1 INTRODUCTION

The treatment of diabetes is a major global public health challenge, with more than 415 million individuals affected worldwide in 2015, the trend of which is increasing numbers, estimated at 642 million in the next 20 years.1 Diabetes mellitus affects virtually all systems of the body and can lead to nephropathy, retinopathy, and diabetic foot ulcers (DFUs),2 a complication characterised by vascular lesions, neuropathy, arthropathy, and increased susceptibility to infections.3, 4 It is estimated that up to 15% of people with diabetes will develop foot ulcers at some point in their life.5 DFUs not only result in a significant personal financial burden, they also impair quality of life and are associated with high levels of amputation and high mortality.6, 7 Surgery is now a conventional therapy for severe DFUs. Because of poor wound healing, patients with DFUs frequently require multiple surgical interventions.

Good glycaemic control is critical for the prevention of diabetes and its complications. Maintenance of normoglycaemia or near-normoglycaemic blood glucose levels can minimise the risk of diabetic complications.8 Haemoglobin A1c (HbA1c) is considered the gold standard for the evaluation of glycaemic control since the publication of the results of the Diabetes Control and Complications Trial (DCCT).9 HbA1c reflects an individual's average level of glycaemia over the preceding 2 to 3 months and represents an important parameter closely related to chronic diabetic complications. However, there are a number of limitations with this measure because it does not accurately reflect acute events (such as hypoglycaemia) or glycaemic variability,10 instead representing long-term glycaemia. In addition, there are significant individual differences in the relationship between HbA1c and mean glucose level.11 A previous study found that an HbA1c of 8% could reflect a range of 6.11 to 11.94 mmol/L in mean blood glucose level because the index can be influenced by other factors, such as anaemia or renal failure.12, 13

As continuous glucose monitoring (CGM) has become more popular, time in range (TIR) may represent a meaningful indicator derived from CGM data that indicate glycaemic control.14, 15 TIR is defined as the percentage of time or the duration that glucose levels are within a target range each day, that target range typically being 3.9 to 10.0 mmol/L.16 TIR reflects hypoglycaemia events and blood glucose fluctuations that can provide guidance for glycaemic control.17 The validity of TIR using seven-point testing and the association with risk of development of retinopathy and microalbuminuria in DCCT have been demonstrated.18 The association between HbA1c and wound healing in DFUs has also been corroborated.19 However, the significance of TIR for wound healing in patients with diabetes remains unknown. Thus, the objective of the present study was to examine the association between TIR and postoperative wound healing in DFU patients.

2 MATERIAL AND METHODS 2.1 Study population

The present study was a retrospective study of patients with DFUs presenting at our hospital from May 2015 to December 2019. All hospitalised patients were managed based on established standards for glucose monitoring and wound treatment. The inclusion criteria were as follows: (a) patients with DFUs; (b) patients with foot ulcers treated surgically. Exclusion criteria were as follows: (a) patients with incomplete medical data; (b) patients with non-type 2 diabetes mellitus; (c) patients unwilling to comply with treatment recommendations; (d) patients with prior history of malignancy or mental disorders; (e) patients with severe metabolic disorders, such as diabetic ketoacidosis or in a hyperglycaemic hyperosmolar state. A total of 347 eligible patients were identified after screening, as described in the flowchart in Figure 1. The study protocol received authorization from the ethics committee of the hospital and complied with the principles of the Declaration of Helsinki. As a retrospective study, written informed consent was not required.

Population flowchart. TIR, time in range

2.2 Data collection

Patient data were collected from a prospectively collected database and the electronic medical records at the hospital. For each patient, the following clinical information was recorded: (a) patient information, including age, gender, body mass index (BMI), instances of ulceration, prior history of hypertension, lower extremity macroangiopathy, peripheral neuropathy, insulin use, history of smoking and drink, and previous amputations; (b) blood tests prior to surgery, including blood glucose, white blood cell (WBC) count, HbA1c value, and haemoglobin, albumin, creatinine, and blood lipid levels; (c) method of surgery; (d) clinical outcomes, including secondary surgery within 1 month, wound conditions at the time of discharge, duration of hospital stay, and hospital costs. Wagner classification scores were determined by reviewing the description of each DFU. The presence of osteomyelitis was also ascertained by clinical and imaging examinations at admission.

