Early clinical predictors of amputation in frostbite: injury depth and extent for admission risk stratification

Abstract

Background

Early identification of patients at high risk of amputation remains a major challenge in frostbite management, particularly in resource-limited settings where advanced imaging may not be readily available. We aimed to identify admission predictors of amputation in a predominantly migrant frostbite cohort and secondarily evaluated the prognostic value of the Prognostic Nutritional Index (PNI) and neutrophil-to-lymphocyte ratio (NLR).

Methods

We conducted a retrospective cohort study of consecutive patients admitted with frostbite to a tertiary referral hospital in eastern Turkey between January 2020 and December 2025. Demographic, clinical, and laboratory variables recorded at admission were compared between patients with and without amputation. Predictive performance was evaluated using receiver operating characteristic (ROC) analysis and multivariable logistic regression. Because frostbite degree demonstrated quasi-complete separation with the outcome, its independent effect was assessed using Firth penalized logistic regression.

Results

A total of 128 patients were included, of whom 126 (98.4%) were foreign nationals. Amputation was performed in 54 patients (42.2%). Frostbite degree was the strongest admission predictor of amputation, with 53 of 65 patients (81.5%) with third-degree frostbite requiring amputation compared with only 1 of 63 patients (1.6%) with first- or second-degree injuries (AUC 0.910, 95% CI 0.864–0.956). Frostbite extent (%TBSA) remained independently associated with amputation in multivariable analysis (adjusted OR 2.23, 95% CI 1.59–3.32; p < 0.001), and combining frostbite degree with %TBSA achieved the highest predictive performance (AUC 0.948, 95% CI 0.910–0.985). Although CRP, NLR, albumin, and PNI differed significantly between groups in univariable analyses, none remained independent predictors after adjustment for injury severity.

Conclusions

In patients with frostbite, admission risk of amputation is primarily determined by injury depth and extent. Simple bedside assessment of frostbite degree and %TBSA provides robust early risk stratification, whereas inflammatory and nutritional biomarkers offer limited additional prognostic value beyond clinical evaluation. These findings may facilitate early identification of patients at high risk of amputation, particularly in resource-limited emergency settings and regions managing migration-related cold injuries.

Introduction

Frostbite is a localized freezing injury resulting from exposure of tissues to subzero temperatures, leading to ice crystal formation, endothelial damage, microvascular thrombosis, and reperfusion injury following rewarming. Depending on the severity of tissue damage, frostbite may progress from reversible injury to irreversible necrosis requiring amputation. Despite advances in diagnosis and treatment, amputation remains one of the most serious complications of frostbite and is associated with substantial functional disability, psychological burden, and healthcare utilization [1, 2]. Contemporary studies from North America report amputation rates ranging from approximately 18% to 30% among hospitalized frostbite patients, although these rates vary according to injury severity and patient characteristics [3–5]. Historically, frostbite has been associated with military personnel, mountaineers, outdoor workers, and individuals experiencing homelessness. However, the epidemiology of frostbite has evolved considerably in recent years. Migration-related frostbite has emerged as an increasingly recognized clinical challenge in countries located along major migration routes. Severe cold injuries are frequrntly encountered among irregular migrants undertaking prolonged winter border crossings under harsh environmental conditions. These individuals are frequently exposed to prolonged cold, inadequate clothing, dehydration, exhaustion, and delayed access to medical care, factors that may contribute to more severe tissue injury and higher amputation risk. A previous study from eastern Turkey demonstrated that frostbite among irregular migrants was characterized by predominantly lower-extremity involvement and advanced injury severity at presentation [6]. Nevertheless, migrant populations remain underrepresented in the frostbite literature.

