Cohort study

Methods

Aims

A prospective, multicentre observational cohort study was conducted to:

• describe the disease characteristics of participants undergoing treatment for PSD
• describe procedures currently in use and quantify clinical outcomes and PROs associated with each
• identify patient-specific disease features that might predict poor outcome in each treatment group
• derive a case mix-adjusted estimate of the risks associated with common treatment options.

Participants

Eligible patients were undergoing definitive elective treatment for PSD at study recruiting centres.

Inclusion criteria

• Consenting patients aged 16 years or older and with PSD.

Exclusion criteria

• Asymptomatic disease.
• Currently pregnant.
• Unable to give consent.
• Acute abscess.
• Hypersensitivity to the sealants.

Study procedures

Recruitment and consent

Patients considered suitable for surgery were identified from general practitioner (GP) secondary care referrals, surgery waiting lists or clinics. Once identified, participants were given an approved participant information sheet detailing the study – sent in the post or provided in person. Patients were invited to attend a recruiting clinic. At the clinic, a member of the research team explained the study to the participant and offered them an opportunity for them to ask any questions. The principal investigator (PI) or delegated research team member confirmed eligibility and ensured written informed consent was obtained prior to any patient data being collected. Participants were advised that they were able to withdraw from the study at any point without any impact on their routine NHS care. As is standard practice, the surgeon would discuss the condition, possible interventions and their advantages and disadvantages. Patients were given a minimum of 24 hours between receiving the participant information sheet and consenting to the study.

In response to the COVID-19 pandemic, adjustments were made to allow the continuation of recruitment and consent procedures. Consent could be obtained by post. An invitation letter, participant information sheet and a postal consent form were posted to the patient. The research team were able to contact the patient to provide an overview of the study and answer questions. The patient was instructed to complete two consent forms: one to be returned to the research team, one for their own records. Once received, the research team contacted the patient to complete the postal consent review form.

Intervention

The study was observational, and surgeons were not asked to change their usual practice. Surveys suggested that around six procedures were in common use,28,31–33 which can broadly be described as:

• major excision with asymmetric closure (‘Bascom II cleft lift’)
• major excision with lateral closure (‘Karydakis’)
• major excision with lateral closure with rhombic flap
• major excision with midline closure
• major excision and leave open
• minimal excision (‘Bascom I’ or ‘pit picking’).

Other approaches include curettage (‘scraping out’) or phenol injection with glue closure. For the purpose of analysis, procedures were classified as either ‘minor’ or ‘major’ procedures, with the latter further subdivided as ‘asymmetric/lateral closure’, ‘midline closure’ or ‘leave open’.

Data collection

Data were collected by trained research personnel. All patient data were recorded on the case record form. Copies of the consent and patient information sheets were kept in the participant’s hospital case notes. A copy of the consent form was uploaded onto the REDCap data capture system for monitoring purposes. All data were recorded on the REDCap data capture system.

Assessment schedule

Participants completed baseline questionnaires after eligibility and consent were confirmed. Details of the procedure were collected on the day of procedure. Outcome data were collected on days 1 and 7 after procedure and then at an in-person clinic visit and a further follow-up 6 months after the procedure. Participant data could also be collected opportunistically at a final ‘study completion’ visit. The outcomes collected are listed in Recruitment and participant flow.

Safety assessments

Participants were asked to report complications at days 1 and 7 post procedure, and again at the clinic visit and 6 months. Participants were prompted specifically for incidence of bleeding, seroma, haematoma, infection, dehiscence, maceration, flap necrosis or discharge. Other adverse events (AEs) were collected only if considered related to the study treatment.

Statistical methods

Sample size

The study aimed to recruit approximately 800 patients, with at least 100 within each of the front-running management strategies. Doing so allows proportions to be estimated within each management strategy to a standard error of ≤ 5% and pain numeric rating scale to within a standard error of 0.2 points, assuming that the standard deviation (SD) of a 10-point scale would not exceed two units.

Outcomes

The outcomes and their timing and description are listed in Table 4. All outcomes were self-reported aside from the clinician-assessed scarring question. No single primary outcome was prespecified in this study; methods to elicit outcomes of most importance to study participants are described in subsequent chapters.

