disease (CD) is a form of inflammatory bowel disease characterized by chronic inflammation of the gastrointestinal tract[1]. Crohn’s perianal fistulas (CPF) are a debilitating complication of CD that cause pain and drainage of pus, stool, or blood from the fistula openings and can have a significant impact on a patient’s quality of life[2-4]. CPF are classified as ‘simple’ or ‘complex’ based on the location of the fistula, level of sphincter involvement, and number of tracts[5,6]. Although classifications vary, the American Gastroenterological Association (AGA) defines complex fistulas as those that are high (high intersphincteric, transsphincteric, extrasphincteric, or suprasphincteric), have multiple external openings, or are associated with the presence of a perianal abscess, anorectal stricture, or rectovaginal fistula[7]. Owing to limited data availability on CPF prevalence and incidence, the global epidemiological disease burden remains unclear[8].

Treatment of complex CPF requires both medical and surgical approaches, which aim to achieve sustained fistula closure while preserving fecal continence[3,9]. However, complex CPF is challenging to treat and rates of long-term healing are modest, rates of fistula recurrence are high, and optimal treatment strategies are yet to be established[8,10,11]. Current treatment options can also place significant burden on patients. For example, medical therapies can increase the risk of infections, and surgical failure can lead to additional surgical interventions, which increase the risk of complications [e.g. fecal incontinence (FI)] and may result in proctectomy[8,12,13].

Evidence from research and clinical practice indicate that multiple risk factors are associated with CPF; however, few reliable predictors of CPF have been identified[14,15]. An improved understanding of how CPF treatment options are used in the real-world setting and factors associated with CPF development, treatment failure, and reasons for undergoing multiple surgeries may help to inform optimal patient management strategies, reduce treatment burden, and improve outcomes in patients with CPF.

This systematic literature review (SLR) aimed to describe the epidemiology of complex CPF in adult and pediatric populations and the treatments, outcomes, and risk/protective factors associated with the disease.

MATERIALS AND METHODS

Data sources, search strategy, and study selection

This SLR was conducted in accordance with the guidelines of the Cochrane Collaboration and the Centre for Reviews & Dissemination (York, United Kingdom), and the National Institute for Health and Care Excellence for evidence synthesis[16-18]. Electronic databases (MEDLINE, EMBASE, EBM Reviews, and EconLit) were searched for relevant articles published between January 1, 2015 and February 17, 2022 using predefined search algorithms (Supplementary Table 1 for the full search algorithm). A manual bibliographical check of included articles was conducted at the full text screening stage, and all identified review articles, guidelines, SLRs, or meta-analyses were screened for relevant original publications not identified by the electronic database search. Titles and abstracts were assessed independently by two reviewers for inclusion according to predefined population, interventions, comparators, outcomes, study designs, and time selection criteria. After full texts of relevant publications were obtained, two reviewers independently assessed study eligibility. Any disagreement was discussed, and if necessary a mediator was involved in making a final decision.

Inclusion/exclusion criteria

Included articles were observational studies with > 50 patients with CD and complex CPF published in English. To avoid missing relevant information due to variability in terminology used to describe CPF, studies with synonyms for complex CPF, such as ‘anal, perianal, and rectal fistulas’ without the specification term ‘complex’ were also included. Therefore, hereafter the term ‘complex CPF’ refers to results from studies in which patients were described as having complex CPF. Any clinical trials (including randomized clinical trials, single-arm trials, and phase 1 trials), animal/in vitro studies, and case reports were excluded.

Data extraction and quality assessment

Two independent reviewers followed an agreed data extraction procedure. Data were extracted for patients with complex CPF, for mixed populations of complex and simple CPF, and for patients for which a CPF type was not clearly defined. If studies were reported in several publications, the leading publication was fully extracted and subsequent publications were checked for additional or contradictory results. Principal publications were defined as the first publication to report the main results from a study. The National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies was used to assess the risk of bias in and quality (good, fair, poor) of included articles.

Outcomes of interest

Outcomes of interest included article, study, patient, and disease characteristics, epidemiology (prevalence and incidence) of complex CPF, treatment patterns and outcomes (comprising treatment patterns and rates of healing/failure), morbidity (abscess, anorectal stricture/stenosis, FI, proctitis and/or continuous pain in the anal area, and sepsis), mortality and risk factors associated with complex CPF development, treatment failure, and undergoing multiple surgeries.

Data analysis

Although the aim of this SLR was to assess complex CPF, data for mixed and unknown CPF populations were also considered. For analyses adults were defined as any participant aged at least 18 years or otherwise categorized as adult patients; studies in which participant age was not reported were considered adult studies. Pediatric participants were those aged under 18 years or otherwise categorized as pediatric in a study. Data were summarized using descriptive statistics. Full details of the data analysis undertaken for each of the outcomes from this SLR are provided in the Supplementary material.

