The diagnosis of anal fistula is based on the German S3 guidelines: anal abscess and fistula (23). All patients were diagnosed with anal fistula by anal finger examination, anoscope examination, radiographic examination (including rectal endoluminal ultrasound, pelvic CT, or MRI), or intraoperative probe/methylene blue staining, and the number of internal orifices was counted by these techniques. The diagnostic criteria for T2DM were based on the latest Chinese guidelines for the prevention and treatment of T2DM set by the Chinese Diabetes Society (24, 25). And the diagnosis was assigned by an endocrinologist. Relevant data were collected on the cases, including demographic characteristics, clinical features, laboratory and ancillary tests at admission, anal fistula-related information (e.g., previous surgical history, anal fistula types, number of internal orifices, etc.), pre- and post-surgical treatments, and surgical modalities. Non-healing (refractory) group refers to trauma that cannot be repaired in time with conventional therapy or wounds that can not achieve functional recovery and anatomical integrity (26). The last routine dressing change time in the outpatient clinic was collected as the outcome indicator. Judged by the specialist anorectologist and the definition of the relevant literature, patients were divided into the non-healing (refractory) group or healing group according to whether its recovery period is longer than 35 days (27–29).

Among the underlying diseases, hypertensive disease and non-alcoholic fatty liver diseases are listed independently. Chronic cardiovascular diseases included coronary atherosclerotic heart disease and lacunar cerebral infarction. Chronic lung diseases included tuberculosis, chronic obstructive pulmonary disease, and chronic pulmonary heart disease. Chronic liver diseases included chronic viral hepatitis B, cirrhosis of the liver, hepatic hemangioma, etc.

This clinical retrospective study collected 408 cases of anal fistula combined with T2DM attending the Hospital of Chengdu University of Chinese Medicine from June 2017 to May 2022. The study excluded anal fistula caused by Crohn's disease or exotic injuries (from indigestible diets or external devices). Extremely complicated fistulas that were not suitable for fistulotomy, cutting seton surgery or a combination procedure were screened out. And fistulas in the active phase of anal fistula were excluded. Research recruited 408 patients, but 43 patients were eliminated because of recurrence (within 6 months) (n = 35), severe clinical data deficits, failure to perform surgery, or missed visits. In the end, a total of 365 patients were included. Up on wound recovery time, patients were divided into the non-healing (refractory) group (n = 170) and the healing group (n = 195), respectively. After matching, there were 122 patients in each group. The detailed operation procedure was shown in Figure 1. This study was approved by the ethics committee of the Hospital of Chengdu University of Traditional Chinese Medicine (ethics number: 2022KL-018). The research was a retrospective clinical study and only the patients’ previous treatment data were extracted through the medical system. Therefore, no informed consent was required from the patients.

All statistical analysis were performed with SPSS version 22.0 (IBM, Armonk, NY, USA). Continuous variables were expressed as mean ± standard (SD) deviation or M (QL, QU). Before conducting propensity score, the student's t-test and Mann–Whitney U test were applied for continuous variables, whereas the χ² test was used for dichotomous variables. To eliminate relevant confounders and increase comparability between different group, PSM analysis was performed using nearest-neighbor matching (1:1) adjusted for baseline data including: gender, age, body mass index (BMI), history of anal fistula surgery, the number of internal orifices (≥2 or not) and the number of wounds (≥3 or not), with the caliper value set at 0.02. After PSM, variables were tested by paired t-test and Wilcoxon rank sum test for continuous variables. McNemar test and Fisher's exact test for categorical variables. Risk factors for non-healing wound were identified by univariate and multivariate logistic regression models, and the degree of association was shown by calculating odds ratio (OR). Variance inflation factor (VIF) and tolerance were applied to determine the covariance between internal orifice and incision, as well as multicollinearity among variables before multivariate regression. To prevent overfitting, multiple variables with statistically different (p < 0.05) were selected for multivariate logistic regression analysis to identify independent predictors of healing trauma. Also, we used ROC curve to assess the sensitivity and specificity of glucose indicators in wound healing. All statistical tests were two-sided, and p < 0.05 was considered statistically significant. The experimental procedure of the article was carried out with reference to the relevant authoritative literature (27).