Comparisons

The sample size precluded reliable comparisons between specific subtypes, with only one procedure (Karydakis) providing at least 100 participants with non-missing outcome data. Instead, we undertook the following risk-adjusted treatment comparisons on broader categories of procedure types:

• any major excisional procedure versus any minor procedure
• any major excisional procedure with asymmetric closure versus any minor procedure (minimal excision).

Outcomes were summarised descriptively in relation to the treatment received for the less broad treatment groupings whose number did not permit risk-adjusted modelling.

Methods for obtaining risk-adjusted comparisons

Procedures were compared using risk-adjusted methods to reduce bias due to treatment selection, since the extent of disease is likely associated with both the type of procedure and the response.

Statistical methods for risk-adjusted outcomes

Three broad approaches were taken to risk-adjust these comparisons.

1. Regression modelling.

Risk adjustment was undertaken separately to attempt to adjust for imbalance in prognostic features across the procedure groups. Each outcome was modelled separately since features do not affect all outcomes equally. For each outcome, three models were fitted:

1. all features
2. features associated with the outcome
3. the Wysocki disease classification alone.

Model 2 is a ‘compromise’ model which trades off missing potentially important features against model parsimony (overfitting) and the impact of missing covariate data. Covariates were removed on the basis of Akaike’s information criteria (AIC) and the size of the c-statistic of the model. All models were discussed and agreed with a core study clinical team prior to revealing comparative data.

Continuous outcomes were modelled using linear regression, and differences with 95% confidence intervals (CIs) between treatment groups were estimated from the regression coefficient for the procedure group. Binary outcomes were modelled using logistic regression and absolute differences in proportions were assessed using the difference in marginal probabilities. Time to wound healing and time to return to normal activities were modelled using either Cox regression or parametric accelerated survival time, the choice of which depended on which fitted best to the distribution. Proportional hazards were assessed using scaled Schoenfeld residuals and the Grambsch–Therneau test, and the fit of parametric survival distributions was assessed using Q-Q plots.34 The parametric model was chosen as the lowest AIC among four different approaches (Weibull, log-normal, log-logistic and generalised gamma).

1. Propensity score approaches.

The second approach used a different approach which attempts to balance treatment groups based on features affecting the choice of procedure they received, rather than features associated with outcome. Two approaches were taken.

1. inverse probability weighting (IPW)
2. nearest neighbour matching.

Features were assessed using logistic regression in which treatment choice was the outcome. Covariates were identified analogously to model 1(ii) above. The same propensity score adjustments were used for each outcome. The propensity score-adjusted models were then used to calculate predicted outcomes in both arms, following which their difference and 95% CI were estimated.

Linear and logistic regression models were used for continuous and binary outcomes, respectively. Time-to-event outcomes were fitted within the propensity score framework only if the assumption of accelerated failure time distributions was met based as outlined above.

1. Augmented IPW/IPW with regression adjustment.

The final approach is a combination of approaches 1 and 2 which simultaneously models both treatment selection and outcome using the same covariates used in 1(ii) and 2. The differences in predicted outcomes and their 95% Cis were estimated for binary or continuous outcomes. Time-to-event outcomes (wound healing and return to normal activities) were compared using IPW with regression adjustment. In this, treatment selection was balanced using IPW, and the outcome was modelled adjusting for covariates in 1(ii). The AIC was used to select the best-fitting distribution (Weibull, log-normal, log-logistic or generalised gamma).

Factors affecting outcome and choice of treatment

Previous publications have suggested several possible risk factors which may affect outcomes. The following factors were considered as potential risk factors for poor outcome:

• Demographic features: sex, body mass index (BMI), depth of natal cleft, presence and type of gluteal hair (none, mild, dense) and smoking status.
• Disease characteristics: pit density (number of pits divided by spread of pits), presence of unilateral or bilateral disease, distance from furthest lateral opening to the nearest pit, presence of pus and Wysocki disease classification.

The same features were also assessed for their association with choice of procedure.