2.3 Glucose monitoring

Inpatients continued to receive their previously prescribed treatment regimens for diabetes. The attending physician reviewed daily glucose measurements and adjusted insulin or other drug doses as required. Additional endocrine consultations could be requested for patients with poor glycaemic control. In addition, blood glucose monitoring was reviewed weekly by a specialist in diabetes.

Seven-point capillary blood glucose values were measured both before and 2 hours after each meal and at bedtime by a diabetes nurse. Glucose levels were recorded in the electronic medical records. Glucose levels were measured using an Accu-Chek Inform II portable glucometer (Roche Diagnostics), which was calibrated as required.

2.4 Time in range

Glycaemic outcome and performance were reflected in TIR values, calculated for each patient during their hospital stay. TIR values were represented by glucose levels of 3.9 to 10.0 mmol/L and computed for each patient by calculating the percentage of glucose measurements in that range from the number of total measurements (X) and the number within range (Y) as follows: X/Y*100%.

All patients were classified into one of two groups based on whether their TIR value exceeded the predefined threshold t. In the present study, a higher threshold value suggests lower tolerance for dysglycaemia. A threshold value of 0.5 was selected as a measure of overall poor glycaemic control in inpatients, whereby the majority of readings exceeded the acceptable glycaemic value of 10 mmol/L. The value of 0.5 has been previously used to correlate the quality of glycaemic control with clinical outcomes.20 In addition, we conducted a receiver operating characteristic (ROC) curve in the enrolled patients. The values of TIR were plotted into an ROC curve according to secondary surgery (Figure S1), and the optimal cut-off value of TIR was 51.1%. Individuals were therefore placed in either Group I or II for those with TIR ≥50% or <50%, respectively. Hypoglycaemia was defined as at least one blood glucose measurement ≤3.9 mmol/L.

2.5 Surgical treatment

All patients with DFUs underwent at least one surgery during their hospital stay. Surgery was classified as debridement, minor amputation, or split-thickness skin grafting. Debridement and skin grafts were grouped in the same category because the patient backgrounds are different. The selection of protocol was dependent on wound severity and was chosen by the attending surgeon. The primary outcome measure of the study was whether secondary surgery was performed (as described above) within a month of the initial surgery. Patients received secondary operations after a short period, generally because of poor wound healing or wound infection. Other outcomes included length of hospital stay and cost of hospitalisation. Treatment and care were provided in accordance with the guidelines of the International Diabetic Foot Consensus.21 The components of wound treatment and care were as follows: (a) formal assessment of ulcer and surrounding skin; (b) provision of any necessary off-loading; (c) debridement or other necessary surgery; (d) appropriate dressing products; (e) appropriate antibiotic therapy; (f) nutrition and self-care.

2.6 Propensity score adjusted analysis

To minimise the impact of an imbalanced baseline, propensity score matching (PSM) was selected to balance any differences in baseline characteristics of patients between groups. The following factors were treated as covariates for PSM: BMI, preoperative albumin level, WBC count, duration of diabetes, baseline HbA1c, drinker or non-drinker, diabetes treatment, Wagner score, hypertensive status, and type of surgery. PSM was conducted using a 1:1 nearest neighbour matching algorithm and a calliper of 0.2. Clinical outcomes were then evaluated against the statistical measurements.

2.7 Statistical analysis

Normally distributed continuous data were presented as mean and standard deviation (SD), and non-normally distributed continuous data as median with interquartile range (IQR). Categorical variables were represented as numbers (%). A Student's t-test or Mann-Whitney U test was performed for continuous data. A chi-square test or Fisher's exact test was performed for categorical variables. The association between variables and outcomes was evaluated using logistic regression. Variables with a P value <.10 were selected for subsequent multivariate analysis. Subgroup analysis was conducted using univariate analysis. P values <.05 represented a statistically significant difference, and all tests were two-sided. Statistical analyses were conducted using SPSS software (version 22.0; IBM, Armonk, New York, USA), while PSM was calculated using the R version 4.0.3 programming environment.

