Development and validation of a nomogram predicting the risk of necrosis in fingertip replantation: a single-center retrospective study Authors: Yadong Li, Xinwei Hou, Shuai Dong, Yan Zhang, Zhongliang Wang, Jihui Ju, Guangzhe Jin Published: 2026 ABSTRACT BackgroundNecrosis remains an important complication after fingertip replantation. We aimed to develop and internally validate a model for estimating necrosis risk using clinical, procedural, and early-postoperative information.MethodsWe retrospectively analyzed 1,344 patients who underwent replantation of 1,497 fingertips at a single center between August 2018 and August 2023. The patient was the unit of analysis. Patients were divided by outcome-stratified random sampling (seed 2222) into a derivation cohort (n = 942; 70%) and a split-sample internal validation cohort (n = 402; 30%). Candidate predictors were reduced by 10-fold cross-validated least absolute shrinkage and selection operator regression. A multivariable logistic regression model was specified using six predictors available during the early postoperative course. Discrimination, calibration, Brier score, and decision-curve analysis were evaluated.ResultsNecrosis occurred in 115 patients (12.2%) in the derivation cohort and 48 (11.9%) in the internal validation cohort. The model retained surgeon experience, platelet count (PLT), neutrophil percentage (NEU%), D-dimer, hypertension, and vascular crisis. The area under the receiver operating characteristic curve was 0.953 (95% CI, 0.935–0.968) in the derivation cohort and 0.974 (95% CI, 0.958–0.987) in the internal validation cohort. Corresponding Brier scores were 0.051 and 0.042. In the internal validation cohort, the calibration intercept was −0.338 and the calibration slope was 0.759; the Hosmer–Lemeshow test yielded P = 0.729. Decision-curve analysis suggested potential net benefit across clinically plausible threshold probabilities, but external validation is required.ConclusionThis split-sample internally validated nomogram estimates necrosis risk after fingertip replantation once early-postoperative laboratory results and vascular-crisis status are available. The observed discrimination should be interpreted alongside calibration shrinkage, possible overfitting, and the absence of external validation. Background As a rapidly developing manufacturing country, China has seen a significant increase in hand injuries, particularly fingertip amputations, which are common in hand trauma cases. Severe fingertip injury was once considered a contraindication for replantation, with most surgeons opting for amputation (1, 2) or the use of flaps (3, 4). However, the influence of traditional Chinese concepts coupled with the widespread availability of social health insurance, Chinese patients has led to elevated expectations among Chinese patients regarding the preservation of their fingers, resulting in a significantly higher rate of fingertip replantation in China compared with other countries. With the development of microsurgical techniques, the success rate of fingertip replantation in most Chinese hospitals has exceeded 80%. However, fingertip replantation still faces numerous challenges, and some patients experience failure postoperatively. Tissue necrosis not only affects the psychological health of patients but can also have long-term consequences for the appearance and functionality of the hand (5). Previous studies have identified predictors of digit-replantation failure and have proposed prediction models, including models based on large retrospective or multicenter cohorts (5–8). The present study differs by focusing on distal fingertip replantation and integrating procedural characteristics with routinely available early-postoperative laboratory markers and vascular-crisis status. It is therefore intended as an early-postoperative, rather than preoperative, risk-stratification tool. This incremental contribution and its potential clinical use