RESULTS

Article, study, patient, and disease characteristics

In total 1966 abstracts were identified (Supplementary Figure 1 for the full PRISMA diagram). After screening 140 studies (principal publications: 79 full texts and 61 conference abstracts) remained that reported on the following outcomes: Epidemiology (n = 86); treatment patterns/failures/responses (n = 63); risk/protective factors associated with CPF development (n = 28); treatment failure (n = 34); and multiple surgeries (n = 8); morbidity (n = 30); and mortality (n = 5) (Supplementary Table 2 for a full list of included studies, including full references, study types, and outcomes associated with each study). Most studies were conducted in Europe (n = 52), Asia (n = 46), or North America (n = 28) and were cohort (n = 103), cross-sectional (n = 36), or case-control (n = 1) in design. Most studies were rated as fair quality (n = 98) with respect to risk of bias and scientific quality because many did not provide enough information to be evaluated per the assessment criteria. The remaining studies were rated good (n = 14) or poor (n = 28).

For epidemiology studies (Supplementary Table 2) sample sizes (based on the number of patients with CD) ranged from 53 patients[19] to 187070 patients[20]. For studies with other outcomes, sample sizes were based on the number of patients with CPF and ranged from 50[21] to 1773[22] for treatment patterns, 57[23] to 1065[24] for morbidities and mortality, and 53[25] to 3623[26] for risk/protective factors.

Overall, 86 studies assessed outcomes from adult populations (including those where age was not reported), 16 from pediatric-only populations, and 38 from both adult and pediatric populations (Supplementary Table 2). Mean participant ages (reported in 55 studies) ranged from 11 years[19] to 58 years[27]. For adult studies, mean patient age (reported in 35 studies) was 37.2 years [standard deviation (SD): 5.4], and for pediatric (n = 6) and mixed population studies (n = 13), the mean age was 13.1 years (SD 1.9) and 27.0 years (SD 5.9), respectively. In total, 106 studies reported sex distribution, of which most (n = 70/106) had a higher proportion of males [median 55.7%; interquartile range (IQR) 46.4%-66.9%] when compared with females (median 44.3%; IQR 33.1%-53.7%). Median CD disease duration assessed from date of diagnosis was reported in 21 studies (30.5 months, IQR 11.9-118.2 months), while the median duration of CPF was only reported in four studies (64.4 months, IQR 35.2-127.4 months).

Definitions for complex CPF were provided in 24 studies and were variable, with studies using between 1 and 10 characteristics to define complex CPF. Overall, the definition from the AGA for complex CPF was the most frequently used definition (n = 5 studies), with the presence of multiple fistula tracts/external openings (n = 18 studies, Table 1) being the most frequently cited characteristic of complex CPF. Most studies (n = 97/140, 69.3%) did not report the proportion of patients with complex CPF; of those that did (n = 43) the median proportion of patients with complex CPF within the overall CPF population was 72.7% (IQR: 52.3%-88.6%).

Table 1 Definitions of complex Crohn’s perianal fistulas from included studies.

Included in definitions for complex CPF	Number of studies
Multiple fistula tracts/external openings1	18
Presence of rectovaginal fistula1	17
High position1	16
High transsphincteric1	16
Signs of perianal abscess (including pain or fluctuation)1	16
High intersphincteric1	15
Anorectal stricture1	14
Suprasphincteric1	13
Extrasphincteric1	12
Endoscopic activity1	6
Supralevator extension	3
Proctitis	2
Horseshoe fistula	2
Procedure as surrogate for complex CPF (ostomies, seton drainage)	2
Perianal disease activity index (not further defined)	1
Translevator extension	1

¹Included in the definition of complex Crohn’s perianal fistulas by the American Gastroenterological Association.

CPF: Crohn’s perianal fistulas.

Epidemiology

Across all studies CPF prevalence in patients with CD varied widely [range 1.5%[28] to 81.0%[29] (IQR: 10.7-26.7%)] and was reported across multiple time points, spanning from before study enrollment to end of follow-up (Supplementary Figure 2 for CPF prevalence at each time point). When considering studies reporting prevalence from primary epidemiology (defined as those whose main objective or primary outcome was incidence/prevalence of CPF; n = 19) and population-based (n = 36) studies only, large variations in CPF prevalence were still reported [range: 1.5%[28] to 43.3%[30], (IQR: 5.5%-21.6%) and range: 1.5%[28] to 44.8%[31] (IQR: 8.3%-20.6%), respectively].

Similar inconsistency was observed across geographical regions, with prevalence ranging from 1.5%[28] to 55.3%[32] (IQR: 9.5%-27.8%, n = 27) for European studies, 5.6%[33] to 61.2%[34] (IQR: 13.7%-42.2%, n = 29) for Asian studies, and 3.3%[20] to 81.0%[29] (IQR: 7.7%-20.5%, n = 17) for North American studies. Subgroup analysis by age also showed that prevalence varied across adult (n = 44), pediatric (n = 15) and mixed population (n = 25) studies (Figure 1). Of the studies defining complex CPF, 18 reported prevalence, which ranged from 3.4%[35] to 54.6%[34]. Complex CPF was most commonly reported at cross-sectional time points (i.e. before enrollment, before and/or at time of survey, before and/or after CD diagnosis, after CD diagnosis, or throughout follow-up), ranging from 3.4%[36] to 48.2%[32] (IQR: 6.2-28.8%, Supplementary Table 3 for a breakdown of complex CPF prevalence for all time points).