Statistical assumptions

All modelling approaches make assumptions (some of which are untestable), and no single method is clearly optimal. Approaches 1 (regression) and 2 (propensity score) are unbiased only if the models incorporate all relevant features and are correctly specified. Approach 3 is termed a ‘doubly robust’ method and is unbiased if either of the two-component models is correctly specified, but is more complex and more susceptible to overfitting. In view of this, no single method was identified as the primary risk adjustment. Instead, the findings from models were assessed for their consistency and, where they provided conflicting estimates, the plausibility of each model was considered.

A preliminary assessment of overlap was undertaken prior to any modelling in order to ensure that different treatments had non-zero probability of uptake in different subgroups.35

All participants undergoing a procedure were included in the analyses.

Patient and public involvement

Two patient and public involvement (PPI) representatives joined the study. The PPI representatives reviewed all patient-facing documents to ensure readability, understanding and format. One PPI representative sat on the steering committee panel and provided an instrumental patient voice in the management of the study. This PPI representative suggested the inclusion of a supplementary participant information sheet to be available for patients who were waiting for their clinic appointment. The patient representatives were also consulted when writing the plain English summaries and dissemination materials.

Results

Recruitment and participant flow

Thirty-one UK sites recruited participants over a 46-month period from May 2019 to March 2022 (see Appendix 2). Figure 1 shows the flow of participants through the cohort: in total, 729 participants consented to be part of the cohort study. Participants were excluded from analyses if they did not have a procedure during the study (n = 45), if they were ineligible due to an incorrect diagnosis (n = 7), or if there was not enough information provided in order to categorise their procedure (n = 10). A total of 667 participants were included in the analysis cohort, of whom 476 (71%) provided follow-up data at 6 months. The number of participants who consented was lower than our anticipated sample size of 800, with 100 patients in each of the front-running management strategies.

FIGURE 1 Participant flow for the PITSTOP study. PIS, patient information sheet. (Figure not reproduced here; see the original article.)

In response to COVID-19, participants were able to be followed up at the end of the study; this was completed for 574 participants and used to update complications and recurrence data.

Baseline characteristics

The characteristics of the cohort participants are included in Tables 5 and 6. There were more males (73%), and 85% of participants were white; the average age of participants was 29 years. Just over half (54%) of participants reported no previous procedures, and 22% reported a previous elective procedure for PSD.

Data completion

Data completion rates for the cohort outcomes are presented in Appendix 3, Table 25. Return to normal activities, recurrence and wound healing were considered complete if a patient contributed those data at either clinic visit, 6-month follow-up or study close (or in the case of recurrence, it was apparent from AE reporting). Complication data were considered complete if the participant contributed data to at least one follow-up time point. Data for recurrence (94%), complication (96%) and return to normal activities (94%) were collected for most participants. The characteristics of participants that attended 6-month follow-up were compared to those that did not attend 6-month follow-up (see Appendix 3, Table 37), and the distribution of characteristics was similar between the groups; there were marginally more participants that had lateral openings in the attenders (50%) compared to the non-attenders (44%). All analyses were conducted on available data.

Treatment decisions

The breakdown of treatments received is presented in Table 7. Recorded procedure details were categorised into four categories, which were further combined into major or minor procedure categories. Over half (60%) of the participants received a major treatment, most commonly asymmetric closure (41%). Of the participants that received minimal excision, the most common treatment options were glue (n = 106, 16% of the cohort) and pit picking (n = 60, 9% of the cohort).

Further treatment details are presented in Appendix 3, Table 26. Median length of surgery was 30 minutes, and 95% were performed as day cases. Procedures were typically performed by consultant surgeons (68%).

Figure 2 shows the treatment received by disease characteristics. Participants with recurrent disease (defined as reporting any previous procedure, including acute drainage) were more likely to be given asymmetric closure; participants that were not recurrent were more likely to receive minimal excision. Over half (56%) of Wysocki type 1 (only midline pit or sinuses) participants underwent minimal excision, whereas over half (53%) of Wysocki type 4 (disease after treatment with definitive intent) were given asymmetric closure. The extent of overlap is noteworthy; for all disease characteristic categories there were a number of participants that received each treatment type, suggesting there is variety in the types of procedures considered appropriate for patients with different disease characteristics. The distribution of the number of pits was similar across procedure types (see Appendix 3, Figure 13). The proportion of patients with a minor procedure varied substantially across the sites (see Appendix 3, Figure 19), although this may be due to differing case mix across centres.