require confirmation through external validation. Patients and methods Study design and patient population We retrospectively screened 1,344 patient records covering 1,497 replanted fingertips treated at our hospital between August 2018 and August 2023. No patient was excluded for missing predictor or outcome data. The patient was the unit of analysis: each patient contributed one record and one binary patient-level outcome, including patients with multiple replanted fingertips. Patients were divided by outcome-stratified random sampling (seed 2222) into a derivation cohort (n = 942; 70%) and a split-sample internal validation cohort (n = 402; 30%). Inclusion criteria were (i) fingertip amputation at or distal to the distal interphalangeal joint and (ii) age >18 years. Exclusion criteria were (i) refusal of follow-up or unavailable essential predictor/outcome data; (ii) absence of vascular or nerve anastomosis; (iii) mental illness or communication disorders; and (iv) vascular injury in the ipsilateral arm or forearm. Data were anonymized. The institutional ethics committee approved the study and waived informed consent. The participant flow is shown in Figure 1. Study flowchart. Flow diagram of patient inclusion, cohort allocation, variable selection, model development, and validation. Potential prognostic factors Demographic data such as age, sex, smoking status (smoker or non-smoker), and comorbidities (e.g., hypertension, diabetes) were self-reported. Clinical risk factors were derived from medical records. Preoperative risk factors included injury mechanism, which finger was injured, injury level, number of amputated fingers, preservation method of amputated fingers, presence of other injuries, degree of amputation (complete or incomplete), bone and joint injury, nerve injury, tendon injury, and nail bed or dorsal skin injury. Intraoperative risk factors included ischemia time (the time elapsed between the restoration of blood flow and the reported injury time), duration of surgery, surgeon experience (specialist was defined as having performed more than 50 finger replantation surgeries, while trainee was defined as having performed 49 or fewer finger replantations), vein graft usage, number of arteries repaired, number of veins repaired, arterial to venous ratio, arterial injury severity, and whether flap coverage was used. Postoperative candidate predictors included PLT (×10⁹/L), NEU% (percentage points), hemoglobin (g/L), albumin (g/L), D-dimer (mg/L), and vascular crisis. Laboratory measurements were obtained 3–5 days after replantation. Vascular crisis was defined as a clinically documented arterial or venous perfusion disturbance after surgery and before determination of the final replantation outcome: arterial crisis comprised ischemia due to arterial-anastomosis occlusion or inadequate inflow, whereas venous crisis comprised congestion due to venous-anastomosis occlusion or inadequate outflow. The model is intended for use during the early postoperative course, after these laboratory results and vascular-crisis status have become available; it is not a preoperative model. Outcome definition and assessment The binary outcome was complete necrosis/replantation failure documented in the medical record on the basis of irreversible loss of perfusion and tissue viability. Partial superficial necrosis without complete replantation failure was not counted as an event. Outcome status was abstracted from the final documented clinical assessment during available follow-up; no uniform minimum follow-up interval was imposed. Assessments were made by the treating hand-surgery team, who were not blinded to clinical and laboratory information. These features may introduce ascertainment bias. Statistical analysis Continuous variables are reported as mean ± standard deviation or median (interquartile range), as appropriate. Distributional assumptions were evaluated before between-group comparisons; independent-samples t