3 RESULTS 3.1 Study cohort

A total of 347 patients with DFUs were eligible for this study from May 2015 to December 2019. Baseline characteristics of the original and matched patients, stratified by TIR, are displayed in Table 1. Prior to matching, the TIR < 50% group had a lower BMI, lower albumin levels, higher WBCs counts, higher baseline HbA1c, a longer history of diabetes, and higher Wagner scores than the TIR ≥50% group, and received insulin treatment. Following matching, 130 patients were included in each group, with baseline characteristics for each group that were well balanced.

TABLE 1. The comparison of patient clinical characteristics before and after propensity score matching Characteristic Patients before PSM Patients after PSM TIR ≥ 50% group (n = 187) TIR < 50% group (n = 160) P TIR ≥ 50% group (n = 130) TIR < 50% group (n = 130) P Sex, n (%) .125 .602 Female 72 (38.5%) 49 (30.6%) 47 (36.2%) 43 (33.1%) Male 115 (61.5%) 111 (69.4%) 83 (63.8%) 87 (66.9%) Age, median (IQR), year 67.0 (59.0–75.0) 65.0 (58.0–71.2) .273 67.0 (58.0–75.0) 65.5 (59.0–72.0) .449 BMI, median (IQR), Kg/m2 23.2 (21.2–26.1) 22.9 (20.8–24.5) .010a 22.3 (20.7–25.0) 23.0 (20.9–24.8) .526 Preoperative serum albumin, median (IQR), g/L 36.8 (32.2–40.2) 32.0 (27.8–36.1) <.001a 34.7 (30.0–38.2) 33.6 (28.9–37.0) .172 Preoperative haemoglobin, median (IQR), g/L 118.0 (102.5–130.5) 114.0 (99.0–127.0) .094 118.0 (97.0–129.0) 114.5 (99.0–127.0) .690

Preoperative WBC count,

median (IQR), X109/L

7.9 (6.5–10.1) 8.9 (7.1–13.4) <.001a 8.1 (6.5–10.1) 8.3 (6.7–12.2) .099 Preoperative HDL, median (IQR), mmol/L 0.92 (0.74–1.10) 0.81 (0.70–1.07) .113 0.90 (0.72–1.06) (0.71–1.09) .592 Diabetes duration, median (IQR), years 9.0 (4.0–18.0) 11.0 (8.0–20.0) <.001a 10.0 (5.0–20.0) 10.5(7.0–22.0) .245 Baseline HbA1c, n (%) <.001a .206 <7.5% 90 (48.1%) 30 (18.8%) 39 (30.00%) 30 (23.08%) ≥7.5% 97 (51.9%) 130 (81.2%) 91 (70.00%) 100 (76.92%) Drink, n (%) .037a .572 No 148 (79.1%) 111 (69.4%) 98 (75.4%) 94 (72.3%) Yes 39 (20.9%) 49 (30.6%) 32 (24.6%) 36 (27.7%) Smoking, n (%) .361 .576 No 140 (74.9%) 111 (69.4%) 93 (71.5%) 97 (74.6%) Yes 47 (25.1%) 49 (30.6%) 37 (28.5%) 33 (25.4%) Diabetes treatment, n (%) <.001a .252 Oral hypoglycaemic 136 (72.7%) 84 (52.5%) 84 (64.6%) 75 (57.7%) Insulin 51 (27.3%) 76 (47.5%) 46 (35.4%) 55 (42.3%) Wagner score, n (%) .048a .500 1 to 2 144 (77.0%) 108 (67.5%) 93 (71.5%) 88 (67.7%) 3 to 5 43 (23.0%) 52 (32.5%) 37 (28.5%) 42 (32.3%) Hypertensives, n (%) .048a .619 No 76 (40.6%) 82 (51.2%) 60 (46.2%) 64 (49.2%) Yes 111 (59.4%) 78 (48.8%) 70 (53.8%) 66 (50.8%) Lower extremity vascular disease, n (%) .974 .901 No 95 (50.8%) 81 (50.6%) 63 (48.5%) 64 (49.2%) Yes 92 (49.2%) 79 (49.4%) 67 (51.5%) 66 (50.8%) Kidney disease, n (%) .095 .763 No 155 (82.9%) 121 (75.6%) 103 (79.2%) 101 (77.7%) Yes 32 (17.1%) 39 (24.4%) 27 (20.8%) 29 (22.3%) Peripheral neuropathy, n (%) .677 .352 No 25 (13.4%) 19 (11.9%) 19 (14.6%) 14 (10.8%) Yes 162 (86.6%) 141 (88.1%) 111 (85.4%) 116 (89.2%) Type of surgery .005a .210 Debridement/skin grafting 108 (57.8%) 103 (64.4%) 82 (63.1%) 81 (62.3%) Minor amputation 60 (32.1%) 29 (18.1%) 32 (24.6%) 24 (18.5%) Amputation + skin grafting 19 (10.2%) 28 (17.5%) 16 (12.3%) 25 (19.2%) Abbreviations: BMI, body mass index; HDL, high-density lipoprotein; IQR, interquartile range; PSM, propensity score matching; WBC, white blood cell. 3.2 Clinical outcomes

