Introduction

The treatment of anal fissure was varied and chaotic until 1951 when Eisenhammer proposed using partial lateral internal sphincterotomy (LIS). He also combined LIS with a rather liberal dilation of the anal canal after the sphincterotomy. He was the first to list the number of patients treated by this method and reported that none had any defecation difficulties afterward [1]. This procedure was enthusiastically adopted by surgeons around the world. It was widely believed that incontinence was not an issue after sphincter division [2]. An early study to quantify continence disturbance published in 1985 stated that of 306 patients who had undergone LIS at least 1 year earlier, only 15 suffered from any degree of incontinence and this was principally only incontinence to flatus. In none was it severe enough for the patient to wear a pad [3].

However, in 1989 everything changed. Khubchandani published a large case series of follow-up after LIS in which 36% of the patients were incontinent to flatus and 5% to solid stool [4]. In 1996, a group from the University of Minnesota, which had reported the low incontinence rate in 1985 [3], in a retrospective comparison of open versus closed LIS, found that 30.3% of their patients were incontinent to flatus and 11.8% to solid stool [5]. The age of glyceryl trinitrate (GTN) ointment, botulinum toxin (Botox) injection and calcium channel blockers (CCBs) was born. It appears that in many countries, LIS had been abandoned in favor of medical therapy [6]. In one systematic review of anal incontinence following LIS, 22 studies, mostly non-randomized case series or cohorts, found an overall incontinence rate of 14% with less than 1% having incontinence to solid stool [7]. Yet patient satisfaction with LIS has been reported to be high [8]. The often crippling pain of fissure is almost immediately relieved by LIS.

Nevertheless, the fear of incontinence has resulted in a rapid expansion in the number of treatment options for anal fissure. The goal was to find a medicine or surgical procedure that simulated the high success rate of LIS and avoided the presumed high risk of postoperative incontinence. We performed a systematic review of all the published treatment options for anal fissure that have been subjected to randomized clinical trials only. The review was divided structurally into two halves. The first is comparison of surgical procedures only. The second is a comparison of non-surgical (usually pharmacological) treatments to either best supportive care, to other non-operative treatments or to a surgical procedure, which in all cases was LIS.

This type of review and meta-analysis is needed because individual reports of surgical procedures and non-medical therapies are variable in their results and under-powered. In addition, this approach allows assessment of the risk of bias in each publication as well as a combined assessment of the strength of the evidence for each individual intervention.

Materials and methods

Surgical procedures

Trials in which participants were randomized to a surgical procedure and either no treatment or an alternative surgical procedure were eligible for inclusion in this part of the review. Studies that compared any surgical procedure to any non-surgical procedure were not included in this section, but in a separately searched and analyzed group described below. Cluster- and group-randomized trials were also eligible but were not found.

Participants eligible for this portion of the review were patients with chronic anal fissure. Chronic anal fissure is typically described as an anal fissure which lasts more than 4–6 weeks, or which has characteristic features such as a sentinel pile, bare internal sphincter, heaped up edges or hypertrophied anal papillae. As it is common practice among surgeons reporting this disease not to operate on acute fissures, or fissures in children, or atypical fissures (multiple, irregular, off the midline or not associated with sphincter spasm), these were not eligible for inclusion in this section of the review.

Non-operative therapy

Studies in which participants were randomized to non-surgical treatment for anal fissure are the focus of this part of the review. Comparison groups in each of these studies may include a surgical procedure, medical therapy, or a control group consisting of no treatment, supportive care or placebo. Supportive care may consist of dietary fiber, laxatives or warm baths, lubricants, and even topical anesthetics, applied sometimes equally to both groups and sometimes only to the control group. Similar to the above group, cluster- and group-randomized trials were also eligible but were not found. Acute fissure will be included in this part of the review (see “Discussion” section).

Outcome measures

The main outcome measures were the following: fissure non-healing

  • Fissure recurrence.

  • Anal incontinence incidence—assessed mostly but not exclusively in surgical studies. In most cases, it was specified that it was minor incontinence, to flatus or anal seepage.

  • Headache—assessed mostly but not exclusively in GTN and isosorbide mononitrate (ISMN) studies.

The primary outcome analyzed and reported in this entire review is fissure treatment failure, which is a combination of primary non-healing and recurrence after apparent healing. This is the inverse of sustained fissure healing. This combination of the two outcomes often reported as stated above was done because due to the waxing–waning nature of fissure, it was problematic to differentiate persistence from recurrence (see “Discussion” section).