Figure 1 Crohn’s perianal fistula prevalence in adult, pediatric, and mixed populations across study types and geographical location. Data are presented as range (interquartile range) for study type and median [range (interquartile range)] for geographical region. Only studies reporting relevant data are included.

CPF incidence was reported less frequently than prevalence (n = 15, Supplementary Table 2). Cumulative incidence was reported at several time points, varying from 1 to 25 years (n = 12 studies), and was most commonly reported at 1-year post-CD diagnosis (n = 8)[25,36-42], ranging from (median, IQR) 3.5%[37] to 50.1%[38] (11.0%, 8.1%-20.1%). Only one study reported CPF incidence at 20 or 25 years, with incidence rates of 21.5%[36] and 23.9%[43], respectively (Supplementary Table 4 for a breakdown of CPF incidence at each time point). Of those reporting CPF incidence throughout study follow-up, two reported incidence in person-years (PY), ranging from 133.83 per 100000 PY[44] to 1100 per 100000 PY[28], while the other reported CPF incidence at 10% over a 6-year follow-up period[22].

Primary epidemiology studies (n = 9)[22,25,28,35-37,39,41,44] generally reported lower rates of CPF incidence when compared with non-primary epidemiology studies and most commonly reported cumulative CPF incidence at 1 year after CD diagnosis (n = 5)[25,36,37,39,41], with medians (IQR) ranging from 3.5%[37] to 35.2%[41] (11.0%, 8.3%-11.0%). Incidences of CPF at 1 year after CD diagnosis in adult studies (n = 2) were relatively homogeneous, ranging from 8.3%[36] to 11.0%[39] (median 9.7%, IQR: 9.0%-10.3%); however for pediatric [n = 3, range 3.5%[37] to 50.1%[38] (median 7.5%, IQR: 5.5%-28.8%)] and mixed study populations [n = 4, range 11.0%[25] to 35.2%[41] (median 13.5%, IQR: 11.8%-20.1%)] incidence at this time point was variable.

Treatment patterns

Treatment frequency: Overall, 69 CPF treatment interventions were identified, which were grouped into 16 categories (antibiotics, biologics, immunomodulators/-suppressants, stem cell therapy, steroids, colectomy/-ostomy, cutting seton, fistulectomy/-otomy, flap, glue, loose seton, plug, surgery, surgery [without seton], multimodal approach, and other) (Supplementary Table 5 for a summary of which interventions were attributed to each category). Of the 50 studies assessing patients receiving at least one of the interventions of interest, the most frequently reported interventions were biologics (n = 43), loose seton (n = 35), surgery (without seton) (n = 31), immunosuppressants (n = 30), and fistulectomy/-otomy (n = 24). Across all study settings, loose seton was the most commonly received first or second intervention.

Among medical treatments immunosuppressants and biologics were the most frequently received first and second study interventions, respectively (Figure 2). Compared with adults a higher proportion of pediatric patients underwent surgery as their first intervention (IQR: 3.8%-30.0% and IQR: 22.9%-41.9%, respectively). Conversely, a higher proportion of adults compared with pediatric patients experienced surgery as their second study intervention (IQR: 4.9%-27.4% and IQR: 3.9%-6.5%, respectively). A higher proportion of pediatric patients received biologics as their first study intervention when compared with adults (IQR: 63.0%-91.5% and 6.3%-49.4%, respectively).

Figure 2 Proportions of patients with Crohn’s perianal fistulas receiving one of the top five most frequently reported interventions used as first, second, or third treatment within a study. Numbers in the key represent the number of studies reporting the intervention as the first, second, and/or third treatment in a study; some studies reported an intervention in more than one line of treatment. Numbers above the bars are median (interquartile range) proportion of patients; number of studies. Loose seton was not reported as a third treatment in any studies. IQR: Interquartile range.

Treatment lines: In total, 15 studies provided details on the treatment lines used to treat patients with CPF (Figure 3). Loose seton was the most common first-line study intervention (n = 11), followed by immunomodulators (n = 5) and biologics (n = 4). Of those undergoing loose seton as a first-line study treatment, biologics, advancement flap, and glue were the most common second line treatments (n = 2 each). Surgical interventions other than seton use tended to be used after medical treatments such as immunomodulators/immunosuppressants and biologics.

Figure 3 Treatment algorithms showing the first, second, third, and fourth treatments among studies showing sequential treatments (n = 15). The first treatment in a study is presented on the left of the figure with the number of corresponding treatment algorithms. The lines then flow to the next treatment mentioned in the studies; line thickness corresponds to the number of treatment lines. Of note one study may report more than one treatment per line; for example, one study by Graf et al[49] reported that patients underwent loose seton placement as a first treatment followed by fistulectomy/-otomy, advancement flap, or plug, representing three different treatments as a second treatment in this study.