Early identification of patients at high risk for tissue loss is clinically important because several limb-salvage interventions appear to be highly time-dependent. Current treatment guidelines support the use of thrombolytic therapy and prostacyclin analogues such as iloprost in selected patients with severe frostbite, particularly when administered early after injury [7–9]. Consequently, reliable risk stratification at the time of admission may facilitate appropriate referral, resource allocation, and therapeutic decision-making. At present, prognosis is largely determined by clinical assessment of frostbite severity together with, where available, advanced imaging techniques such as technetium-99 m bone scintigraphy, angiography, and magnetic resonance angiography. Contemporary frostbite classification has evolved beyond the traditional depth-based four-degree clinical classification toward prognostically oriented systems, most notably the Cauchy classification, which incorporates the anatomical extent of injury and may be complemented by imaging to improve prognostic assessment and guide therapeutic decision-making.

More recently, the NATO Research Task Group emphasized that the choice of classification system should be explicitly reported and adapted to the available clinical and imaging resources [10–13]. However, these imaging modalities are not universally available, particularly in resource-limited settings and border regions where severe frostbite frequently occurs. Therefore, admission-based prognostic models relying on readily available clinical variables remain particularly valuable in resource-limited settings. Several studies have attempted to identify factors associated with adverse outcomes following frostbite. In a recent systematic review, Essien et al. [14] reported that delayed presentation, homelessness, alcohol use, psychiatric illness, substance abuse, and inadequate protection from cold exposure were among the most consistently reported predictors of frostbite-related complications and amputation [14]. Similarly, Schellenberg et al. identified physiological derangement and injury severity as important determinants of outcome in frostbite patients [15]. Nevertheless, most previously reported predictors are psychosocial, behavioral, or injury-related factors rather than objective laboratory measurements available at hospital admission. Validated admission-based prognostic models for predicting amputation remain scarce.

Inflammatory and nutritional biomarkers have recently attracted attention as potential predictors of adverse outcomes in ischemic and thermal injuries. The neutrophil-to-lymphocyte ratio (NLR), a readily available marker of systemic inflammation and physiological stress, has been associated with increased amputation risk and poorer outcomes in patients with diabetic foot disease and severe thermal injuries [16, 17]. Likewise, albumin-based nutritional indices have been linked to wound healing, tissue viability, and limb preservation outcomes [18]. The Prognostic Nutritional Index (PNI), originally described by Onodera et al., combines serum albumin concentration and peripheral lymphocyte count into a single indicator reflecting both nutritional and immunological status [19]. Although these biomarkers have shown prognostic value in several vascular and inflammatory conditions, their role in frostbite has not been adequately investigated. To our knowledge, neither NLR nor PNI has previously been evaluated as potential predictors of amputation in patients with frostbite injury.

Therefore, the present study aimed to identify admission characteristics associated with amputation in patients hospitalized for frostbite injury at a tertiary referral center located in a Turkish border province with a predominantly migrant patient population. In addition to established clinical indicators such as frostbite degree and injury extent, classified according to the conventional four-degree clinical classification routinely used in our institution, we evaluated the prognostic value of routinely available inflammatory and nutritional biomarkers, including NLR and PNI.

Methods

Study design and setting

This retrospective cohort study was conducted at a tertiary referral hospital located in eastern Turkey, serving a border region with a high volume of irregular migration. The hospital functions as the primary referral center for severe frostbite injuries occurring during winter border crossings. All patients admitted with a diagnosis of frostbite between January 1, 2020, and December 31, 2025, were screened for eligibility. The study was designed and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

Participants

All consecutive patients admitted to the emergency department with a diagnosis of frostbite during the study period were considered eligible. No age restriction was applied. Patients with incomplete demographic, clinical, or laboratory data required for analysis were excluded.

A total of 140 patients were identified during the study period. Twelve patients were excluded because of missing data, leaving 128 patients available for the final analysis. Patient records were reviewed individually to ensure that duplicate entries were not included.

Outcome definition

The primary outcome was frostbite-related amputation performed during the index hospitalization.

Amputation was defined as any surgical removal of frostbite-injured tissue, including digital amputations and major limb amputations. Patients were categorized into two groups according to amputation status: amputation and non-amputation.

Data collection and variables

Demographic, clinical, and laboratory data were extracted retrospectively from electronic hospital records.