Adverse events, also primary outcomes, were principally anal incontinence and headache.

A secondary outcome found is the anal incontinence score which was only rarely reported.

Other outcomes reported were pain, bleeding, infection, but all of these were only reported sporadically, in varying metrics, and so are not analyzed in this review.

Literature search

A literature search (Fig. 1) was conducted to identify all published and unpublished randomized controlled trials with no language restriction, using the following electronic databases to identify potential studies including:

Fig. 1
figure 1

Search strategy

Full size image

  • The Cochrane Central Register of Controlled Trials (Issue 3, 2017).

  • Ovid Medline (1950 to January 18, 2017).

  • EMBASE (January 17, 2017).

  • ClinicalTrials.gov and the World Health Organization’s Internet clinical trial portal (ICTRP) were searched to March 7, 2017.

Study selection and review

Screening of each title and abstract and full text, data abstraction, data entry and risk of bias assessments were all conducted by at least 3 reviewers. All differences were resolved by the whole group discussion. The risk of bias of each study was assessed against key criteria: random sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, incomplete outcome data, selective outcome reporting [9]. Another most significant source of bias in this review was duration of follow-up after completing therapy (see “Discussion” section). The following judgments were used: low risk, high risk and unclear (either lack of information of uncertainty over the potential for bias).

Authors were contacted for missing data. When data on non-healing were unavailable, such as drop outs or losses to follow-up, the missing data, if they could be assigned to a treatment group, were treated as treatment failures, the last observation brought forward [10].

Clinical heterogeneity was sought in the performance of the surgery, or administration of the treatment, the veracity of the diagnosis and the accuracy and timing of outcome assessment. That is, how comparable were data from the included studies for meta-analysis. Methodological heterogeneity was sought in the differential risks of bias between studies. Statistical heterogeneity was calculated in Revman [9]. It was defined as a Chi-square (p < 0.10) and I-square (I 2) (>60%).

Subgroup analyses were done to investigate sources of heterogeneity in meta-analyses. Sensitivity analyses were done after eliminating the studies found to be of poor quality to assess the robustness of the results of the meta-analysis.

There are three nitrous oxide donors that have been tested for their ability to heal anal fissure: glyceryl trinitrate or nitroglycerin (GTN), isosorbide dinitrate (ISDN) and isosorbide mononitrate (ISMN). These three have been combined in these analyses. Their mechanisms of action are the same, and in one small trial, they were found to have a similar effect on healing fissure (#24 in Table 2). Similarly nifedipine and diltiazem have been combined in the review as CCBs (calcium channel blockers).

Data were analyzed using Revman 5.3. Results were expressed as odds ratios (OR) and 95% confidence intervals(CI) for the dichotomous outcomes, using the random effects model because of the heterogeneity seen in most of the larger comparisons.

The Grades of Recommendation, Assessment, Development and Evaluation (GRADE) [11, 12] approach was used to classify the quality of evidence for each intervention of the three primary outcomes (treatment failure, incidence of incontinence and other adverse events, principally headache) into one of the four grades:

  1. 1.

    High: Further research is very unlikely to change our confidence in the estimate of effect;

  2. 2.

    Moderate: Further research is likely to have an impact on our confidence in the estimate of effect and may change the estimate;

  3. 3.

    Low: Further research is very likely to have an important impact on our confidence on the estimate of effect and is likely to change the estimate;

  4. 4.

    Very low: Any estimate of effect is very uncertain.

Each intervention began with a high quality of evidence and was downgraded either one or two points to moderate, low or very low depending on whether any of the five factors listed below were present and how seriously (one or two steps down) the factor impacted the data.

  1. 1.

    Risk of bias as described above;

  2. 2.

    Inconsistency (unexplained heterogeneity and thus inconsistency of results) [13];

  3. 3.

    Indirectness (indirect or atypical populations, interventions, controls, outcomes. For instance, if a surrogate outcome was measured in some studies, such as anal pain only, without examining for fissure healing);

  4. 4.

    Imprecision, as made evident by wide confidence intervals, small study size, two or fewer studies and/or fewer than 100 events in that comparison. Random error in that case is too great [14].

  5. 5.

    Publication bias: Evidence either graphically in funnel plots or from review of trials registries that many potentially included studies have not been published.

Absolute risk reduction was measured using GRADEPro software [12].

Results

Search results

In the surgical review, 931 abstracts were found; 499 remained after duplicates were removed. From these, 462 studies were excluded and 37 reviewed in full text; 31 were chosen for inclusion and reported 2606 patients with fissure [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45]. Results of the search for this update are shown in the PRISMA diagram (Fig. 2).