tests or Mann–Whitney U tests were used for continuous variables, and chi-square or Fisher exact tests were used for categorical variables. Baseline characteristics were compared between the derivation and internal validation cohorts. Candidate variables were chosen from demographic, injury-related, procedural, and early-postoperative factors available in the dataset. The resulting dataset contained no missing predictor or outcome values; therefore, neither case deletion nor imputation was required. Least absolute shrinkage and selection operator (LASSO) logistic regression with 10-fold cross-validation was used for variable reduction. The one-standard-error criterion (lambda.1se = 0.01601135; log [lambda.1se] = −4.134) retained nine candidate variables. Six clinically interpretable variables were retained in the multivariable logistic model and nomogram. Binary predictors were indicator-coded; any arterial or venous crisis was coded as vascular crisis present. PLT was scaled per 10 × 10⁹/L, NEU% per percentage point, and D-dimer per 1 mg/L. Multicollinearity was assessed using variance inflation factors (maximum VIF = 1.09). Model discrimination was quantified by the area under the receiver operating characteristic curve (AUC) with bootstrap 95% confidence intervals. Calibration was assessed using plots, calibration intercept and slope, Brier score, and the Hosmer–Lemeshow test. Decision-curve analysis was restricted to clinically interpretable thresholds. Reporting was revised with reference to the TRIPOD framework. Results Comparison of the derivation and internal validation cohorts The analysis cohort comprised 1,344 patients, of whom 163 (12.1%) experienced necrosis. Outcome-stratified random sampling assigned 942 patients to the derivation cohort (115 necrosis events) and 402 to the split-sample internal validation cohort (48 events). Baseline characteristics were broadly similar between cohorts (Table 1). Table 1Complete baseline characteristics of the derivation and internal validation cohorts.VariableDerivation (n = 942)Internal validation (n = 402)P valueNecrosis0.963 No827 (87.8%)354 (88.1%) Yes115 (12.2%)48 (11.9%)Sex0.018 Male732 (77.7%)336 (83.6%) Female210 (22.3%)66 (16.4%)Age (years)38.5 ± 14.040.2 ± 13.70.033Ischemia time (hours)5.1 ± 2.05.3 ± 2.00.181Fingertip preservation0.389 Isolated ice-bag cooling3 (0.3%)0 (0.0%) Clean dressing744 (79.0%)326 (81.1%) Room temperature/other195 (20.7%)76 (18.9%)Multiple/associated injury0.140 No856 (90.9%)354 (88.1%) Yes86 (9.1%)48 (11.9%)Duration of surgery (hours)2.8 ± 1.33.3 ± 1.60.000Surgeon experience0.003 Trainee197 (20.9%)115 (28.6%) Specialist745 (79.1%)287 (71.4%)Injured finger0.666 Thumb165 (17.5%)70 (17.4%) Index275 (29.2%)100 (24.9%) Middle191 (20.3%)89 (22.1%) Ring134 (14.2%)65 (16.2%) Little83 (8.8%)35 (8.7%) Multiple digits94 (10.0%)43 (10.7%)Number of injured fingers0.702 One849 (90.1%)360 (89.6%) Two81 (8.6%)39 (9.7%) Three10 (1.1%)2 (0.5%) Four2 (0.2%)1 (0.2%)Injury level0.509 Free tissue block72 (7.6%)34 (8.5%) Distal to nail132 (14.0%)43 (10.7%) Middle nail level205 (21.8%)84 (20.9%) Nail-base level306 (32.5%)140 (34.8%) Distal interphalangeal joint227 (24.1%)101 (25.1%)Injury mechanism0.344 Electric saw56 (5.9%)28 (7.0%) Crush464 (49.3%)218 (54.2%) Twisting173 (18.4%)59 (14.7%) Blunt impact48 (5.1%)17 (4.2%) Heat press1 (0.1%)0 (0.0%) Avulsion16 (1.7%)3 (0.7%) Sharp cut184 (19.5%)77 (19.2%)Smoking history0.748 No791 (84.0%)334 (83.1%) Yes151 (16.0%)68 (16.9%)Hypertension0.936 No883 (93.7%)378 (94.0%) Yes59 (6.3%)24 (6.0%)Diabetes0.799 No929 (98.6%)395 (98.3%) Yes13 (1.4%)7 (1.7%)Vein graft0.108 No925 (98.2%)400 (99.5%) Yes17 (1.8%)2 (0.5%)Degree of amputation0.736 Incomplete39 (4.1%)19 (4.7%) Complete903 (95.9%)383 (95.3%)Bone injury0.709 Comminuted fracture343 (36.4%)140 (34.8%) Intra-articular fracture211 (22.4%)98 (24.4%) Transverse fracture388 (41.2%)164 (40.8%)Nerve injury0.216 Fingertip nerve injury267 (28.3%)104 (25.9%) Bilateral crush