Admission variables included age, sex, frostbite degree, frostbite extent expressed as percentage of total body surface area involved (%TBSA), injury location, number of involved extremities, and number of involved digits. Frostbite severity was classified according to the conventional clinical four-degree classification (first-, second-, third-, and fourth-degree frostbite) based on the depth of tissue injury documented in the medical records. Classification was assigned by the treating surgeon at admission and recorded in the electronic medical record.

Although more contemporary classification systems, particularly the Cauchy grading system, incorporate the anatomical extent of injury and may be complemented by radionuclide imaging to improve prognostic assessment and guide treatment decisions, the Cauchy classification was not routinely used in our clinical practice during the study period. Accordingly, patients were classified using the conventional four-degree clinical classification, which represented the standard approach at our institution [11–13]. Laboratory parameters recorded at admission included white blood cell count (WBC), C-reactive protein (CRP), creatinine, aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin, direct bilirubin, albumin, and lymphocyte count.

Two composite inflammatory–nutritional indices were calculated:

Neutrophil-to-lymphocyte ratio (NLR)

Prognostic Nutritional Index (PNI)

PNI was calculated according to the formula described by Onodera et al. [16]:

Variables occurring after admission, including secondary infection, length of hospital stay, number of surgical debridements, and number of amputated digits, were considered descriptive outcome variables and were not included as admission predictors because they are not available at the time of initial clinical assessment.

Clinical management

After admission, all patients underwent rapid clinical assessment, followed by rapid rewarming and evaluation of frostbite severity using the conventional four-degree clinical classification. Decisions regarding hospitalization, wound management, and medical therapy were based on the severity of tissue injury, extent of involvement, and the patient’s overall clinical condition. Patients were managed by a multidisciplinary team including general surgeons and emergency physicians according to the institutional frostbite management protocol. Surgical debridement and amputation were performed after clear demarcation of non-viable tissue unless urgent intervention was clinically required.

All patients were managed according to the institutional treatment protocol for frostbite. Standard supportive therapy included proton pump inhibitor prophylaxis, enoxaparin (0.6 mL), acetylsalicylic acid (100 mg/day), intravenous iloprost, pentoxifylline, and amlodipine. Additional treatments were administered according to clinical indications. Beta-blockers with vasodilatory properties were used in selected patients to manage iloprost-associated tachycardia and to promote peripheral vasodilation, whereas dexamethasone was administered in patients with severe edema.

According to the institutional protocol, empirical ceftriaxone therapy was routinely initiated at admission. Analgesic treatment consisted of ibuprofen and paracetamol, with tramadol administered as needed in cases of severe pain. No standardized thrombolytic protocol was available during the study period. Because all patients presented more than 24 h after cold exposure, none fulfilled the eligibility criteria for early thrombolytic therapy according to contemporary treatment recommendations. Therefore, management consisted primarily of supportive medical treatment, wound care, delayed assessment of tissue viability, and surgical intervention when clinically indicated.

Statistical analysis

Continuous variables were assessed for normality using the Shapiro–Wilk test. Because all continuous variables demonstrated non-normal distributions, results were presented as median and interquartile range (IQR). Comparisons between the amputation and non-amputation groups were performed using the Mann–Whitney U test.

Categorical variables were expressed as frequencies and percentages and were compared using Pearson’s chi-square test or Fisher’s exact test when appropriate.

Variables associated with amputation in univariate analyses (p < 0.05) were subsequently evaluated in multivariable analysis. Logistic regression was used to identify independent predictors of amputation. Results were reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs).

Because frostbite degree demonstrated quasi-complete separation with respect to the outcome, standard logistic regression estimates were unstable. Therefore, the independent effect of frostbite degree was assessed using Firth penalized logistic regression, which provides bias-reduced estimates in the presence of separation.

To avoid multicollinearity among inflammatory and nutritional variables, PNI was used as a composite marker rather than entering albumin and lymphocyte count simultaneously into the same multivariable model. Multicollinearity was assessed using variance inflation factors (VIFs).