Fig. 2
figure 2

PRISMA diagram for surgical studies

Full size image

In the non-surgical portion of the review, 1221 abstracts were found and 663 remained after duplicates were removed. From these, 494 studies were excluded and 170 reviewed in full text, 117 being chosen for inclusion with 9456 participants [46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162]. The study by Oueidat [125] was excluded as the abstract was never published in full text. Results for the search for this update are shown in the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) diagram (Fig. 3).

Fig. 3
figure 3

PRISMA diagram for non-surgical studies

Full size image

Fifty-nine publications were read in full text and excluded from the review. The reasons for exclusion mostly fell into three categories:

  • They were non-randomized trials.

  • They were abstracts only from meeting presentations. There were therefore insufficient data in these for this review, and in most cases they were years old, so it is unlikely that a full publication will ever occur.

  • Fissure healing was not an endpoint of the study.

Interventions

The interventions described in the included studies of the surgical review are listed in Table 1, and for the non-surgical review in Table 2. Along with the interventions are:

Table 1 Surgical studies review
Full size table
Table 2 Non-surgical treatment reviews
Full size table

  • The comparator for each test intervention,

  • The number of included randomized studies for that intervention,

  • The number of patients in each treatment group,

  • The number of events found in each treatment group, the event being the inverse of sustained fissure healing, i.e., failure of fissure therapy, or incontinence or headache in the case of adverse events,

  • The OR and CI obtained from a meta-analysis of the studies included in each intervention,

  • The GRADE classification of the strength of the evidence for that intervention and the reason for each GRADE.

Risk of bias

Table 3 shows the prevalence of the components of the risk of bias in the whole segments of the review, surgical and non-surgical, as well as several of the most important interventions examined in this review. In addition, statistical heterogeneity is displayed for each of the specific interventions, both the Chi-square and I-square.

Table 3 Risk of bias
Full size table

The sensitivity analysis of manual anal stretch versus LIS, eliminating 2 studies with significant quality issues, is shown to eliminate the statistical heterogeneity of the whole group (Table 1, #2). A subgroup analysis of LIS versus any non-surgical therapy was done to investigate the heterogeneity found in the whole group of 32 studies, choosing only those studies with more than 6 months of follow-up. Once again this analysis resolved the heterogeneity (Table 2 #10).

Numerous subgroup analyses were done to investigate the extreme heterogeneity in GTN versus control including elimination of all but the 4 largest studies, which had more than 100 patients each. These included 896 patients and 498 events (non-healing). In addition, an influence analysis was done eliminating one study at a time to see whether any single study altered the heterogeneity. In all cases, the Chi-square remained <0.00001 and I-square > 75%. The remarkable feature of this intervention was that 17 of 19 (89%) of the studies had inadequate follow-up. This introduces a major bias because it fails to detect persistence in a disease known to wax and wane, and recurrence (which may in fact be the same thing). GTN gave a sustained cure in 47.5% of the patients, and the control group had sustained cure in 38.6%. In this comparison as well as many others in the non-surgical review, a “control” group is variously described as best supportive care, involving laxatives, baths and sometimes topical anesthetics such as lidocaine (see Table 2 comparison 2) or occasionally a true placebo. These were combined in the analyses as a control group.

Grades of recommendation, assessment, development and evaluation (GRADE)

In the majority of interventions, the evidence is rated low to very low because the studies included in the interventions are under-powered and have significant quality issues related to selection bias, blinding and length of follow-up. Only 1 was rated as high, the subgroup analysis of LIS versus any medical therapy (Table 2 #10). The only weakness in this group was a high attrition rate, which would be expected with the longer follow-up. It is offset because one can upgrade GRADE if the effect is extreme and the information size otherwise adequate. It is for both in the case here (Table 2 #10). The medical therapies included in Table 2 #s 9 and 10 were, in all but 2 cases, either GTN, CCBs or Botox. The other two were arginine (2#14) and posterior tibial nerve stimulation (2#31). Publication bias is difficult to ascertain and quantify. Funnel plots were done in this review when there were at least 10 studies in a specific intervention. However, they are open to wide interpretation. A more interesting means of assessing publication bias is, in this age of trials registries, the World Health Organization Internet Clinical Trials Registry Portal (ICTRP). Many trials were found there which were not apparent in publication search. Many investigated substances that were previously unheard of. No cluster was found that would substantially change the GRADE rating of any of these interventions in this review.

