Risk Factors for Digital Replantation Failure: A Nomogram Prediction Model

Introduction

Finger amputation is a common traumatic disorder, which may lead to catastrophic physical and psychosocial repercussions. Replantation was first reported by Malt et al¹ to treat digital amputation with 70% to 90% survival rates based on previous studies.^2,3 Replantation provides hope to numerous patients, because it can preserve digital length and nail, provide decent soft-tissue soft-tissue coverage, and maintain good joint motion, which improves patient satisfaction cosmetically and functionally in comparison with other methods, such as skin grafting.^4,5 Recently, increasing evidence has been reported that injury mechanism, platelet count, smoking, preservation method of amputated part and the use of vein graft were closely related to digital necrosis (DN).⁶ Wojciech Dec⁷ found that patients with diabetes, smoking and crushing or avulsion injury were predictors for necrosis after replantation. Wang⁸ performed a retrospective study of 946 patients with replantation and first concluded that menstrual period, menopause, and the level of D-dimer were predictors of DN after digital replantation.

Ongoing literature is tried to investigate the relationship between inflammatory indicators in various diseases, such as ischemia-reperfusion injury.^9,10 As we know, litter research explored the role of inflammatory indicators, such as the complete blood count and derived inflammatory indicators, in the prediction of DN after digital replantation. Although various studies focused on the predictors of DN, its predictors remain debated. Hence, we performed a nomogram prediction model to investigate predictors of DN. As far as we know, this is the first study to assess the relationship between derived inflammatory indicators and predictors for DN.

Materials and Methods

Ethics Statement

The study was approved by the Institutional Review Board of Third Hospital of Hebei Medical University in compliance with the Helsinki, and an exemption from the informed consent was obtained. All data were anonymized before the analysis to safeguard patient privacy (W2021-046-1).

Patients

We included 332 patients, including 54 patients with DN and 278 who were not, receiving multiple digital replantation from Jan. 2017 and Jan. 2024 in our hospital. The inclusion criteria: (1) multiple complete finger separation; (2) more than 18 years old. The exclusion criteria were as follows: (1) congenital finger deformity; (2) history of hand surgery. We obtained information of demographics (age, gender, BMI, injury mechanism, ischemia time, probe artery, injury hand), comorbidities (patients with a history of heart valve disease, heart failure, heart infarction, diabetes, or hypertension), and admission laboratory examinations, including prothrombin time (PT), international normalized ratio (INR), fibrinogen (FIB), activated partial thromboplastin time (APTT), thrombin time (TT), D-dimer, white blood cell (WBC), neutrophil (NEU), lymphocyte (LYM), monocyte (MON), red blood cell (RBC), platelet (PLT), neutrophil-to-lymphocyte ratio (neutrophils count/lymphocytes count, NLR), Monocyte-to-lymphocyte ratio (monocytes count/lymphocytes count, MLR), Platelet-to-lymphocyte ratio (platelets count/lymphocytes count, PLR), systemic immune-inflammation index (platelets count × neutrophils count)/lymphocytes count, SII), and system inflammation response index (monocyte count × neutrophils count)/lymphocytes count, SIRI), c-reactive protein (CRP), HDL=high-density lipoprotein; NHR=neutrophils/HDL; LHR=lymphocyte/HDL; MHR=monocytes/HDL; PHR=platelets/HDL.

All measurement data were presented as the mean ± SD. (standard deviation) when data satisfied the criteria for normality with p > 0.05. Since our data did not satisfy the criteria for normality and homogeneity of variance, statistical analysis between groups was performed using the rank sum test. For count data, chi-square test was used for data analysis. Statistical significance levels were considered to be p<0.05. All statistical analyses were carried out using SPSS, version 27.0 (SPSS Inc., Chicago, IL). To identify the best predictors of DN, univariate and multivariate analyses were computed. We choose the cut-off values for continuous variables by the maximum Youden index (sensitivity+specificity-1) in the ROC curve analysis. We used R language software to establish the nomogram prediction model for aDVT according to the multivariate logistic regression analysis results, and then we assessed the discriminative ability of the prediction model by AUC of ROC, and we evaluated the predicted and actual probabilities of this prediction model by calibration curve. The decision curve analysis is to evaluate the clinical application value of the prediction model.