laceration344 (36.5%)167 (41.5%) Bilateral sharp laceration179 (19.0%)79 (19.7%) Bilateral avulsion injury152 (16.1%)52 (12.9%)Tendon injury0.314 No tendon rupture696 (73.9%)294 (73.1%) Flexor/extensor zone I235 (24.9%)107 (26.6%) Flexor/extensor zone II7 (0.7%)0 (0.0%) Flexor/extensor zone V4 (0.4%)1 (0.2%)Number of arteries repaired0.459 One116 (12.3%)46 (11.4%) Two747 (79.3%)314 (78.1%) Three or more79 (8.4%)42 (10.4%)Number of veins repaired0.789 One166 (17.6%)78 (19.4%) Two283 (30.0%)109 (27.1%) Three152 (16.1%)64 (15.9%) Four220 (23.4%)94 (23.4%) Five or more121 (12.8%)57 (14.2%)Artery-to-vein repair ratio0.534 1:09 (1.0%)2 (0.5%) 1:177 (8.2%)37 (9.2%) 1:253 (5.6%)15 (3.7%) 2:185 (9.0%)43 (10.7%) 2:2272 (28.9%)111 (27.6%) ≥2:3446 (47.3%)194 (48.3%)Arterial injury0.050 Fingertip artery transection265 (28.1%)97 (24.1%) Bilateral digital-artery crush injury338 (35.9%)171 (42.5%) Bilateral digital-artery sharp transection179 (19.0%)83 (20.6%) Bilateral digital-artery avulsion146 (15.5%)50 (12.4%) Avulsion with unilateral defect >1.5 cm2 (0.2%)0 (0.0%) Avulsion with bilateral defects >1.5 cm12 (1.3%)1 (0.2%)Nail-bed/dorsal skin injury0.636 Nail-bed laceration397 (42.1%)154 (38.3%) Nail-bed and dorsal-skin injury501 (53.2%)232 (57.7%) Crush injury9 (1.0%)5 (1.2%) Sharp/blunt laceration23 (2.4%)7 (1.7%) Avulsion injury7 (0.7%)3 (0.7%) Avulsion with partial skin defect5 (0.5%)1 (0.2%)PLT (×10⁹/L)235.5 ± 59.9230.4 ± 57.80.140NEU% (%)66.3 ± 10.767.3 ± 9.90.094D-dimer (mg/L)0.3 ± 0.40.4 ± 0.70.461Hemoglobin (g/L)137.5 ± 17.3137.6 ± 17.60.988Albumin (g/L)39.1 ± 3.738.7 ± 3.60.107Flap coverage0.668 No915 (97.1%)388 (96.5%) Yes27 (2.9%)14 (3.5%)Vascular crisis0.318 None888 (94.3%)386 (96.0%) Venous6 (0.6%)3 (0.7%) Arterial48 (5.1%)13 (3.2%)Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests.PLT, platelet count; NEU%, neutrophil percentage. Complete baseline characteristics of the derivation and internal validation cohorts. Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests. PLT, platelet count; NEU%, neutrophil percentage. Characteristics according to necrosis outcome In the complete cohort, the necrosis group had higher PLT, NEU%, and D-dimer values and a higher frequency of vascular crisis than the survival group. Necrosis was also more frequent in procedures performed by trainee surgeons. Selected outcome-stratified characteristics are shown in Table 2. Table 2Complete characteristics by outcome in the overall cohort.VariableSurvival (n = 1181)Necrosis (n = 163)P valueSex0.062 Male948 (80.3%)120 (73.6%) Female233 (19.7%)43 (26.4%)Age (years)38.8 ± 13.940.2 ± 14.50.269Ischemia time (hours)5.1 ± 1.95.9 ± 2.30.000Fingertip preservation0.568 Isolated ice-bag cooling3 (0.3%)0 (0.0%) Clean dressing944 (79.9%)126 (77.3%) Room temperature/other234 (19.8%)37 (22.7%)Multiple/associated injury0.081 No1070 (90.6%)140 (85.9%) Yes111 (9.4%)23 (14.1%)Duration of surgery (hours)2.9 ± 1.33.4 ± 1.70.001Surgeon experience0.000 Trainee204 (17.3%)108 (66.3%) Specialist977 (82.7%)55 (33.7%)Injured finger0.304 Thumb212 (18.0%)23 (14.1%) Index330 (27.9%)45 (27.6%) Middle245 (20.7%)35 (21.5%) Ring179 (15.2%)20 (12.3%) Little102 (8.6%)16 (9.8%) Multiple digits113 (9.6%)24 (14.7%)Number of injured fingers0.143 One1070 (90.6%)139 (85.3%) Two98 (8.3%)22 (13.5%) Three10 (0.8%)2 (1.2%) Four3 (0.3%)0 (0.0%)Injury level0.000 Free tissue block98 (8.3%)8 (4.9%) Distal to nail166 (14.1%)9 (5.5%) Middle nail level270 (22.9%)19 (11.7%) Nail-base level381 (32.3%)65 (39.9%) Distal interphalangeal joint266 (22.5%)62 (38.0%)Injury mechanism0.266 Electric saw75 (6.4%)9 (5.5%) Crush599 (50.7%)83 (50.9%) Twisting195 (16.5%)37 (22.7%) Blunt impact62 (5.2%)3 (1.8%) Heat press1 (0.1%)0 (0.0%) Avulsion16 (1.4%)3 (1.8%) Sharp cut233 (19.7%)28 (17.2%)Smoking history0.179 No995 (84.3%)130 (79.8%) Yes186 (15.7%)33 (20.2%)Hypertension0.000 No1119 (94.8%)142 (87.1%) Yes62 (5.2%)21 (12.9%)Diabetes0.458 No1165 (98.6%)159 (97.5%) Yes16 (1.4%)4 (2.5%)Vein graft0.024 