The discriminative performance of individual variables and multivariable models was evaluated using receiver operating characteristic (ROC) curve analysis. Predictive accuracy was quantified using the area under the ROC curve (AUC). Differences between AUC values were assessed using the DeLong test. Optimal cut-off values were determined using the Youden index.

Calibration of the final model was assessed using the Hosmer–Lemeshow goodness-of-fit test and explained variation was quantified using Nagelkerke’s R². Internal validation was performed using bootstrap resampling with 1,000 iterations, and optimism-corrected AUC values were calculated.

All statistical analyses were performed using R software (version 4.6.0; R Foundation for Statistical Computing, Vienna, Austria). A two-sided p value < 0.05 was considered statistically significant.

Ethics approval

The study protocol was approved by the Non-Interventional Clinical Research Ethics Committee of University of Health Sciences Van Training and Research Hospital (Decision No: GOKAEK/2026-02/14; Date: 06 February 2026). Given the retrospective nature of the study, the requirement for informed consent was waived. All patient data were anonymized prior to analysis, and the study was conducted in accordance with the principles of the Declaration of Helsinki.

Results

Patient characteristics

A total of 140 patients with frostbite were screened for eligibility during the study period. Twelve patients were excluded because of incomplete clinical or laboratory data, leaving 128 patients for the final analysis (Fig. 1). The cohort was predominantly male (115/128, 89.8%) and almost entirely composed of foreign nationals (126/128, 98.4%). The median age was 26 years (IQR 21–32 years). Frostbite involved the feet only in 78 patients (60.9%), the hands only in 26 patients (20.3%), and all four extremities in 24 patients (18.8%). Lower-extremity involvement was present in 102 patients (79.7%).

Fig. 1.

STROBE flow diagram of patient selection and cohort formation. Of 140 patients screened for eligibility during the study period, 12 were excluded because of incomplete clinical or laboratory data. The final study cohort consisted of 128 patients, including 54 patients who underwent amputation and 74 patients who did not undergo amputation

Amputation occurred in 54 patients (42.2%), whereas 74 patients (57.8%) were managed without amputation. Among amputated patients, 53 underwent digital amputation and one patient underwent a major limb amputation. No in-hospital mortality was observed. All patients presented more than 24 h after cold exposure, and none were eligible for early thrombolytic therapy. Representative clinical photographs of severe frostbite injuries are shown in Fig. 2.

Fig. 2.

Representative clinical photographs of severe frostbite injury. (A) Extensive digital necrosis involving the toes. (B) Severe frostbite injury affecting multiple fingers with tissue demarcation and necrosis. (C) Bilateral foot involvement demonstrating progressive tissue loss and plantar necrosis. (D) Advanced frostbite with well-defined demarcation lines and dry gangrene of the toes. All patients presented more than 24 h after cold exposure, and the images illustrate advanced-stage frostbite injuries associated with delayed presentation

Comparison of admission characteristics according to amputation status

Admission characteristics according to amputation status are summarized in Table 1. Compared with patients who did not undergo amputation, amputated patients were older (median 28 vs. 25 years, p = 0.016) and had significantly greater frostbite extent (%TBSA) (median 4% vs. 2%, p < 0.001). They also had a higher number of involved digits (median 6 vs. 4, p = 0.005).

Table 1.