Healing and treatment failure by intervention

Due to heterogeneity in definitions used for healing across the identified studies, healing was grouped into four categories (fistula healing/closure, partial response, radiological healing, and failure-free). Overall, healing rates were identified for 14 interventions across 34 studies, with follow-up times ranging from 3 months[45] to 145 months[46] (Supplementary Tables 2 and 5). In the most commonly reported healing categories, rates of fistula healing/closure and partial response for the top five most frequently reported interventions varied widely ranging from 8.8%[47] to 90.3%[48] (Table 2). Of the top five interventions for which healing rates were reported, biologics were the most frequently reported intervention (n = 17 studies, range 8.8% to 78.4%); however, fistulectomy/-otomy had the highest rates of fistula healing/closure (n = 4 studies, range 47.9%[49] to 90.3%[48]). Time to healing was reported in five studies, with median times to healing ranging from 5.5 months (biologics; 50.0%[50]) to 22 months (colectomy/-ostomy; 80.0%[21]), while mean times to healing ranged from 3.4 months (immunosuppressants; 47.6%[51]) to 36.7 months (biologics; 23.1%[52]).

Table 2 Summary of healing and failure rates for the top five most frequently reported intervention categories.

Frequency of healing or failure category, range (number of articles)
	‘Healing’		‘Failure’
Category	Fistula healing/closure	Partial response	Active disease after treatment	Recurrence
Biologics	8.8%-78.4% (n = 17)	11.9%-68.3% (n = 7)	10.7%-55.6% (n = 10)	7.9%-45.0% (n = 5)
Loose seton	17.7%-60.5% (n = 5)	25.0% (n = 1)	41.7%-51.0% (n = 2)	4.0%-68.0% (n = 5)
Surgery (other than seton)	40.7%-75.0% (n = 6)	15.5%-41.6% (n = 3)	22.5%-42.0% (n = 3)	39.5%-64.0% (n = 3)
Immunosuppressants	25.0%-54.8% (n = 3)	19.1% (n = 1)	9.5% (n = 1)	28.6%-46.9% (n = 3)
Fistulectomy/-otomy	47.9%-90.3% (n = 4)	Not reported	3.4% (n = 1)	17.9% (n = 1)

Data presented are for the two most frequently identified healing and failure categories. Details of all the interventions for which healing and failure were assessed can be found in Supplementary Table 4.

Definitions for treatment failure also varied and were grouped into seven categories (active disease after treatment, recurrence, additional surgery, deterioration, discontinuation, loss of response, and treatment change/escalation). In total 33 studies reported failure rates across 12 intervention categories (Supplementary Tables 2 and 5). In the most commonly reported failure categories (active disease after treatment and recurrence; Table 2), biologics had the highest rate and widest range of active disease after treatment (n = 10 studies, range 10.7%[53] to 55.6%[54]); loose seton had the second highest rate of active disease (n = 2 studies, 41.7%[55] to 51.0%[56]) and the highest rate of recurrence after treatment followed by surgery (without seton).

Time to treatment failure was reported in eight studies[11,50-53,57-59], with median times to failure ranging from 1 month (advancement flap; recurrence 77.3%[11]) to 52 months (biologics; recurrence 45%[50]) and mean times to failure ranging from 3.3 months (antibiotics; recurrence 7.7%[51]) to 21.3 months (stem cell therapy; recurrence 10.0%[52]). In patients who had undergone surgery, the highest proportions requiring additional surgery were those who had undergone an advancement flap procedure (n = 2, range: 57.7%[60] to 60.0%[56]). The need for additional surgery ranged from 6.9%[61] to 33.3%[56] after fistulectomy/-otomy (n = 3[46,56,61]) and 14.0%[23] to 68.0%[56] after treatment with loose seton (n = 5[23,46,56,57,62]).

Treatment outcomes

Overall healing and failure rates: Overall healing (n = 18 studies) and failure (n = 24 studies) rates were described at various time points (range: 12.0-85.2 months and 5.5-120.0 months, respectively). Across all intervention categories, healing rates ranged from 10.5%[53] to 80.2%[58] and failure rates from 3.6%[63] to 83.0%[56]. Fistula healing/closure was the most frequently reported healing category (n = 14 studies) with healing rates in the range of 25.0%[64] to 63.2%[65], while CPF recurrence was the most commonly reported failure category (n = 15 studies) with failure rates in the range of 3.6%[63] to 83.0%[56] (Table 3).

Table 3 Summary of overall healing and failure rates per category.

	Articles, n	Sample size	Healing/failure1	Follow-up time/time to failure (months)
Healing category
Fistula healing/closure	14	50-253	25.0%-63.2%	12.0-85.2
Partial healing/improvement	3	58-65	10.5%-17.0%	12-45
Radiological healing	3	60-253	18.0%-57.0%	12-48
Failure-free	4	51-157	48.7%-80.2%	12-60
Failure category
Active disease after treatment	4	61-91	15.4%-47.7%	12-31
Additional surgery	4	51-365	13.8%-45.6%	6.8-90.0
Deterioration	1	22	22.0%	45.6
Discontinuation	1	12	12.0%	6
Loss of response	1	236	73.0%	48.0
Recurrence of CPF	15	50-161	3.6%-83.0%	5.5-120.0
Treatment change/escalation	1	51	43.1%	7

¹If cumulative probability of healing at several time points was provided, only the first time point was considered in the range of percent healing.