FIGURE 2 Treatment choice by disease characteristic. Note: ‘recurrent’ defined as any reported previous procedure; type 1: only midline pit or sinuses; type 2: any midline disease with secondary sinus/es or abscess scar/s; type 3: any midline or secondary disease extending below tip of coccyx; type 4, any disease after treatment with definitive intent. (Figure not reproduced here; see the original article.)

Shared decision-making and decision regret

Participant ratings of their pre-op consultation were high (Table 8), with the median (IQR) of the CollaboRATE mean score response being 3 (3–4), where 3 represents ‘a lot of effort was made’. The CollaboRATE top score was given in 36% of cases, reflecting that ‘every effort’ was made to help the patient understand their health issue, listen to the things that matter most and include what matters most to the patient in choosing what to do next. The decision regret (DR) scale, completed at month 6 follow-up, was low (median 8, IQR 0–20), and was broadly similar across the treatment categories.

The relationship between CollaboRATE mean score and DR is shown in Appendix 3, Figure 14. There is little clear evidence of a correlation between shared decision-making (SDM) and DR, with the majority of patients being in the top left corner of the graph (representing participants that were happy with their collaboration and had few regrets about their procedure).

Decision regret was low among patients [mean (SD) 14.5 (16.7)] and was broadly similar across the procedure types. The majority of patients reported being either satisfied or very satisfied (83%) with their procedure. Seven (21%) of the participants that received a leave-open procedure reported being either dissatisfied or very dissatisfied. The majority of patients returned to normal activity by the end of follow-up (n = 550, 88%) and 75% reported the wound as having healed during the study follow-up. Almost half (47%) of participants experienced a complication during follow-up.

Outcome summaries

The continuous outcome measures, recorded over time, are presented in Appendix 3, Table 27. Self-reported pain related to pilonidal sinus was at its highest on average on day 1 after procedure compared to baseline and reduced to its lowest at the 6-month visit. The highest average pain was reported by participants that received asymmetric closure. Patient-reported EQ-5D-5L health utility reduced from baseline to day 7 (overall means 0.80 and 0.69, respectively) but had recovered at both clinic and 6-month visits (overall means 0.83 and 0.89, respectively). Participants that received minimal excision reported the least change at day 7 and the highest health utility and QoL satisfaction at clinic visit and 6 months, although the scores were more similar among treatment groups at 6 months.

Other outcome measures are presented in Appendix 3, Tables 28–30. Repacking procedures were reported by 68 participants (12%) by day 7, while 87 participants (17%) reported repacking at clinic visit. At day 7, 226 (39%) patients reported a re-dressing procedure; re-dressing by day 7 was most common in asymmetric closure (49%) and midline closure (52%). Nearly half of participants experienced a complication during follow-up (n = 301, 45%), the most common of which were infection (26%) and discharge (18%). The numbers of complications were broadly similar across the three major surgery groups, and were lower for patients who received minimal excision, particularly for bleeding, dehiscence and infection.

Risk-adjusted treatment comparisons

The primary comparison between treatments was made between major procedures (asymmetric closure, leave open, midline closure, n = 396) and minor procedures (minimal excision, n = 270). No factors were found to be collinear and so all were included in the risk adjustment. Non-linearity of continuous features (BMI, natal cleft depth, pit spread and pit distance) was investigated and all features were deemed to be sufficiently modelled using linear terms. For the propensity score modelling, sufficient overlap in risk score was observed for all outcomes, and thus risk-adjusted analysis was deemed appropriate for major versus minor procedures.

Pain

The propensity score model identified sex, presence of pus and Wysocki classification as the most important features in treatment choice. Patients were more likely to undergo major procedure if they were female, had pus, or were classified as Wysocki type 4 (disease after treatment with definitive intent) and least likely to have major procedure if type 1 (only midline pit or sinuses). These factors were used in the propensity-adjusted models for all outcome comparisons.