Baseline characteristics according to amputation status

Variable	No amputation (n = 74)	Amputation (n = 54)	p
Age (years)	25.0 [21.0–29.8]	28.0 [21.2–36.0]	0.016
Frostbite extent (%TBSA)	2.0 [2.0–2.0]	4.0 [3.0–4.0]	< 0.001
Involved extremities (n)	2.0 [2.0–2.0]	2.0 [2.0–2.0]	0.411
Involved digits (n)	4.0 [3.0–6.0]	6.0 [4.0–10.0]	0.005
WBC (×10³/µL)	10.0 [8.0–14.0]	12.0 [9.0–18.0]	0.024
CRP (mg/L)	11.0 [4.2–24.8]	27.0 [9.5–62.2]	< 0.001
Creatinine (mg/dL)	0.7 [0.6–0.8]	0.7 [0.6–0.8]	0.685
AST (U/L)	37.5 [24.0–62.0]	59.0 [31.8–115.5]	0.016
ALT (U/L)	23.5 [17.2–36.0]	39.5 [22.0–55.8]	0.005
Albumin (g/L)	39.0 [38.0–40.8]	37.0 [36.0–38.0]	< 0.001
Lymphocyte (×10³/µL)	2.0 [1.6–2.4]	1.9 [1.3–2.1]	0.066
NLR	3.0 [2.0–4.8]	4.0 [3.0–8.8]	0.001
PNI	48.1 [46.5–53.4]	45.8 [43.0–48.7]	< 0.001

Among laboratory parameters, amputated patients demonstrated significantly higher WBC counts (12.0 vs. 10.0 × 10³/µL, p = 0.024), CRP levels (27 vs. 11 mg/L, p < 0.001), AST levels (59.0 vs. 37.5 U/L, p = 0.016), ALT levels (39.5 vs. 23.5 U/L, p = 0.005), and NLR values (4.0 vs. 3.0, p = 0.001). In contrast, albumin (37 vs. 39 g/L, p < 0.001) and PNI values (45.8 vs. 48.1, p < 0.001) were significantly lower among amputated patients. No significant differences were observed for creatinine, total bilirubin, direct bilirubin, lymphocyte count, or number of involved extremities.

Association between frostbite degree and amputation

Frostbite degree demonstrated the strongest association with amputation. Among 65 patients with third-degree frostbite, 53 (81.5%) underwent amputation, whereas only one of 63 patients (1.6%) with first- or second-degree frostbite required amputation (p < 0.001).

Specifically, no amputations occurred among patients with first-degree frostbite, while only one amputation occurred among 62 patients with second-degree frostbite. In contrast, 53 of 65 patients with third-degree frostbite underwent amputation. The association between frostbite degree and amputation was highly significant (χ² = 83.8, p < 0.001) (Fig. 3). No patients were classified as fourth-degree frostbite.

Fig. 3.

Amputation rate according to frostbite degree. The proportion of patients undergoing amputation increased markedly with increasing frostbite degree. No amputations occurred among patients with first-degree frostbite, whereas only one amputation was observed among patients with second-degree frostbite. In contrast, 53 of 65 patients (81.5%) with third-degree frostbite underwent amputation

Univariate and multivariable logistic regression analyses

Univariate logistic regression analysis identified several admission variables associated with amputation, including frostbite extent (%TBSA), age, CRP, NLR, albumin, PNI, WBC count, lymphocyte count, AST level, and number of involved digits.

In the multivariable logistic regression model including age, frostbite extent (%TBSA), CRP, NLR, and PNI, only frostbite extent remained independently associated with amputation. Each 1% increase in affected body surface area was associated with a 2.23-fold increase in the odds of amputation (adjusted OR 2.23, 95% CI 1.59–3.32, p < 0.001). Results of the univariable and multivariable logistic regression analyses are presented in Table 2.

Table 2.

Univariable and multivariable predictors of amputation

Predictor	Crude OR (95% CI)	p	Adjusted OR (95% CI)	p
Frostbite extent (%TBSA)	2.49 (1.81–3.61)	< 0.001	2.23 (1.59–3.32)	< 0.001
PNI	0.87 (0.80–0.94)	0.001	0.94 (0.85–1.02)	0.173
NLR	1.15 (1.05–1.28)	0.007	1.06 (0.96–1.21)	0.315
CRP	1.03 (1.01–1.05)	< 0.001	1.02 (1.00–1.03)	0.096
Age	1.05 (1.01–1.09)	0.026	1.03 (0.98–1.09)	0.231

OR, odds ratio; CI, confidence interval

Age (adjusted OR 1.03, p = 0.231), CRP (adjusted OR 1.02, p = 0.096), NLR (adjusted OR 1.06, p = 0.315), and PNI (adjusted OR 0.94, p = 0.173) were no longer statistically significant after adjustment.