Effects of interventions

The effects of interventions are varied. They are presented in Tables 1 and 2. The most important include:

Table 1, #2

LIS is superior to manual anal stretch

Table 1, #3

Open and closed LIS are equally efficacious

Table 1, #4 7 #5

Length of LIS is important, but the quality of the evidence makes it difficult to determine which is best

Table 1, #8

Posterior internal sphincterotomy seems no worse than LIS in healing, though nonsignificantly worse than LIS with incontinence (not shown in the table; OR 4.46, 95% CI 0.47–42)

Table 2, #1, Fig. 4

GTN is superior to control, but with very poor quality of evidence (see below)

Table 2, #2

GTN is superior to lidocaine

Table 2, #5

GTN is roughly equivalent to Botox

Table 2, #6

CCBs are superior to GTN

Table 2, #10

LIS is far superior in healing to medical therapies, but LIS is associated with an increased risk of minor incontinence (see “Discussion” section)

Many of these interventions about which only one or two studies have been done have yielded interesting results that should be pursued in further investigations. These are listed in “Discussion” section.

Adverse events

Adverse event rates for the two principal adverse events related to anal fissure therapy are shown in Table 4. Many other adverse events are reported in the 148 included trials, but they were of low frequency and the various ascertainment methods used made determination of the compilation and rate impossible. Headache dominated nitrous oxide donor reports and so was also reported in the comparators to these interventions. Incontinence was reported in all surgical trials and so in the comparators to surgical procedures. Incontinence, in almost all cases, was described as minor, which was specified to mean in many studies incontinence to flatus and anal seepage. This often resolved over a year, implying that the symptoms were caused by a healing anal wound.

Table 4 Adverse events
Full size table

Discussion

There was only 1 comparison for which the quality of evidence was high, indicating that future research is unlikely to alter this finding, the subgroup analysis of LIS versus any medical therapy with at least 6-month follow-up (Table 2 #10). Figure 4 shows the forest plot for the entire group of 29 studies of LIS versus medical therapy, the last colored column showing which studies had adequate follow-up data. Figure 4 shows forest plots of the subgroup of studies with more than 6 months of follow-up, Fig. 4 being non-healing, with the remarkable OR of 0.12. Figure 4 shows that in these studies, LIS is still more likely than medical therapy to result in minor incontinence. This comparison (Table 2 #10) also highlights the weakness of comparison seen in Table 2 #1: GTN versus control. The predominance of short follow-up (2#1) is clearly the source of the extreme heterogeneity that causes the quality of the evidence in this important comparison to be very low despite the large number of included studies. Very high recurrence rates have been reported in patients whose fissures were initially healed by GTN, if they had 1 year of follow-up: 51 [163] and 67% [164]. In many of these cases, healing may have just been the usual behavior of the anal fissure rather than healing due to GTN.

Fig. 4
figure 4figure 4

Forest plots for key findings. The risk of bias for each study was low (green), high (red) or yellow (unclear—which often meant unstated in the text) for each of the facets of risk of bias listed in the figures

Full size image

The GRADE ranking of GTN versus control (Table 2 #1) is quite problematic. The number of studies and participants is large, and, using GRADE’s own rating, it seems unlikely that further research will change the relations seen in Fig. 4. And so upgrading it a bit, at least to poor would seem justified. Better studies with adequate follow-up would solve this problem.

At the other end of the spectrum, there are studies, many of which have original innovative ideas, but for which only a single study exists, usually with few patients. Random error cripples the impact of these studies, in spite of statistical significance, and they yield unreliable results [14]. Therefore, comparisons in all of Table 1 and in Table 2, #7 and #s 11–27, 30–43, all were downgraded in GRADE for serious or very serious imprecision.

The ones with the more interesting results were:

  • More controlled anal dilation than the older manual anal stretch: Table 1, #s 11, 12; Table 2 #4.

    GTN patch versus GTN ointment: Table 2 #36, i.e., a patch as for angina on the chest or shoulder.

  • Clove oil versus lidocaine: Table 2, #15.

  • Botox anterior versus Botox posterior: Table 2, #26.

Several comparisons yielded moderate quality of evidence and so are more reliable, but future research may alter the summary effect:

  • Open versus closed LIS: Table 1, #3.

    GTN versus lidocaine: Table 2, #2.

  • High- versus low-dose GTN: Table 2, #3.

  • GTN versus patient self-dilation at home: Table 2, #4.