The majority patients were male, reaching astonishing 347 cases (95.07%). The healing group had more female patients than the refractory group. However, there was no discernible difference between each other (p = 0.248). BMI (26.83 ± 0.33 vs. 26.97 ± 0.28 Kg/m²) and previous surgical history (19 vs. 21) in the healing group were slightly higher than those in the non-healing group. According to more than two types of ancillary findings, the number of patients with multiple internal orifices was larger in the healing group (n = 173, 88.72%), while cases with ≥3 wounds were more in the refractory group (42 vs. 89, p = 0.003). No significant difference was discovered between two groups in terms of underlying diseases (p > 0.05).

122 pairs of patients were successfully matched (Table 1) with no significant difference between gender (p = 1.000), age (p = 0.923), BMI (p = 0.547), previous anal fistula surgery history (p = 1.000), number of internal orifices (p = 1.000), and number of wounds (p = 1.000).

The results of the detailed laboratory and ancillary tests are presented in Table 2. The concentration of plasma sodium (137.45 mmol/L [135.78, 140.10] vs. 139.45 mmol/L [137.58, 140.83], p < 0.001) and plasma chloride (102.94 ± 0.34 vs. 103.94 ± 0.29 mmol/L, p = 0.039) showed a higher level in non-healing group. No significant abnormalities or intergroup differences were discovered in the rest of electrolytes such as potassium, calcium, magnesium and phosphorus in plasma. Uric acid, one of the key metabolic byproducts in the body, demonstrated an aberrant range in the non-healing group, where the level reached 359.50 (292.50, 433.50) mmol/L, compared to the healing group's 314.00 mmol/L (255.50.372.00). In terms of lipid metabolism, both groups displayed abnormal ranges in cholesterol (5.46 mmol/L [4.43, 6.25] vs. 5.16 mmol/L [4.37, 6.20]) and triglyceride (1.99 mmol/L [1.34, 2.78] vs. 2.30 mmol/L [1.67, 3.41]). However, no statistical significance (p > 0.05) was detected in cholesterol, triglyceride, high-density lipoprotein (LDL), low-density lipoprotein (HDL). Meanwhile, except for the A/G ratio (1.60 [1.40, 1.80] vs. 1.70 [1.50, 1.95], p = 0.041), the remaining indicators of albumin, globulin, alanine transaminase (ALT) and aspartate transaminase (AST) were not found to be different from the normal range or demonstrate differences between groups.

Anal fistula lesions are confined by collagen and epithelial tissue and the infection is mostly under control. The results showed that white blood cell count (WBC) and neutrophil percentage (NEUT%) fluctuated in the normal range in both groups, while C-reactive protein (CRP) was higher than normal range. The NEUT% in the healing group reached 67.80% (61.00, 74.88) was greater than in non-healing group [66.15% (54.78, 74.13), p = 0.014]. At the same time, lymphocyte percentage (28.25% [18.00, 52.23] vs. 26.20% [18.15, 47.33], p = 0.039) and monocyte percentage (8.75% [5.60, 34.50] vs. 7.35% [5.30, 29.53], p = 0.008) in refractory group also exhibited a superior level. 143 patients carried out anorectal endoscopic ultrasonography, and the positive detection of fistulas internal orifices in the implemented patients reached 94.44%.

It exhibits a clear distinction between glucose indicators. It is worth mentioning that random intravenous blood glucose (tested with other biochemical indicators) and HbA1c were extracted preoperatively. The finger-prick glucose test was used to monitor FBG and PBG. Patients in the healing group had lower random intravenous blood glucose (8.88 mmol/L [6.61–12.41]) compared to 12.39 mmol/L (8.27–16.64) in refractory group, with a p value less than 0.001. In response to medium and long-term blood glucose levels, HbA1c levels at admission and maximum PBG, and FBG levels throughout hospitalization has been collected. We noticed that HbA1c indicated a higher standard in non-healing group (8.80 mmol/L [7.68, 10.03] vs. 7.40 mmol/L [6.50, 8.73], p = 0.001). During the hospitalization, patients’ maximum FBG and the maximum PBG respectively reached 9.00 mmol/L (7.50, 10.40) and 15.15 mmol/L (13.20, 17.53) in the refractory group. Striking statistical differences between the two groups were shown (p < 0.001). Among the glucose related indicators, the area under the ROC curve of maximum FBG reached 0.724, followed by HbA1c 0.667, random blood glucose 0.665 and maximum PBG 0.640 (Figure 2). The critical values of HbA1c, random blood glucose, maximum FBG and maximum PBG were 7.69%, 12.50, 8.35 and 16.05 mmol/L, respectively. The sensitivity of each index at the critical values was 75.4%, 50.0%, 64.8%, and 41.8%, and the specificity was 58.2%, 76.2%, 74.6%, and 82.8%, respectively.