Because frostbite degree demonstrated quasi-complete separation with the outcome, its effect was evaluated using Firth penalized logistic regression. Frostbite degree remained strongly associated with amputation (OR 78.96, 95% CI 14.25–437.56, p < 0.001) (Table 3).

Table 3.

Association between frostbite degree and amputation using Firth penalized logistic regression

Variable	OR (95% CI)	p
Frostbite degree	78.96 (14.25–437.56)	< 0.001

OR, odds ratio; CI, confidence interval. Firth penalized logistic regression was used because quasi-complete separation was observed between frostbite degree and amputation status

Performance of admission variables

Frostbite degree was the strongest individual predictor of amputation, achieving an AUC of 0.910 (95% CI 0.864–0.956). As a simple bedside rule, third-degree frostbite predicted amputation with a sensitivity of 98.1%, specificity of 83.8%, negative predictive value of 98.4%, and overall accuracy of 89.8%.

Among the remaining admission variables, frostbite extent (%TBSA) demonstrated the highest discriminatory ability (AUC 0.812, 95% CI 0.742–0.882), followed by PNI (AUC 0.698), albumin (AUC 0.698), CRP (AUC 0.686), and NLR (AUC 0.669).

The multivariable admission-marker model consisting of %TBSA, age, CRP, NLR, and PNI achieved an apparent AUC of 0.862 (95% CI 0.799–0.926). Following bootstrap internal validation, the optimism-corrected AUC was 0.842. The predictive performance of this model did not significantly exceed that of frostbite degree alone (DeLong p = 0.175).

Combining frostbite degree and frostbite extent (%TBSA) yielded the highest discriminatory performance observed in the study, with an AUC of 0.948 (95% CI 0.910–0.985). Even among patients with third-degree frostbite, injury extent retained discriminatory value for amputation (AUC 0.705).

The multivariable admission-marker model demonstrated adequate calibration according to the Hosmer–Lemeshow goodness-of-fit test (χ² = 5.04, df = 8, p = 0.754). The model explained 46.7% of the variance in amputation status according to Nagelkerke’s R². Receiver-operating-characteristic curves for the principal admission predictors and the multivariable model are presented in Fig. 4 (Table 4).

Fig. 4.

Receiver-operating-characteristic curves for admission predictors. Receiver-operating-characteristic (ROC) curves comparing the discriminatory performance of admission predictors for amputation. Frostbite degree demonstrated the highest individual predictive performance (AUC 0.910), while the combination of frostbite degree and frostbite extent (%TBSA) achieved the best overall discrimination (AUC 0.948). NLR, CRP, albumin, and PNI showed comparatively lower predictive performance

Table 4.

ROC analysis of admission predictors

Predictor	AUC	95% CI	Cut-off	Sensitivity (%)	Specificity (%)
Frostbite degree	0.910	0.864–0.956	Third degree frostbite	98.1	83.8
Frostbite extent (%TBSA)	0.812	0.742–0.882	2.5%	75.9	77.0
PNI	0.698	0.606–0.791	46.3	59.3	77.0
Albumin	0.698	0.607–0.789	39.5	88.9	44.6
CRP	0.686	0.591–0.781	44.5	38.9	93.2
NLR	0.669	0.576–0.762	5.5	42.6	82.4
Multivariable model	0.862	0.799–0.926	0.47	72.2	86.5
Degree + TBSA	0.948	0.910–0.985	—	—	—

Secondary clinical outcomes

Secondary infection developed in 55 patients (43.0%). Patients who developed secondary infection experienced a significantly higher amputation rate compared with those without infection (52.7% vs. 34.2%, p = 0.036). However, because secondary infection occurs during hospitalization and is not available at the time of admission, it was not included in predictive modeling.