CPF: Crohn’s perianal fistulas.

Morbidity: In total, 30 studies reported on five morbidities of interest: Abscess (n = 16); anorectal stricture/stenosis (n = 14); FI (n = 6); proctitis and/or continuous pain in the anal area (n = 13); and sepsis (n = 4, Supplementary Table 2). Morbidity rates varied and were reported at different time points. Only data for the rates of abscess were available at all time points studied (e.g., at baseline and during follow-up); only one time point (at the point of surgery) had morbidity rates for abscess, anorectal stricture/stenosis, proctitis and/or continuous pain in the anal area, and sepsis. Data showing the proportion of patients experiencing each morbidity of interest across all time points identified are available in Supplementary Table 6 and summarized below.

The proportion of patients experiencing abscesses was assessed across multiple time points and ranged between 6.7%[49] and 87.7%[66]. All studies were conducted in adult or mixed (adult and pediatric) populations, with sample sizes ranging from 58 patients[67] to 465 patients[68]. Median follow-up times (n = 10) ranged from 3.0 months[45] to 85.2 months[49], although one study reported a median follow-up time of at least 12 months rather than a median for individual patient follow-up[55]. Three studies reported abscesses in patients with complex CPF only[55,67,69]. Although the prevalence of abscesses were similar in two of these studies (67.2% at enrollment[67] and 66.0% throughout follow-up[69]), one study reported a relatively lower proportion of patients with abscesses at baseline (24.0%), likely because only large abscesses (> 3 cm in size) were considered in this study[55].

The proportion of patients experiencing anorectal stricture/stenosis was also reported across multiple time points and ranged between 4.3%[52] and 31.7%[70]. Median follow-up times (n = 8) ranged from 36.0 months[11] to 153.6 months[39], and participant sample sizes ranged from 60 patients[11] to 465 patients[68]. Only one study assessed outcomes related to complex CPF only[11], with 22.0% of participants having anorectal stricture/stenosis at baseline.

The proportion of patients experiencing FI was assessed in six studies[49,53,63,64,67,71] across multiple time points, ranging between 3.4%[67] and 72.8%[71]. All studies were conducted in adults with sample sizes ranging from 55[63] to 451[67] patients and median follow-up times between 45.0 months[53] and 85.2 months[49]. With respect to complex CPF, only one study assessed FI in patients with complex CPF (at enrolment only) and reported a FI rate of 3.4%[67].

The proportion of patients with CPF who experienced proctitis and/or continuous pain in the anal area ranged from 25.6%[65] to 80.2%[66] over multiple time points. Patient follow-up (median, n = 8) ranged from 7.8 months[72] to 85.2 months[49] with sample sizes ranging from 60[11] to 253 patients[57,62]. Only one study assessed proctitis rates in patients with complex CPF[11], with 45% of patients reporting proctitis at baseline.

Rates of sepsis in patients with CPF ranged from 13.3%[56] to 43.7%[73]. Median follow-up times (n = 2) ranged from 31[56] to 40[73] months and the sample sizes from 57[23] to 83[56] patients. One study reported the proportion of patients whose sepsis was not adequately controlled by seton placement (13.3%[56]), while two studies reported rates of sepsis-related deaths only (1.5%[54] to 1.8%[23]). Finally, one study reported a 5-year recurrence rate of perianal sepsis requiring surgery of 43.7%[73].

Mortality: Mortality was assessed in five studies[23,24,53,54,74]; however, none reported death due to CPF. Mortality ranged from 0.0%[74] to 6.1%[53] across the five studies, with sepsis being the most commonly reported reason for death, either as a result of abdominal sepsis following intestinal fistula surgery or after seton placement to control local sepsis (n = 2 studies)[23,54].

Risk/protective factors

CPF development: In total 28 studies reported risk or protective factors associated with CPF development. Overall, 38 risk and 13 protective factors were identified, which were categorized as follows: Patient characteristics, n = 17 and n = 4; clinical characteristics, n = 17 and n = 3; and treatments, n = 4 and n = 6, respectively (Supplementary Tables 7 and 8 for the full list of risk and protective factors that were associated with CPF development).

The most common patient characteristics associated with increased risk of CPF development were male sex[31,37,41,68,75-77] and younger age at CD diagnosis[41,68,78-80]. In contrast, patient characteristics reported as being associated with a reduced risk of CPF development were female sex[28,80] and older age [> 60 years] at CD diagnosis[36,80]). Some results were contradictory including a retrospective study (n = 3623 participants) that identified female sex as an independent predictor of fistula development[26], and a retrospective analysis that found that for each additional year of age at CD diagnosis, the associated risk of developing CPF increased by 5% (n = 2214; participants were aged 5-24 years)[77]. The most common risks associated with CPF development identified in pediatric-only studies were diagnosis of CD between 6 and 10 years old[75], higher weight/body mass index (BMI) when compared with patients with CD aged 0-17 years without perianal fistulizing disease[31], male sex[37,75], and non-White race[40].