  • GTN versus Botox: Table 2, #5.

  • GTN for 40 or 80 days: Table 2, #28.

  • Oligo-antigenic diet versus control diet (again): Table 2, #29.

GTN versus lidocaine is significant because prior to the introduction of GTN, CCBs and Botox, medical therapy consisted essentially of lidocaine, hydrocortisone (Table 2, #30) and bran (Table 2, #43) [101, 102]. With the passage of time and the introduction of newer medications, lidocaine, bran and hydrocortisone became only best supportive care as stated above and were thought to be ineffective in obtaining a sustained cure of anal fissure.

Other systematic reviews

The other global reviews of fissure therapy were published last in the Cochrane Library in 2011 and 2012 [165, 166]. These have now been updated here with the addition of 47 new randomized trials. A series of systematic reviews examined GTN, Botox, CCBs and LIS [167,168,169,170,171] individually. They were limited to English language publications and also separated primary healing from recurrence.

Acute anal fissure

This is described as a specific clinical entity. There are certainly anatomical components of chronic fissure that make it easy to differentiate from acute fissure. But another definition that separates acute from chronic fissure is duration of disease and that is where matters gets difficult. The definition varies greatly in published reports from as short as 2–4 weeks to as long as 3–6 months. Occasionally short episodes of symptoms widely separated in time may also diagnose chronicity. Patients often cross the border from acute to chronic while getting appointments or initiating therapy. Many acute fissures resolve spontaneously (or may come back later as chronic fissures). So do chronic fissures. It is generally believed that children do not get chronic fissures nor should children or adults with acute fissure ever have surgery. But there are reports in this review where exactly that happened [87]. This is the rationale for the integration of acute and chronic fissure in this review. It seemed too difficult to separate them. When interventions were limited to patients with acute fissure, it is stated in the tables.

Incontinence

Reconciling the quite shocking data concerning incontinence risk with LIS and the description in those publications of LIS-induced acquired incontinence as permanent with the data presented here is not easy. All the data in this review were obtained from prospective randomized trials, all approved by ethics committees, which are very sensitive to potential harms in clinical trials and recording of them, and the trials themselves, especially trials with an LIS component, had focused on incontinence risk. The term permanent incontinence is particularly unfortunate, since all anal incontinence is eminently treatable [172,173,174]. Retrospective review numbers are certainly subject to selection bias, but even that does not explain the disparity with what is described here and those publications. In the first surgical review, published in 2000, the risk of incontinence was 10%. Within the current surgical review, when LIS was compared to other operations, for publications after 2000, the risk has dropped to 3.4%. This may be due to the advent of effective medical therapy, and better patient selection for surgery, or more care being taken in surgery to avoid excessive sphincter injury. It should be noted that the risk of post-therapy incontinence related to LIS in this review is not radically different from what is seen after therapy with GTN, Botox or CCBs (Table 4).

The purpose of this review is not to establish guidelines for the treatment of anal fissure. There are numerous bodies that have that responsibility [174,175,176,177,178]. It is instead to give as detailed and unbiased as possible a summary of the evidence for all studied treatments for anal fissure in order to facilitate the creation of guidelines and to assist patients and doctors who need to understand the risks and benefits of fissure therapy.

Conclusions

LIS is the most effective treatment for anal fissure, curing all but 6% of patients. Late recurrences are very rare after LIS versus with medical therapy [179]. Minor incontinence is more likely with LIS than medical therapy (Fig. 4). The difference between LIS and medical therapy is significant, but the absolute risk alteration is small, increasing from 3 cases per 1000 patients with medical therapy to 14 cases per 1000 with LIS (95% CI 6–31). Open and closed (a euphemism meaning less open) LIS are equally effective. Manual anal dilation is inferior to LIS, but recent small studies suggest that more controlled dilation, either pneumatic, by speculum or by patients at home are just as effective as LIS and are not associated with any risk of incontinence. GTN, Botox and CCBs have been extensively investigated as treatments for acute and chronic anal fissure. They appear to be effective, but most studies have been marred by inadequate follow-up, thus missing late recurrences, which are common. Of the three, CCBs may be the most effective. More research is needed for all three with adequate follow-up. There is virtually no research on sequencing these drugs, i.e., if one fails, what is best to try next? Many other medications have been tried because of the less-than-perfect track record of the three above, but none have proven better, and in any case, the studies are too small and too few. Clove oil and sildenafil may be worth further investigation. Many unpublished studies can be found in this field, especially in ICTRP, but no results are available.