The median number of surgical debridements and the duration of hospitalization were also significantly greater among amputated patients, reflecting the increased complexity and severity of these injuries. However, these variables were considered consequences of injury severity rather than admission predictors and were therefore excluded from multivariable analyses.

Discussion

The principal finding of this study is that amputation risk in frostbite is determined predominantly by injury depth and extent. Among all evaluated admission variables, frostbite degree demonstrated the strongest discriminatory performance for predicting amputation, achieving an AUC of 0.910, while the combination of frostbite degree and frostbite extent (%TBSA) further improved predictive accuracy to an AUC of 0.948. In contrast, although inflammatory and nutritional biomarkers, including CRP, NLR, albumin, and PNI, differed significantly between amputated and non-amputated patients, none remained independently associated with amputation after adjustment for injury severity. These findings suggest that clinical assessment of injury depth and extent remains the cornerstone of prognostic evaluation in frostbite, whereas laboratory markers primarily reflect the biological consequences of severe tissue injury rather than independently determining tissue loss.

Frostbite depth has long been recognized as the principal determinant of tissue viability. Contemporary frostbite classification has evolved from the conventional four-degree clinical classification based primarily on tissue depth to more prognostically oriented systems such as the Cauchy classification, which integrates the anatomical extent of injury and may be supplemented by radionuclide imaging to guide treatment decisions and estimate tissue viability. More recently, the NATO Research Task Group emphasized that no single classification system is universally applicable and recommended that the selected classification should reflect the available clinical and imaging resources while being explicitly reported in research studies [11–13]. Our study used the conventional four-degree clinical classification because it represented the standard approach routinely employed at our institution during the study period. Despite its relative simplicity, this classification demonstrated excellent discrimination for amputation risk in our cohort, with frostbite degree emerging as the strongest admission predictor of tissue loss. Among patients with third-degree frostbite, more than 80% ultimately underwent amputation, whereas amputation was exceedingly rare among patients with first- or second-degree injuries. The large effect size observed in the Firth penalized regression model further emphasizes the dominant prognostic role of frostbite depth. These results are consistent with contemporary treatment guidelines, which prioritize injury depth when determining the need for advanced imaging, thrombolytic therapy, and specialist referral [7, 8, 13].

Although frostbite degree was the strongest predictor, our results also demonstrate that injury extent provides important additional prognostic information. Frostbite extent remained independently associated with amputation in multivariable analysis, with each 1% increase in affected body surface area associated with more than a twofold increase in the odds of amputation. Furthermore, combining frostbite degree and injury extent yielded the highest predictive performance observed in the study. This finding suggests that injury depth alone may not fully capture the biological burden of frostbite. Patients with similarly deep injuries may differ substantially in the volume of tissue affected, vascular compromise, and inflammatory response. Consequently, injury extent appears to refine risk assessment within groups of patients who already have severe frostbite. From a practical perspective, this observation supports a simple clinical approach in which injury depth serves as the primary determinant of risk, while extent provides additional stratification among patients with deep injuries. The concept that “depth dominates and extent refines risk” may represent a useful framework for bedside decision-making in frostbite management.

Several previous studies have attempted to identify factors associated with amputation following frostbite. However, most reported predictors have been psychosocial, environmental, or behavioral factors rather than objective clinical measurements. In their systematic review, Essien et al. [14], identified delayed presentation, homelessness, alcohol use disorder, psychiatric illness, substance abuse, and inadequate protection from cold exposure as the most consistently reported risk factors for adverse outcomes and amputation [14]. Similarly, Schellenberg et al. highlighted the importance of physiological derangement and severe tissue injury in determining outcome [15]. While these factors undoubtedly contribute to frostbite severity, many are not directly applicable to clinical risk stratification at hospital admission. Our study differs from previous reports by focusing specifically on objective admission variables that are readily available in routine clinical practice. The strong performance of frostbite degree and %TBSA suggests that simple bedside assessment may provide prognostic information comparable to or greater than that obtained from more complex laboratory evaluations. This distinction is particularly relevant in emergency settings, where admission decisions often rely on immediately available clinical information.