As indicated in Table 2, of the 260 matched patients, 54 (20.8%) had undergone secondary surgery within a month of initial surgery. Patients in the TIR <50% group exhibited a higher rate of secondary surgery (15.4% vs 26.2%, P = .032), longer duration of hospital stay (median, 13.0 days vs 15.5 days, P = .045), and higher hospitalisation costs (median, 25 438 yuan vs 32 052 yuan, P < .001). There was no difference in in-hospital mortality between groups.

TABLE 2. Postoperative short-term outcomes Factors TIR ≥ 50% group (n = 130) TIR < 50% group (n = 130) P Secondary surgery, n (%) 20 (15.4%) 34 (26.2%) .032a In-hospital mortality, n (%) 2 (1.5%) 2 (1.5%) 1.000 Hospital stays, median (IQR) (days) 13.0 (10.0-17.0) 15.5(11.0-21.8) .045a Costs, median (IQR) (yuan) 25 438 (15861-41 974) 32 052 (21094-46 293) <.001a Abbreviations: IQR, interquartile range; TIR, time in range. 3.3 Risk factors for secondary surgery

Logistic regression analysis for risk factors associated with secondary surgery is presented in Table 3. In univariate analysis, TIR (P = .032), WBC count (P = .019), albumin level (P < .001), HDL (P = .005), Wagner score (P = .001), diabetes treatment (P = .029), and type of surgery (P = .002) were correlated with secondary surgery. Those factors that were significant were included in the multivariate logistic regression model. TIR (P = .034), Wagner score (P = .009), diabetes treatment (P = .006), and type of surgery (P = .013) were confirmed to be independent factors for secondary surgery.

TABLE 3. Univariate and multivariate logistic regression analysis of secondary surgery Variables Patients without Secondary surgery (n = 206) Patients with Secondary surgery (n = 54) Univariate analysis P Multivariate analysis P Age, year, median (IQR) 66.0 (58.0–73.0) 66.5 (59.5–74.0) .335 BMI 22.8 (20.8–25.0) 22.3 (20.4–24.4) .780 Diabetes duration, median (IQR), years 10.0 (6.0–20.0) 10.0 (5.3–20.0) .796 WBC count, median (IQR), X10̂9/L 8.1 (6.5–11.1) 8.8 (7.2–13.1) .019a 0.887 Preoperative serum albumin, median (IQR), g/L 35.1 (30.3–38.1) 31.6 (27.7–33.9) <.001a 0.193 Preoperative haemoglobin, median (IQR), g/L 118.0 (99.0–129.8) 112.5 (97.0–125.0) .094 0.440 HDL, median(IQR), mmol/L 0.88 (0.74–1.11) 0.74 (0.65–0.95) 0.005a 0.315 Baseline HbA1c 8.7 (7.5–10.1) 9.1 (7.2–10.4) .806 Hypoglycaemia, n (%) .636 0 138 (67.0%) 38 (70.4%) 1 68 (33.0%) 16 (29.6%) Wagner score, n (%) .001a 0.009a 1 to 2 153 (74.3%) 28 (51.8%) 3 to 5 53 (25.7%) 26 (48.2%) Diabetes treatment, n (%) .029