The most frequently reported clinical characteristics associated with an increased risk of CPF development were manifestation of CD in the colon (L2)[35,41,80-82], rectal involvement[25,35,79,81], and stricturing CD behavior (B2 phenotype)[35,80,81]. Protective associations included ileal involvement[80], isolated small bowel disease[81], and non-stricturing and non-penetrating CD behavior[81]. For pediatric-only studies, characteristics including high C-reactive protein levels[40], presence of granulomas on diagnostic biopsies and non-penetrating perianal lesions[40,83], procedures before CD diagnosis (including perianal lesion removal, fistulotomy, incision, and drainage of a perianal abscess)[40], and involvement of the upper gastrointestinal tract were all associated with a higher risk of CPF development[31].

Overall, four treatments, including the use of immunosuppressants[77,83], perianal examination under anesthesia with drainage performed within 12 months prior to magnetic resonance imaging[84], and steroid-sparing therapy[77] were associated with a lower likelihood of developing CPF. Data regarding antibiotics, antitumor necrosis factor (TNF)/biologics and/or corticosteroids were mixed, with some studies reporting these agents as being associated with an increased risk of CPF development[28,37,77,80], while others reported that these same treatments were associated with a decreased risk of CPF development[40,75,77,83]. Of note one abstract reported an association between antibiotics and an increased risk of CPF development, whereas the subsequent manuscript associated antibiotic use with a decreased risk of CPF development[77,85]. In pediatric-only studies corticosteroid use for CD treatment was identified as a risk factor associated with CPF development[37], whereas treatment with anti-TNF/biologics[40,83] or immunosuppressants[83] was associated with a reduced risk of CPF development.

Treatment failure

In total 34 studies reported risk or protective factors associated with treatment failure in patients with CPF. Overall, 34 risk and 39 protective factors were identified, which were categorized as follows: Patient characteristics, n = 5 and n = 4; clinical characteristics, n = 16 and n = 16; and treatments, n = 13 and n = 19, respectively (Supplementary Tables 9 and 10, respectively, for a full list of the risk and protective factors that were associated with treatment failure).

The most common patient characteristics identified as a risk factor associated with CPF treatment failure were a high BMI at baseline[40,84] and female sex[66,69]; however, one study also reported that a low BMI at baseline was associated with an increased risk of CPF treatment failure[55]. Male sex was associated with reduced levels of CPF treatment failure[86] as was a younger age at CD diagnosis[72] and younger age at surgery for patients with complex anal fistula[11]. Older age was reported as being associated with both increased[84] and decreased risk of CPF treatment failure[87].

Complex CPF was the most commonly reported clinical characteristic associated with CPF treatment failure (n = 5)[54,58,66,84,87] followed by stricturing phenotype (B2) (n = 4)[27,66,74,87], higher number of fistulas and/or tracts (n = 3)[45,69,84], a longer duration of CD since diagnosis (n = 3)[69,84,88], the presence of perianal abscess at baseline (n = 3)[36,55,65], and proctitis (n = 3)[72,87,88]. Protective clinical characteristics included a low C-reactive protein level, low CD severity/median Harvey Bradshaw Index score, and non-inflamed mucosa[49,86].

Data regarding treatments as risk or protective factors associated with treatment failure were in some instances contradictory. For example, anti-TNF was associated with a negative effect on fistula healing in two studies[40,51]; however, combination therapy of anti-TNFs with immunosuppressants was also reported as a contributing factor to CPF treatment success[59,88-90]. Additionally, long-term treatment with either anti-TNF or immunosuppressants was positively correlated with therapy success[69,88]. In total, 4 publications reported that controlling CPF via seton placement was a contributing factor to treatment success[40,56,88,89]. For pediatric-only studies (n = 1)[40], a high BMI at diagnosis was the only factor associated with CPF treatment failure.

Multiple surgeries for CPF

In total 8 studies reported risk or protective factors associated with undergoing multiple surgeries for CPF. Overall, 11 risk and 6 protective factors were identified, which were categorized as follows: Clinical characteristics, n = 6 and n = 0; and treatments, n = 5 and n = 6, respectively (Supplementary Tables 11 and 12, respectively, for a full list of the risk and protective factors that were associated with receiving multiple surgeries for CPF).

The most common clinical characteristics identified as risk factors associated with undergoing multiple surgeries were complex CPF (n = 4 studies)[54,57,62,91], proctitis (n = 3 studies)[57,62,91], and abscess (n = 2 studies)[65,91].

Data relating to treatments as risk/protective factors were limited; however, the concomitant use of steroids (n = 253 participants)[62] and infliximab maintenance therapy with more than three infusions (n = 117 participants)[65] were associated with a need for surgical reintervention. Anti-TNF therapy (two studies; n = 253 participants[62] and n = 116 participants[91]) and biologics (n = 517 participants)[76] were associated with a reduced need for additional surgery.

DISCUSSION

This SLR identified recent studies reporting CPF epidemiology, treatment patterns (including response to therapies), morbidity, and mortality alongside risk/protective factors associated with CPF development, treatment failure, and patients undergoing multiple CPF-related surgeries. In total 140 studies were included in the analysis, covering major geographies (Europe, North America, and Asia) and age groups.