No1168 (98.9%)157 (96.3%) Yes13 (1.1%)6 (3.7%)Degree of amputation0.848 Incomplete50 (4.2%)8 (4.9%) Complete1131 (95.8%)155 (95.1%)Bone injury0.000 Comminuted fracture452 (38.3%)31 (19.0%) Intra-articular fracture251 (21.3%)58 (35.6%) Transverse fracture478 (40.5%)74 (45.4%)Nerve injury0.000 Fingertip nerve injury349 (29.6%)22 (13.5%) Bilateral crush laceration439 (37.2%)72 (44.2%) Bilateral sharp laceration228 (19.3%)30 (18.4%) Bilateral avulsion injury165 (14.0%)39 (23.9%)Tendon injury0.000 No tendon rupture893 (75.6%)97 (59.5%) Flexor/extensor zone I280 (23.7%)62 (38.0%) Flexor/extensor zone II6 (0.5%)1 (0.6%) Flexor/extensor zone V2 (0.2%)3 (1.8%)Number of arteries repaired0.034 One147 (12.4%)15 (9.2%) Two936 (79.3%)125 (76.7%) Three or more98 (8.3%)23 (14.1%)Number of veins repaired0.091 One223 (18.9%)21 (12.9%) Two339 (28.7%)53 (32.5%) Three182 (15.4%)34 (20.9%) Four283 (24.0%)31 (19.0%) Five or more154 (13.0%)24 (14.7%)Artery-to-vein repair ratio0.223 1:011 (0.9%)0 (0.0%) 1:1104 (8.8%)10 (6.1%) 1:260 (5.1%)8 (4.9%) 2:1117 (9.9%)11 (6.7%) 2:2326 (27.6%)57 (35.0%) ≥2:3563 (47.7%)77 (47.2%)Arterial injury0.000 Fingertip artery transection339 (28.7%)23 (14.1%) Bilateral digital-artery crush injury441 (37.3%)68 (41.7%) Bilateral digital-artery sharp transection233 (19.7%)29 (17.8%) Bilateral digital-artery avulsion160 (13.5%)36 (22.1%) Avulsion with unilateral defect >1.5 cm0 (0.0%)2 (1.2%) Avulsion with bilateral defects >1.5 cm8 (0.7%)5 (3.1%)Nail-bed/dorsal skin injury0.000 Nail-bed laceration514 (43.5%)37 (22.7%) Nail-bed and dorsal-skin injury615 (52.1%)118 (72.4%) Crush injury12 (1.0%)2 (1.2%) Sharp/blunt laceration29 (2.5%)1 (0.6%) Avulsion injury8 (0.7%)2 (1.2%) Avulsion with partial skin defect3 (0.3%)3 (1.8%)PLT (×10⁹/L)227.3 ± 53.4282.2 ± 75.90.000NEU% (%)65.1 ± 9.977.3 ± 7.40.000D-dimer (mg/L)0.3 ± 0.40.8 ± 0.60.000Hemoglobin (g/L)138.3 ± 16.9132.2 ± 19.80.000Albumin (g/L)39.1 ± 3.738.2 ± 3.40.002Flap coverage0.002 No1152 (97.5%)151 (92.6%) Yes29 (2.5%)12 (7.4%)Vascular crisis0.000 None1139 (96.4%)135 (82.8%) Venous2 (0.2%)7 (4.3%) Arterial40 (3.4%)21 (12.9%)Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests.PLT, platelet count; NEU%, neutrophil percentage. Complete characteristics by outcome in the overall cohort. Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests. PLT, platelet count; NEU%, neutrophil percentage. The same directional patterns were observed in the derivation cohort: patients with necrosis had higher PLT, NEU%, and D-dimer values, more vascular crises, and a lower proportion of procedures performed by specialist surgeons (Table 3). Table 3Complete characteristics by outcome in the derivation cohort.VariableSurvival (n = 827)Necrosis (n = 115)P valueSex0.060 Male651 (78.7%)81 (70.4%) Female176 (21.3%)34 (29.6%)Age (years)38.2 ± 14.040.0 ± 14.30.220Ischemia time (hours)5.0 ± 1.95.7 ± 2.40.003Fingertip preservation0.605 Isolated ice-bag cooling3 (0.4%)0 (0.0%) Clean dressing656 (79.3%)88 (76.5%) Room temperature/other168 (20.3%)27 (23.5%)Multiple/associated injury0.729 No753 (91.1%)103 (89.6%) Yes74 (8.9%)12 (10.4%)Duration of surgery (hours)2.8 ± 1.23.1 ± 1.40.041Surgeon experience0.000 Trainee127 (15.4%)70 (60.9%) Specialist700 (84.6%)45 (39.1%)Injured finger0.651 Thumb147 (17.8%)18 (15.7%) Index240 (29.0%)35 (30.4%) Middle169 (20.4%)22 (19.1%) Ring121 (14.6%)13 (11.3%) Little72 (8.7%)11 (9.6%) Multiple digits78 (9.4%)16 (13.9%)Number of injured fingers0.387 One750 (90.7%)99 (86.1%) Two67 (8.1%)14 (12.2%) Three8 (1.0%)2 (1.7%) Four2 (0.2%)0 (0.0%)Injury level0.000 Free tissue block67 (8.1%)5 (4.3%) Distal to nail125 (15.1%)7 (6.1%) Middle nail level191 (23.1%)14 (12.2%) Nail-base level262 (31.7%)44 (38.3%) Distal interphalangeal joint182 (22.0%)45 (39.1%)Injury mechanism0.095 Electric saw51 (6.2%)5 (4.3%) Crush407 (49.2%)57 (49.6%) Twisting143 (17.3%)30 (26.1%) Blunt impact47 (5.7%)1 (0.9%) Heat press1 (0.1%)0 (0.0%) Avulsion13 (1.6%)3 (2.6%) Sharp cut165 (20.0%)19 (16.5%)Smoking history0.100 No701 (84.8%)90 (78.3%) Yes126 (15.2%)25 (21.7%)Hypertension0.000 No786 (95.0%)97 (84.3%) Yes41 (5.0%)18 (15.7%)Diabetes0.436 No817 (98.8%)112 (97.4%) Yes10 (1.2%)3 (2.6%)Vein graft0.010 No816 (98.7%)109 (94.8%) Yes11 (1.3%)6 (5.2%)Degree of amputation0.712 Incomplete33 (4.0%)6 (5.2%) Complete794 (96.0%)109 (94.8%)Bone injury0.000 Comminuted fracture320 (38.7%)23 (20.0%) Intra-articular fracture171 (20.7%)40 (34.8%) Transverse fracture336 (40.6%)52 (45.2%)Nerve injury0.000 Fingertip nerve injury251 (30.4%)16 (13.9%) Bilateral crush laceration297 (35.9%)47 (40.9%) Bilateral sharp laceration159 (19.2%)20 (17.4%) Bilateral avulsion injury120 (14.5%)32 (27.8%)Tendon injury0.001 No tendon rupture627 (75.8%)69 (60.0%) Flexor/extensor zone I192 (23.2%)43 (37.4%) Flexor/extensor zone II6 (0.7%)1 (0.9%) Flexor/extensor zone V2 (0.2%)2 (1.7%)Number of arteries repaired0.143 One104 (12.6%)12 (10.4%) Two659 (79.7%)88 (76.5%) Three or more64 (7.7%)15 (13.0%)Number of veins repaired0.189 One151 (18.3%)15 (13.0%) Two246 (29.7%)37 (32.2%) Three127 (15.4%)25 (21.7%) Four199 (24.1%)21 (18.3%) Five or more104 (12.6%)17 (14.8%)Artery-to-vein repair ratio0.498 1:09 (1.1%)0 (0.0%) 1:169 (8.3%)8 (7.0%) 1:247 (5.7%)6 (5.2%) 2:178 (9.4%)7 (6.1%) 2:2232 (28.1%)40 (34.8%) ≥2:3392 (47.4%)54 (47.0%)Arterial injury0.000 Fingertip artery transection248 (30.0%)17 (14.8%) Bilateral digital-artery crush injury295 (35.7%)43 (37.4%) Bilateral digital-artery sharp transection160 (19.3%)19 (16.5%) Bilateral digital-artery avulsion117 (14.1%)29 (25.2%) Avulsion with unilateral defect >1.5 cm0 (0.0%)2 (1.7%) Avulsion with bilateral defects >1.5 cm7 (0.8%)5 (4.3%)Nail-bed/dorsal skin injury0.000 Nail-bed laceration369 (44.6%)28 (24.3%) Nail-bed and dorsal-skin injury421 (50.9%)80 (69.6%) Crush injury8 (1.0%)1 (0.9%) Sharp/blunt laceration22 (2.7%)1 (0.9%) Avulsion injury5 (0.6%)2 (1.7%) Avulsion with partial skin defect2 (0.2%)3 (2.6%)PLT (×10⁹/L)229.4 ± 54.0279.7 ± 78.90.000NEU% (%)64.8 ± 10.276.9 ± 7.50.000D-dimer (mg/L)0.3 ± 0.30.8 ± 0.60.000Hemoglobin (g/L)138.1 ± 17.0133.2 ± 19.20.010Albumin (g/L)39.2 ± 3.738.4 ± 3.30.014Flap coverage0.000 No810 (97.9%)105 (91.3%) Yes17 (2.1%)10 (8.7%)Vascular crisis0.000 None796 (96.3%)92 (80.0%) Venous0 (0.0%)6 (5.2%) Arterial31 (3.7%)17 (14.8%)Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests.PLT, platelet count; NEU%, neutrophil percentage. Complete characteristics by outcome in the derivation cohort. Values are n (%) or mean ± SD. P values are from Welch t tests or chi-square tests. PLT, platelet count; NEU%, neutrophil percentage. Variable selection using LASSO with 10-fold cross-validation Ten-fold cross-validated LASSO regression was used for variable reduction (Figure 2a). The minimum-deviance penalty was lambda.min = 0.005754182, whereas the more parsimonious one-standard-error penalty was lambda.1se = 0.01601135 (log [lambda.1se] = −4.134). The lambda.1se criterion was used and retained nine non-zero candidate variables: surgeon experience, injured finger, number of injured fingers, hypertension, tendon injury, PLT, NEU%, D-dimer, and vascular crisis (Figure 2b). LASSO regression for predictor selection. (a) LASSO coefficient profiles of candidate variables. (b) Ten-fold cross-validation for selection of the optimal penalty parameter (λ). The λ.1se criterion was used to select nine candidate predictors. Multivariable logistic regression model The final six-predictor model is presented in Figure 3 and Table 4. Specialist surgeon status was associated with lower odds of necrosis (OR, 0.18; 95% CI, 0.10–0.32), whereas higher PLT (per 10 × 10⁹/L: OR, 1.20; 95% CI, 1.14–1.27), higher NEU% (per percentage point: OR, 1.17; 95% CI, 1.12–1.21), and higher D-dimer (per 1 mg/L: OR, 13.69; 95% CI, 6.21–30.18) were associated with higher odds. Hypertension (OR, 2.74; 95% CI, 1.04–7.24) and vascular crisis (OR, 2.79; 95% CI, 1.06–7.37) also remained independently associated with necrosis after