Pain on day 1, adjusted for factors predictive of treatment choice and outcome via augmented IPW, was higher for patients who received major procedures compared to minor procedures by 1.58 points (95% CI 1.14 to 2.01) (see Appendix 3, Table 31). This was very similar to the unadjusted difference (mean difference 1.62, 95% CI 1.23 to 2.02). The number of participants included in each analysis varied according to the factors included to adjust the models. A similar difference in pain at day 7 was observed (augmented IPW-adjusted mean difference 1.53, 95% CI 1.12 to 1.95). The difference in pain between treatment groups was similar regardless of the risk adjustment method. A post hoc sensitivity analysis that included baseline pain as a covariate yielded mean differences and 95% CIs that were within 0.1 points of these estimates.

Pain over time for major and minor procedures is shown in Figure 3 (pain at clinic visit and month 6 were not prespecified as outcomes with formal comparisons); the raw difference in means was closer at clinic visit than at day 7, and there was no difference in mean pain reported at 6 months between the procedure types.

FIGURE 3 Pain with major and minor procedures, baseline to 6 months after surgery. (Figure not reproduced here; see the original article.)

Complications

Just over half (54%) of participants receiving major procedures reported a complication, compared to 36% of participants who had a minor procedure. After augmented IPW risk adjustment, participants who received major procedures reported a 17.5% (95% CI 9.1 to 25.9%) higher absolute incidence of complications during follow-up (Table 9). The estimate of the difference was relatively consistent regardless of risk adjustment method.

Disease recurrence

Recurrence was less likely for major procedures compared to minor procedures (Table 10). Among participants who had a major procedure, 15% reported recurrence by 6 months after surgery, rising to 23% when the full follow-up period was included, compared to 27% and 34%, respectively, for participants who had minor procedures. The risk-adjusted absolute difference in 6-month recurrence was 10.1% (95% CI 2.1 to 18.1%) in favour of major procedure using augmented IPW and was of similar magnitude in other risk adjustments.

Treatment failures (defined as the composite of recurred, not returned to normal activity, or not healed by the time of the last follow-up) were compared between treatment groups. The proportion of participants for whom treatment failed was more similar across treatment groups once healing and return to normal activities were introduced. In total, 47% of minor procedures failed at 6 months compared to 45% of major procedures (adjusted difference 2.3%, 95% CI −6.2 to +10.9%).

Return to normal activities

Time to return to normal activity was compared between major and minor procedures (Table 11 and Figure 4). While nearly all participants had returned to normal activity by the end of follow-up, the time taken was far quicker among those undergoing minimal excision (median 7 days) than those who had major procedures (median 32 days). At 6 months, 12% of participants who had major procedures and 4% of participants who had minor procedures were yet to return to normal activity (see Figure 4). Several participants dropped out, providing either a censored time or no data at all; in the best-case scenario where these were assumed to have recovered, the proportion of participants who returned to normal activity at 6 months would be 96% for major procedures and 98% for minor procedures. Participants who received major procedures took on average 21 days longer to return to normal activity than those receiving minor procedures, and the difference was greater using risk adjustment models. The mean difference as estimated by augmented IPW was 25.9 days (95% CI 18.4 to 33.4 days), with regression adjustment approaches providing estimates closer to the unadjusted difference. Similar findings were present when comparing asymmetric closure procedures to the minimally invasive approaches (see Appendix 3, Figure 19 and Table 40).

FIGURE 4 Time to return to normal activities by major or minor procedure type. (Figure not reproduced here; see the original article.)

Wound healing

Participants having major procedures also took longer to heal than those who had a minor procedure. The median time to healing was 30 days among people undergoing minimal procedures, compared to 70 days among those undergoing a major procedure. However, as highlighted in Figure 5, around 25% of participants in both groups had wounds that had not healed. Some of the individuals lost to follow-up (LTFU) may have healed, but a best-case scenario where those censored prior to 6 months were assumed to have healed would still mean 10% of participants considered their wound unhealed at 6 months. Unadjusted and regression-based risk adjustments both estimated the difference in wound healing to be over a month greater following a major procedure, while propensity score methods estimated larger differences but with wider CIs (augmented IPW estimate 53.5 days, 95% CI 28.8 to 78.2 days; Table 12).