Most studies had mixed populations of patients with CPF (i.e. patients had either simple or complex CPF), or there was insufficient information reported to determine the nature of CPF present in the study population. Therefore, to ensure relevant information relating to CPF outcomes were captured, all data were collected. Consequently, many of the outcomes reported here were related to CPF in general. There was a dearth of data for patients with ‘complex CPF only’, and only 24 studies provided definitions for complex CPF.

Collectively, definitions were heterogeneous, comprising 16 different characteristics with each study using between 1-10 characteristics to define complex CPF. Of note the top 10 most frequently used characteristics were derived from the definition from AGA for complex CPF; however, the application of these characteristics varied across studies, with only five using the full AGA definition for complex CPF. The remaining six characteristics identified were unrelated to the definition from AGA. These findings highlighted a clear need for consensus regarding the definition of complex CPF and for consistency in the application of definitions across studies. A uniform approach to CPF classifications will likely reduce some of the variability observed in reported study outcomes and will aid cross-study comparability. It may therefore be prudent to use the definition from AGA of complex CPF until expert consensus has been reached[92].

In a previous expert consensus process, a new classification for perianal fistulizing CD was proposed that includes stratification according to disease severity and outcome, synchronization of patient and clinician goals in decision-making, and identification of indications for curative fistula treatment, diverting ostomy, and proctectomy[93]. This is an important step towards a uniform approach to the classification of complex CPF; however, it may be more useful for guiding decision-making in daily practice than for patient selection and stratification in clinical trials.

Epidemiological data for both CPF prevalence and incidence were assessed in this SLR. The prevalence of CPF varied widely across studies, time points, and geography in part because of differences in study type, design, and study populations. For example, the study reporting the lowest rate of CPF prevalence (1.5%)[28] was a population-based study that used International Classification of Diseases codes to identify CD and CPF; in contrast, the study reporting the highest CPF prevalence rates (81.0%)[29] utilized a retrospective study design involving data from two inflammatory bowel disease referral centers and included only patients who had a known CD diagnosis and had undergone a proctectomy. The prevalence of CPF in this study may be higher than in the general population because patients with severe disease may be more likely to be referred to these two inflammatory bowel disease centers. However, CPF prevalence varied even when considering primary epidemiology and population-based studies only (1.5% to 44.8%)[28,30,31]. In addition, variation may have arisen because of the heterogeneous time points used in studies. Consistent time points are needed to permit cross-study comparisons and determine robust estimates for CPF prevalence and incidence.

CPF incidence was also varied. Two studies[28,44] reported incidence in PY, and one study reported incidence over a 6-year follow-up period; however, most studies reported cumulative incidence. As with prevalence these studies had heterogeneous study designs and patient populations, which likely accounted for some of the variability observed. For example, of two studies assessing CPF in pediatric patients at 1 year post-CD diagnosis, one[37] utilized the CEDATA-GPGE registry to derive a cumulative CPF incidence of 3.5%, while the other[38] found a cumulative CPF incidence of 50.1% using data from a single center.

Regarding treatment patterns for complex CPF, multiple surgical and medical interventions were identified in the literature; however, the implementation of an optimal or consistent approach to defining CPF treatment goals or failures was not apparent. For example, although loose seton was the most commonly used first intervention in a study, it also had the greatest variation with respect to the proportion of patients receiving the treatment among the top five interventions identified. Similar variability was also observed across other interventions [biologics, surgery (without seton), immunomodulators/-suppressants, and fistulectomy/-otomy]. This observation highlighted the potential heterogeneity with respect to treatment approaches in patients with CPF. However, such treatment pattern observations should be interpreted with caution. As few studies reported a patient’s treatment history, it was unknown whether patients were treatment naïve at enrollment. In addition, while all studies included patients with CPF, they rarely differentiated between whether specific treatments were provided for underlying CD or CPF itself, and therefore all treatments relevant to CPF were included unless they were specifically reported as a treatment for CD. Overall, treatment patterns were inconsistent across studies and difficult to interpret and compare owing to differences in study settings, patient populations, follow-up periods as well as a lack of patient medical history and consistent delineation of when and why patients received specific interventions.

Treatment outcomes (rates of healing and failure) varied widely following CPF treatment, which may be attributed at least in part to differing study designs, patient populations, and heterogeneous definitions for each outcome. For example, healing rates (fistula healing/closure) for biologics ranged from 8.8%[47] to 68.4%[88]. The study reporting the lowest healing rates[47] defined fistula closure as having both fistula closure and cessation of drainage, whereas the study reporting the highest healing rates[94] only required cessation of drainage to record a healing outcome, which may contribute to the higher healing rate. This example also highlighted a more general lack of alignment across studies regarding what constitutes treatment success or failure. These differences outline the need for standardized clinical outcome definitions for CPF to permit cross-study comparability and reduce variability in reported outcomes.