adjustment. Forest plot of the multivariable logistic regression model. Points represent odds ratios (ORs), and horizontal lines represent 95% confidence intervals (CIs). Table 4Complete final logistic regression model.PredictorCoefficient (SE)OR (95% CI)P value; coding/unitconst−17.7171 (SE 1.8022)–<0.001Surgeon specialist (vs. trainee)−1.7204 (SE 0.2954)0.18 (0.10–0.32)<0.001PLT (per 10 × 10^9/L)0.1843 (SE 0.0283)1.20 (1.14–1.27)<0.001NEU% (per 1 percentage point)0.1534 (SE 0.0194)1.17 (1.12–1.21)<0.001D-dimer (per 1 mg/L)2.6170 (SE 0.4032)13.69 (6.21–30.18)<0.001Hypertension (yes vs. no)1.0081 (SE 0.4956)2.74 (1.04–7.24)0.042; 1 = yes; 0 = noAny vascular crisis (yes vs. no)1.0255 (SE 0.4958)2.79 (1.06–7.37)0.039; 1 = arterial/venous; 0 = nonelogit(p) = −17.7171–1.7204(specialist) + 0.0184332(PLT) + 0.153379(NEU%) + 2.617017(D-dimer) + 1.008071(hypertension) + 1.025548(any vascular crisis). Complete final logistic regression model. logit(p) = −17.7171–1.7204(specialist) + 0.0184332(PLT) + 0.153379(NEU%) + 2.617017(D-dimer) + 1.008071(hypertension) + 1.025548(any vascular crisis). Nomogram for early-postoperative risk estimation A nomogram was constructed from the complete multivariable model (Figure 4). The linear predictor was: −17.7171−1.7204 × specialist surgeon + 0.0184332 × PLT + 0.153379 × NEU% + 2.617017 × D-dimer + 1.008071 × hypertension + 1.025548 × any vascular crisis. Binary indicators were coded 1 when present and 0 otherwise. Predicted probability was calculated as 1/ [1 + exp(−linear predictor)]. The nomogram should be applied only after postoperative laboratory results obtained on days 3–5 and vascular-crisis status are available. Nomogram for predicting necrosis after fingertip replantation. The nomogram was developed using six predictors identified by multivariable logistic regression. Model performance and split-sample internal validation The derivation-cohort AUC was 0.953 (95% CI, 0.935–0.968), and the internal-validation AUC was 0.974 (95% CI, 0.958–0.987) (Figures 5a, b). Brier scores were 0.051 and 0.042, respectively. Apparent calibration in the derivation cohort had an intercept of approximately 0 and a slope of approximately 1, as expected when assessed in the model-development data. In the held-out cohort, the calibration intercept was −0.338 and the slope was 0.759, suggesting modest overall overprediction and predictions that were somewhat too extreme. Hosmer–Lemeshow tests did not detect statistically significant lack of fit in either the derivation cohort (P = 0.752) or the internal validation cohort (P = 0.729). However, nonsignificant Hosmer–Lemeshow results do not establish good or excellent calibration and were interpreted together with the calibration plot, intercept, slope, and Brier score (Figures 5c, d). Decision-curve analysis suggested potential net benefit over treat-none within portions of the examined threshold range; it does not establish clinical utility at a specific threshold, which requires prospective external evaluation (Figures 5e, f). Performance of the prediction model. (a,b) ROC curves in the derivation and internal validation cohorts, respectively. (c,d) Calibration curves in the derivation and internal validation cohorts, respectively. (e,f) Decision curve analyses in the derivation and internal validation cohorts, respectively. Discussion We developed a six-predictor model for necrosis after fingertip replantation using information available during the early postoperative course. Unlike models intended for preoperative counseling, this nomogram incorporates laboratory measurements obtained 3–5 days after surgery and vascular-crisis status. Its potential value lies in combining distal-fingertip-specific procedural information with routinely measured inflammatory and coagulation markers. High AUC values were observed in the split-sample cohorts; however, the validation calibration slope of 0.759 suggests some overfitting, and the findings should be