FIGURE 5 Time to wound healing by major or minor procedure type. (Figure not reproduced here; see the original article.)

Pairwise comparisons of asymmetric closure and minimal excision (removing participants that had the major procedures – leave open and midline closure) produced results in keeping with the comparison between major and minor procedures (see Appendix 3, Tables 37–43).

Surgeon variation

A post hoc analysis looked at recurrence rates among surgeons who operated on at least 10 participants. In total 13 surgeons undertook at least 10 procedures (range 10–55 procedures, median 14). In total 282/667 participants underwent procedure by one of these surgeons. While recurrence and treatment failure rates varied among the 13 surgeons, outcomes were more favourable among participants whose surgeons treated 10 or more cases (see Appendix 3, Table 32). Overall recurrence at 6 months was lower among participants treated by these surgeons (40%) compared with surgeons that treated fewer cases (60%), with similar differences seen for recurrence at any time (42 vs. 48%) and treatment failure (40 vs. 60%).

Adverse events

Adverse events and serious adverse events (SAEs) are presented in Table 13; 107 (16%) patients experienced at least one AE during follow-up, and the most common category of AE was wound infection (59%). Eleven participants experienced an SAE, nine of which were inpatient hospitalisation.

Discussion

The prospective cohort study was the main component of the PITSTOP study and consists of one of the largest data sets of real-world experience gathered on PSD to date. Although these are subject to the potential biases of non-randomised comparisons, the data suggest that there is more postoperative pain and failure of treatment after major excisional procedures compared to minimally invasive procedures, associated with an increased time to healing and longer time to return to normal activities. This is the case even after risk adjustment for patient demographics and severity of disease.

The demographics of this cohort are unsurprising: the disease tends to affect a young, predominantly male population who are overweight.1 Interestingly, despite descriptions of patients as having coarse hair,36 we found the majority were assessed as having fine or medium hair. Around half of the group have had previous surgery for PSD, usually acute drainage of an abscess. Around one in five patients have had more than two procedures, with a significant minority having had three or more procedures. The disease varies in severity from simple midline disease to around half having some form of lateral extension. Around 10% have complex disease (bilateral disease or disease below the coccyx), making surgical intervention challenging and potentially limiting options for treatment. Recurrent disease – which may also, but not necessarily, be considered complex – was reported in 15%.

Twelve different types of surgical approach were utilised. This is more than in a previous survey on UK practice,28 the increase being mainly due to the expanded repertoire of minimally invasive techniques including glue, laser and endoscopic treatment over the last 10 years. By far the commonest procedures were excision and asymmetric closure techniques (Karydakis and Bascom’s II), which is consistent with previous data. While considered outdated due to the risk of failure and protracted recovery time,15,37,38 roughly one in six procedures involved excision-and-leave-open of the skin defect or primary closure in the midline. Reasons for the persistence of these procedures in UK practice have been discussed in Chapter 2. It is feasible that complex situations – for example, advanced bilateral disease, markedly infected wounds or other unusual variants of disease – meant that no other procedure was possible. However, it seems unlikely that these uncommon variants account for all such procedures carried out.

Despite the plethora of minimally invasive procedures currently practised in the UK, only 40% of patients had this approach. Given that this was among a group of surgeons interested in pilonidal disease, this approach could well be even less in the context of all UK surgeons. Minimally invasive procedures are certainly not suitable for all patients. For those with extensive disease or complex recurrence, minimally invasive approaches may not be effective. However, it would appear logical that, for most patients with non-recurrent disease confined to the midline or with simple lateral extensions, such a technique would have been feasible. One exception would be the patient with multiple pits within a small area or those with extensive underlying cavities where pit picking or Bascom’s I would result in a large midline defect. Even considering these caveats, only around 60% of patients with disease confined to the midline and only 40% of patients with lateral extensions had a minimal procedure. Many more could have been treated less invasively.