Of these 244 patients, 40.98% (n = 100) were diagnosed with high anal fistula which patients in the refractory group was slightly higher than in the healing group (n = 56 vs. n = 44). Different surgical techniques were carried out by the surgeons depending on the type of fistula. 30 (12.30%) patients underwent cutting seton surgery, 144 (59.02%) patients performed anal fistulectomy and 70 (28.69%) patients underwent both procedure (hybrid surgery). Fisher's exact test did not reveal any statistical difference either overall or between the two different procedures (p > 0.05). In order to explore the effects of different incisions in wound healing, we summarized the characteristics of the distribution of wounds under lithotomy position. 6 o’clock incision was chosen most (n = 123, 50.41%), followed by the 3 o’clock incision (n = 64, 26.23%), 5 o’clock incision (n = 60, 24.59%), and 1 o’clock incision (n = 53, 21.27%). Both 1 o’clock and 5 o’clock wounds exhibited a higher incidence in the non-healing group and there were statistical differences in both groups (p = 0.005, p = 0.029). Milligan-Morgan surgery were conducted in 70 patients (28.69%) at the same time, while in the healing group this proportion only reached 25 cases (20.49%). The full surgery-related data are shown in Table 3.

Treatment for anal fistula and T2DM must be administrated simultaneously when they are both present. To highlight the relationship between various therapies and wound healing, we gathered several interventions and results presented in Table 4. Among these patients, 92 (37.70%) patients had their first T2DM diagnosis with no statistically significant difference (p = 0.665). Insulin pumps were used for glycemic control in 61 (25.00%) patients during hospitalization, including 41 (33.61%) in the refractory group and 20 (16.39%) in the healing group. Among the patients who used them, the duration did not show a difference (6.49 ± 0.39 d vs. 5.00 ± 0.64 d, p = 0.053). There was no significant difference between the two groups regarding the adoption of subcutaneous insulin, but it did show a higher utilization rate on oral hypoglycemics in the healing group (71 vs. 90, p = 0.015).

In 244 cases, the rate of preoperative antibiotics reached 65.16%, with a higher proportion in the refractory group. The duration of postoperative intravenous antibiotics usage (5d [5, 7] vs. 5d [5, 7], p = 0.14) did not demonstrate a significant difference. 89.75% (n = 219) of the patients used only one antibiotic and 10.25% (n = 25) used two or more antibiotics after operation, no significant difference was seen between the groups (p > 0.05).

After the univariate logistic regression analysis of statistically significant indicators, all outcomes revealed various degrees of associations with wound healing with the exception of LY% (Table 5). Among the laboratory indices, elevated plasma sodium, chloride, A/G ratio, and NEUT% showed varying degrees of protection. Random intravenous blood glucose (OR: 1.137, 95% CI: 1.072–1.206, p < 0.001), HbA1c (OR: 1.284, 95% CI: 1.104–1.494, p = 0.001), maximal FBG (OR: 1.581, 95% CI: 1.1.313–1.904, p < 0.001) and excessive levels of maximal PBG (OR:1.182, 95% CI: 1.079–1.296, p < 0.001) were shown to be impairing factors for wound healing. Intriguingly, the usage of insulin pump had an OR > 1 in the univariate logistic regression analysis, whereas oral hypoglycemic drugs unsurprisingly played a protective role (OR = 0.50). In surgical treatment, 1 and 5 o’clock incision and concurrent M-M surgery both served to retard wound healing.

To prevent overfitting, the research included plasma sodium, chloride, uric acid, random intravenous blood glucose, HbA1c, maximum FBG, maximum PBG, insulin pump usage and oral hypoglycemic drug use, 1 and 5 o’clock incision into the multivariate analysis by referring relevant literature and clinical experience. The results of multivariate logistic regression analysis are presented in Table 6. It revealed that the decrease of NEUT% (OR: 0.906, 95% CI: 0.856–0.958, p = 0.001), elevation of uric acid (OR: 1.008, 95% CI: 1.002–1.015, p = 0.012), maximum FBG (OR:1.489, 95% CI:1.028–2.157, p = 0.035), random intravenous blood glucose (OR: 1.130, 95% CI: 1.008–1.267, p = 0.037) and incision at 5 o’clock (OR: 3.510, 95% CI: 1.214–10.146, p = 0.020) were independent risk factors for refractory wounds.