As with other outcomes of interest, morbidity rates were varied. Several elements may have contributed to this variation, one of which may have been the variation in the data sources used in each study. For example, the lowest and highest rates of FI at study enrollment were 3.4%[67] and 72.8%[71], respectively; these rates were derived from studies that utilized different data sources (claims database[67] vs web questionnaire[71]), which may explain the differences observed. Disease characteristics within a patient population may also explain the variability in reported rates of some morbidities, particularly abscesses. For example, at baseline the proportion of patients with abscesses ranged from 22.5%[72] to 53.0%[65]; however, the proportion of patients with complex CPF was not equal between studies (65.2%[72] compared with 90.6%[65]). Because complex CPF is often associated with abscesses, this difference in disease characteristics is a likely explanation for the difference in reported abscess rates.

For anorectal stricture/stenosis, the impact of complex CPF was unclear. For example, at baseline the two studies reporting the highest proportions of patients with stenosis (30.8%[65] and 22.0%[11]) had high proportions of patients with complex CPF (90.6%[65] and 100%[11], respectively). However, another study with a high proportion of patients with complex CPF (72.5%) reported that only 4.3% of these patients had stenosis as a comorbidity. The impact of complex CPF was also unclear with respect to the rate of proctitis. For example, rates of proctitis at baseline ranged from 25.6%[65] to 58.5%[58]. However, the study with the highest rate of proctitis had fewer patients with complex CPF (62.3%[58]) compared with the study with the lowest rate of proctitis (90.6%[65]). A similar observation was made when comparing rates of proctitis at the point of surgery[87,95]. These observations suggest that rates of proctitis are not associated with complex CPF. Overall, mortality data were lacking as only five studies[23,24,53,54,74] reported this outcome, suggesting that mortality is infrequently associated with CPF.

In this study, we were unable to identify consistently reported risk and protective factors associated with CPF development, treatment failure, or having to undergo multiple surgeries for CPF. In each of these categories, conflicting data were reported, which made interpretation challenging; for example, female sex, antibiotics, anti-TNF use and corticosteroid use were identified as both risk and protective factors associated with CPF development. Additionally, in some cases, associated risk and/or protective factors were only reported by a single study, which prevented generalization of such observations. Variability in study designs, definition of complex CPF, patient populations, and outcome definitions (e.g., healing and failure) no doubt contributed to an inability to identify consistently associated risk or protective factors. Data regarding risk factors are vital for informing appropriate, tailored treatment plans that will ultimately lead to optimal outcomes for patients, and the inability to identify these represents a large evidence gap.

In a broader context, the study findings are consistent with those reported previously. A recent SLR of the burden and outcomes of complex CPF identified large variations in treatment response (failure and relapse rates) in patients with CPF as well as varied rates of CPF prevalence in patients with CD[8]. In addition, a recent meta-analysis of the epidemiology and natural history of perianal CD reported significant heterogeneity for multiple outcomes, including disease prevalence. The heterogeneity was attributed to potential differences in patient populations, definitions of perianal disease, and study time frames; inconsistent reporting of risk factors associated with perianal CD was also highlighted[96]. Taken together with the findings of the present study, it is clear that challenges arising from a lack of standardized study designs and definitions for CPF and CPF-related treatment outcomes are ongoing, and the lack of consistency represents a significant unmet need within this therapy area.

The results of this SLR should be interpreted with consideration of several strengths and limitations. This SLR provides a broad and exhaustive overview of recently published evidence regarding epidemiology, treatment patterns, risk factors, morbidity, and mortality in CPF; the lack of restrictions in the inclusion criteria by participant age or comorbidities/concomitant diseases reduced the risk of missing relevant studies. This SLR was also conducted in accordance with published guidelines[16], including the use of two independent reviewers. All included articles underwent a quality assessment as guided by the National Institutes of Health Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. The inclusion of a large number of studies permitted analysis of CPF epidemiology and interventions across different age groups and geographies.

Several limitations should also be considered. Because most studies did not define complex CPF or the proportion of patients in a study with complex CPF, the results of this SLR primarily relate to CPF in general. In addition, direct comparison of CPF epidemiology, treatment patterns, and outcomes between studies was not possible. In addition, subgroup analysis for pediatric studies were not always possible due to the limited number of pediatric-only studies identified. Many studies were only available as conference abstracts and had limited reporting of study population characteristics, including inclusion criteria, which also limited comparability across studies. Finally, because included studies were observational in nature there was a potential risk of bias in these findings.

CONCLUSION

This SLR provides a comprehensive review of the current CPF epidemiological, treatment pattern, morbidity, and mortality data alongside risk and protective factors associated with CPF development, treatment failure, and the need for multiple CPF-related surgeries. Overall, data reported in this study were highly variable. Multiple factors such as heterogeneous outcome definitions, CPF definitions, and study designs (e.g., varying time points) likely influenced this variability. This highlights a need for consistency in the design of future studies as well as definitions of complex CPF and CPF-related clinical outcomes. Standardizing these approaches will facilitate the identification of optimal treatment strategies in the real-world clinical setting and improve outcomes for patients with CPF.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the contribution of Jenny Noack (Research Associate, Cytel Inc, Wismar, Germany) for providing support during the screening and data extraction phase of this SLR.