interpreted cautiously until externally validated. Prior studies have reported prediction models and individual risk factors for digit-replantation failure (5–8). The present nomogram is therefore an incremental rather than first-in-field contribution. It is distinguished by its focus on distal fingertip replantation and by incorporation of early-postoperative PLT, NEU%, D-dimer, and vascular-crisis status. These features may support intensified monitoring once postoperative data are available, but they also restrict portability and prevent preoperative use. Specialist surgeon status was strongly associated with lower necrosis odds, plausibly reflecting greater experience with small-vessel anastomosis and venous identification (9). Hypertension and vascular crisis had directionally positive adjusted associations but wide confidence intervals crossing unity. Vascular crisis is also a post-surgical intermediate event on the pathway to necrosis; its inclusion improves contemporaneous risk estimation but limits causal interpretation and means that the nomogram should not be described as a baseline prognostic model. Higher PLT and D-dimer were associated with increased necrosis risk, consistent with thrombotic and hypercoagulable mechanisms. Wang et al. (6) reported that postoperative D-dimer on day 1 independently predicted replantation necrosis. Our measurements were obtained 3–5 days after surgery, a period that overlaps with the occurrence of vascular compromise. These markers may therefore capture both predisposition and evolving postoperative pathology rather than purely pre-existing risk. Postoperative NEU% was higher in the necrosis group than in the survival group, consistent with the positive adjusted coefficient (OR, 1.17 per percentage point). Inflammation can promote thrombosis through endothelial activation, neutrophil recruitment, platelet adhesion, and a hypercoagulable state (10, 11). This association is biologically plausible but should not be interpreted as causal because NEU% was measured after surgery and may reflect tissue injury, infection, or an evolving vascular event. Mechanistic and longitudinal studies are needed. This study has several limitations. It was retrospective and conducted at a single high-volume center. Only split-sample internal validation was performed, and the validation calibration slope indicated some overfitting; multicenter external validation and model updating are required. Although source verification resolved the single spreadsheet transcription error and no analyzed values were missing, retrospective data collection remains susceptible to undocumented measurement or recording error. The data file contained one patient-level record per patient despite 1,497 replanted fingertips; thus finger-specific outcomes and within-patient heterogeneity among multidigit injuries could not be modeled. Variable selection followed by ordinary logistic regression may introduce optimism. The inclusion of postoperative laboratory values and vascular crisis restricts the model to the early postoperative setting, and vascular crisis may be an intermediate event. Finally, the definition of specialist experience (>50 replantations) and case mix at this high-volume center may not transport to lower-volume settings. Conclusion This nomogram estimates necrosis risk after fingertip replantation using six clinical, procedural, and early-postoperative predictors (12–18). Although high AUC values were observed in split-sample internal validation, calibration shrinkage was present and no external validation has been performed. The model should therefore not be considered ready for routine clinical decision-making; if evaluated prospectively, it should be treated only as an adjunct during the early postoperative course. Future work should externally validate and, if necessary, recalibrate the model in diverse settings and investigate the mechanisms linking inflammatory and coagulation markers with necrosis.