Results

As shown in Tables 1 and 2, there were no significant differences in age, gender, BMI, injury mechanism, ischemia time, probe artery, injury hand, patients with a history of heart valve disease, heart failure, heart infarction, diabetes, or hypertension, PT, as well as the level INR, FIB, APTT, TT, LYM, RBC, PLT, NLR, PLR, HLD, LHR, and PHR between two groups by univariate analysis. However, we found that the level of D-dimer (p < 0.0001), WBC (p = 0.017), NEU (p = 0.025), MON (p = 0.002), MLR (p = 0.003), SII (p = 0.032), SIRI (p = 0.003), CRP (p < 0.0001), NHR (p = 0.019), and MHR (p = 0.001) were significantly higher in NG than in SG.

Table 1 Possible Factors May Be Associated with Digital Necrosis After Replantation in Two Groups

Table 2 Laboratory Examinations May Be Associated with Digital Necrosis After Replantation in Two Groups

Additionally, our findings showed that D-dimer [p = 0.012, OR = 1.859, 95% CI (1.145, 3.018)] and CRP [p = 0.005, OR = 1.009, 95% CI (1.003, 1.015)] were independent risk factors of DN by logistic regression analysis (Table 3). ROC curve analysis showed that the level of D-dimer (p < 0.0001, AUC area = 0.732, 95% CI (0.659,0.805)), and CRP (p < 0.0001, AUC area = 0.697, 95% CI (0.625,0.769)) were independent predictors of DN after replantation, and we identified that their cut-off values were 0.695 mg/L, and 30.985 mg/L, respectively (Table 4).

Table 3 Logistic Regression Analysis of Laboratory Variables Associated with Digital Necrosis After Replantation

Table 4 ROC Curve Analysis and Cut-Off Values Associated with Digital Necrosis After Replantation

As shown in Figure 1, we conduct a nomogram prediction model according to the results of logistic regression analysis. The ROC curve of the nomogram suggested good discrimination ability [AUC = 0.7538, 95% CI(0.6966, 0.7929)] and the calibration curve of the nomogram with Hosmer Lemeshow goodness-of-fit test showed well calibrated (p > 0.05, Figures 2 and 3). Decision curve analysis implied the nomogram prediction model with good clinical benefits (Figure 4).

Figure 1 The nomogram prediction model for digital necrosis in patients with replantation.

Figure 2 The receiver operating characteristic curves.

Figure 3 The calibration curve of nomogram.

Figure 4 The decision curve analysis of the nomogram.

Discussion

DN after replantation is a serious complication that not only affects the digital appearance but also their function. Therefore, it is crucial to have a deep understanding of its risk factors and take timely measures to address them. Yu¹¹ reviewed related studies and concluded that crush and avulsion injury, little finger, children and cold preservation were associated with DN after surgery, but ischemia time was not found to affect DN, which was different from the findings of Breahna⁴ that the timing of replantation and the ischemia time can impact the outcome of replantation. Wang⁸ conducted a comprehensive and in-depth analysis of 946 patients receiving digital replantation and first found that menstrual period, menopause, and the level of D-dimer were predictors of DN. Additionally, Wang⁸ also assessed the relationship between starting and finishing time of surgery and duration of surgery and DN and found that duration of surgery was identified as a predictor of DN.