Analysis of preoperative demographic and disease characteristics revealed some factors that made a major excisional technique more likely, including recurrent disease and the presence of pus. More surprising is the association of being female with an excisional procedure unrelated to disease extent and complexity.

The relatively low utilisation of minimally invasive procedures becomes very relevant when outcomes are considered. Minimally invasive techniques lead to less pain at all time points and especially in the first week after surgery. Patients undergoing minimal intervention reported pain (on a 0–10 scale) around 1.5–1.7 units lower than those undergoing major excision on day 1 and day 7. Complications were also more common with major excision, with 15–20% excess seen after major excision. Time to wound healing and return to normal activities were significantly shorter after minimally invasive procedures, allowing patients to return to work, study and socialisation much faster. In contrast, these data confirm that major excisional techniques are around 10% more likely to cure the disease. This draws into question whether patients prefer a higher chance of cure in preference to more pain, more complications and a more protracted recovery. Such trade-offs are explored in Mixed-methods substudy and Discrete choice experiment.

It is plausible that the differences seen when comparing these two intervention groups relate to the case mix. For example, more extensive disease would be more likely to require major excision, but patients would, regardless of intervention, be more likely to have a complicated and protracted recovery. We attempted to control for case mix by correcting for risk factors with statistical modelling. We prespecified demographic and disease characteristics that previous literature had identified as influencing outcome, and assessed sensitivity to this via alternative models that were developed after discussion with the core study clinical team. Regardless of the model used, the association with postoperative outcomes remained consistent. It is also worth noting the substantial overlap in procedure types even among ostensibly similar subgroups of disease, which suggests that the choice of procedure may be driven as much by patient choice and surgical familiarity as it is by the severity of the disease.

A noteworthy result was the time to return to normal activities and the time to healing. The literature is full of reports of spectacular efficacy for several procedures.13,39,40 Yet regardless of intervention type, at least 10% and possibly up to 25% of patients had not healed by 6 months, and up to 12% of the major excision group had not returned to normal activities. This suggests that, in the real world as compared to specialist units, these procedures may not necessarily be as effective as the literature would suggest, and there may be a significant postoperative burden for patients of which many will not have been made aware during informed consent.

It is likely that patients have different interpretations of the terms ‘recurrence’ and ‘wound healing’ and may have considered these as interchangeable when telling us about their disease during follow-up. In the true sense, recurrence means disease that has healed after surgery but which then arises again, at the same site or at a different site to the original disease. This should be distinguished from disease that fails to heal at all after surgery or indeed when surgery achieves excision of the pits, but the patient is left with an unhealed wound, often in the midline. True recurrence is probably much less common and requires a protracted length of time to detect accurately.41 Failure to heal or persistence of disease is probably easier to define, but there remains an issue as to the time point when the intervention should be considered to have healed and not recurred. Consensus from the European Society of Coloproctology working party on pilonidal guidelines has proposed that 6 months after surgery would seem an appropriate time point (Asha Senapati, September 2023, personal communication). If a 25% failure rate was observed, regardless of technique, this would certainly be inferior to most of the reported literature.13,39,40 This has repercussions when it comes to SDM before surgery.

The difference between this study’s observed failure rate and that reported may relate to the skill and experience of individual surgeons. It may be that experts in pilonidal surgery can achieve healing rates equivalent to those in the literature.25,42–45 Our study partly triangulated this theory: surgeons responsible for more cases (defined here as treating 10 or more study participants) had better outcomes than surgeons who treated fewer participants, but recurrence was still higher than in previous literature. This could imply some surgeons are more skilled than others, although numbers were small, and the healing rates were not controlled for factors such as technique, case volume and disease severity. An important limitation is that this analysis was not risk-adjusted, and disease characteristics (and hence outcomes) may differ between experienced and less experienced surgeons. Nevertheless, it is unlikely that more experienced surgeons would be systematically allocated easier-to-treat patients, and so disease severity is unlikely to be the reason for these differences. This may justify referral to specialist centres for those patients with particularly complex disease.

The recognised issues with major excision in terms of protracted wound healing and the reported high failure rate with excision and midline closure mean that grouping such procedures along with asymmetric closure and flap procedures15 may have skewed the results in favour of minimal intervention. Fortunately, the number of patients in the asymmetric closure group (Karydakis and Bascom cleft closure) was sufficient to allow us to carry out a more granular analysis. Despite excluding these other procedures, the outcomes were similar for cleft closure techniques compared to minimally invasive techniques. Patients who had an asymmetric closure had more early pain, a higher complication rate and longer time to healing and return to normal activity than those treated with minimally invasive procedures. Failure was 5–8% less likely but the rate was still much higher than in most of the reported literature.13

When one considers this potential high rate of failure, it is somewhat surprising that the DR was so low. Patients were mainly satisfied with their decision for surgery and in addition felt that the options and outcomes were explained well to them. The CollaboRATE scores were high, suggesting the SDM process was good. These findings somewhat contradict other data suggesting that many are not offered a range of operations, in particular minimally invasive procedures. Data should perhaps be interpreted with caution as they may reflect a social desirability bias.46

There are limitations to both these data and their interpretation. An obvious limitation is the incompleteness of the data: 1 in 10 patients had missing day 1 data, and 6-month data were only available in around three-quarters of patients. This is despite rigorous governance processes and dedicated research nurses assiduously following up the patients. The study period did fall during the COVID pandemic, and this will have influenced the ability to follow up rigorously in some cases. Otherwise, the incompleteness of the data is probably a reflection of the demographic of the group, which tends to consist of young active working people, predominantly male. Such a demographic may be less likely to respond to follow-up calls.47,48 Interestingly, this is not the case in virtually all published series with follow-up of greater than 12 months.49–58 Most of these trials report complete data collection and all have more than 90% attending for follow-up. It is unclear how these other studies succeeded in such an astonishing rate of attendance. One study in particular reported follow-up data involving a clinic visit 5 years after surgery.51 The authors countenance a follow-up of at least this long if recurrence rates are to be considered accurate.41 While they may be correct, such a follow-up period is not practical and is likely to produce levels of incomplete data far exceeding 25% if carried out in the UK. While we did manage to obtain data for healing, complications and return to normal activities in at least 83% of participants, in some cases this was at only around the 6-week follow-up.

Other limitations relate to the multitude of interventions that were included in the cohort. While analysable data were available for 667 participants, there were 12 different interventions carried out. There were just not enough data to compare individual procedures in any meaningful way. If we had aimed to have 100 patients for each procedure rather than just the front-running procedures, a cohort of over 6000 patients would have been required. It was felt justified to group similar techniques by invasiveness. Minimally invasive procedures have a distinct commonality in that they do not involve major skin/subcutaneous tissue excision and instead focus on destruction/removal of the pit and underlying cavity. Major procedures involve excision of the disease and surrounding skin with or without closure of the wound created. There are clearly subtleties for the way both minimal and major interventions are carried out and advocates will proclaim the benefits and harms of each. Indeed, the literature is full of such comparisons.40 As such, a broad categorisation may be criticised, particularly for the major excisional group where excision-and-leave-open or closure in the midline techniques is considered by many to be obsolete compared with an asymmetric closure technique.15,37,38 Nevertheless, even when a more granular analysis of asymmetric closure versus minimally invasive techniques was carried out, similar differences in terms of pain, recovery, failure to heal and complications were seen.

The final limitation relates to the risk adjustment. Demographic and disease characteristics that may influence outcome are currently unproven. We decided on such parameters by consensus among the core study clinical team. We accommodated the uncertainty by including multiple permutations of the risk adjustment model. Overall, these permutations led to similar results, strengthening the justification for the model and the overall conclusions. Nevertheless, all risk adjustment is predicated on being able to fully quantify ‘risk’, which is both a strong and untestable assumption.

Conclusions

The real-world experience of surgery for pilonidal disease is not as good as the literature would suggest. Many patients have a protracted recovery regardless of intervention, and failure is common. The utilisation of minimally invasive techniques could be increased and would reduce the burden of postoperative recovery substantially while accepting a small reduction in cure rate. Patient QoL and health economics studies, including investigation on the cost to society of longer absences from work and education, may inform shared decisions on first-line treatment of PSD in the future.