In our study, the survival rate was 83.7% (278 of 332). Univariate analysis showed that the levels of D-dimer, WBC, NEU, MON, MLR, SII, SIRI, CRP, NHR, and MHR were related to DN. Furthermore, D-dimer and CRP were found to be the predictors for DN by logistic regression analysis, and we identified their cut-off values. The nomogram prediction model we performed has 0.7538 in the AUC of the prediction model, which implied good consistency in the correction curve and good clinical practicality by decision curve analysis.

Previous studies have reported that the survival rates of replantation to treat digital amputation ranged from 70% to 90%.^2,3 Wang⁸ concluded a retrospective study of 946 patients with 78.8% survival rate in our study. In this study, the survival rate was 83.7% after multiple digital replantation, which was similar to prior research.^2,3,8

D-dimer is widely used to evaluate venous thrombus in clinical practice.¹² Recently, it has also been a predictor for other diseases, such as heart failure, coronary syndrome, pneumonia and cancer.^3,5 Wang⁸ first reported the relationship D-dimer and DN and found that the level of D-dimer replantation played a crucial role in the assessment of DN. In our study, we found similar results showing that the level of D-dimer in FG was significantly higher than in SG, which was also demonstrated by logistic regression analysis. Above mentioned implied that patients with high D-dimer were more likely to suffer from replantation failure. As we know, replantation failure was mostly associated with the blockage of blood vessels in the hypercoagulable state resulting in microthrombus because D-dimer could originate from the formation and lysis of cross-linked fibrin, which can account for our and Wang’s findings. It is worth noting that Wang did not investigate the cut-off value of D-dimer to predictor DN after single digital replantation.⁸ Therefore, we used ROC curve analysis to identify 0.695 mg/L as the cut-off value of D-dimer to predictor DN.

Increasing literature has reported that inflammation contributed to the development of some traumatic disorders, such as ischemia-reperfusion injury.^9,10 Inflammation can cause a hypercoagulable state by activating platelets, neutrophils, and endothelial cells, as well as by modulating the expression and function of coagulation factors and inhibitors. Recently, NLR, PLR, MLR, SII, and SIRI indexes have been reported that these indicators played an important role in many diseases because they can suggest a balance of systemic inflammation.^13,14 To our knowledge, this is the first study assessing the relationship between inflammatory indicators and DN after digital replantation. In this study, the levels of D-dimer, WBC, NEU, MON, MLR, SII, SIRI, CRP, NHR, and MHR were found to be related to DN by univariate analysis. However, logistic regression analysis showed that CRP was an independent predictor of DN after replantation. CRP is a well-known biomarker for inflammation that binds to dead and damaged cells, which was used to screen various diseases, such as coronary heart disease and severe non-vascular diseases.¹⁵ Notably, CRP can improve the ability of the D-dimer and other fibrin degradation products to up-regulate interleukin-6 synthesis, which promotes CRP synthesis.¹⁶ We also used ROC curve analysis to identify the 30.985 mg/L as cut-off values of CRP.

According to the results of univariate and multivariate analyses in this study, we established a prediction model to evaluate DN after digital replantation. The ROC curve of the nomogram suggested good discrimination ability, and the calibration curve of the nomogram with Hosmer Lemeshow goodness-of-fit test showed well calibrated. Decision curve analysis showed that the nomogram prediction model has good clinical benefits.

Although our study offers some novel findings, it still has some limitations. First, this was a retrospective and single-center study, we need a large-scale cohort studies instead to validate the findings. Second, other factors that may related to necrosis, such as smoking history, were not fully included due to limited data in electronic medical records. Additionally, we did not perform external validation due to the limited sample size. Finally, although ten hand surgeons performed replantation in this study, they were hand surgeons with more than 5 years of experience.

In conclusion, we found that the level of D-dimer and CRP were closely related to DN by multivariate analysis, and we identified their cut-off value to predict DN. We also constructed a nomogram prediction model that can effectively predict DN in patients with multiple digital replantation. The purpose of this article is to share our experience in treating complete multiple digital amputation, which may help us take some preoperative measures to lower the incidence of DN.