One of the most novel aspects of the present study is the evaluation of inflammatory and nutritional biomarkers, particularly NLR and PNI, in patients with frostbite. To our knowledge, no previous study has specifically examined the relationship between these markers and frostbite-related amputation. Elevated NLR has been associated with adverse outcomes in a variety of ischemic and inflammatory conditions, including diabetic foot disease, peripheral arterial disease, and severe burns [16, 17]. Likewise, low albumin concentrations and reduced PNI values have been linked to impaired wound healing, poor nutritional status, and increased amputation risk in vascular and chronic wound populations [18]. In the present study, amputated patients exhibited significantly higher NLR values and lower albumin and PNI values than non-amputated patients. These findings suggest that severe frostbite is accompanied by a measurable systemic inflammatory response and deterioration in nutritional–immunological status.

Despite these associations, NLR, albumin, and PNI did not remain significant predictors in multivariable analysis. Several explanations may account for this observation. First, these biomarkers are likely downstream manifestations of injury severity rather than causal determinants of tissue loss. Severe frostbite causes extensive tissue necrosis, endothelial injury, and inflammatory activation, all of which may elevate inflammatory markers and alter nutritional indices. Consequently, the relationship between these biomarkers and amputation may largely reflect their correlation with the underlying severity of tissue injury. Second, injury depth and extent may already capture most of the prognostic information relevant to tissue viability. Once these clinical variables are considered, laboratory markers contribute relatively little additional predictive value. Finally, laboratory measurements obtained at admission represent a single time point and may not fully reflect the dynamic inflammatory processes that occur during the subsequent course of injury. Collectively, these findings indicate that NLR and PNI should be viewed as markers of injury severity rather than independent determinants of amputation risk in frostbite.

The unique characteristics of our study population deserve particular attention. Nearly all patients were foreign nationals, and the vast majority sustained frostbite while attempting irregular border crossings. This population differs substantially from those described in most North American and European frostbite series, where homelessness, alcohol intoxication, psychiatric disease, and recreational exposure are more commonly reported [1, 3]. Migrants often experience prolonged environmental exposure, delayed rescue, dehydration, exhaustion, and limited access to healthcare resources. These factors may contribute to the high proportion of severe frostbite and the relatively elevated amputation rate observed in our cohort. Accordingly, our findings expand the existing frostbite literature by providing data from an underrepresented but increasingly important patient population. Given the continuing growth of migration-related cold injuries worldwide, understanding the prognostic characteristics of these patients is becoming increasingly relevant.

The clinical implications of our findings are straightforward and potentially valuable, particularly in resource-limited settings. Advanced imaging modalities such as technetium-99 m bone scintigraphy, magnetic resonance angiography, and conventional angiography can assist in determining tissue viability and guiding treatment decisions. However, these technologies may not be readily available in emergency departments, rural hospitals, or border regions. Our results suggest that simple bedside assessment can provide robust prognostic information. Patients presenting with third-degree frostbite and extensive tissue involvement should be considered at particularly high risk for amputation and may warrant early transfer to specialized centers, aggressive monitoring, and consideration of limb-salvage therapies where appropriate.

Conversely, the extremely low amputation rate among patients with first- and second-degree frostbite may help identify individuals who can be managed more conservatively. Several limitations should be acknowledged. First, the retrospective design introduces the possibility of selection bias and unmeasured confounding. In addition, although all patients were managed according to the institutional frostbite treatment protocol, the present study was not designed to evaluate the effectiveness of individual therapeutic interventions on amputation risk. Because treatment decisions were based on clinical judgment and injury severity rather than standardized allocation, treatment variables were not analyzed as predictors of outcome. Prospective studies with standardized treatment protocols are required to determine the independent effects of specific therapies on limb salvage. Second, this was a single-center study conducted in a unique migrant population, which may limit the generalizability of the findings to other settings. Third, contemporary prognostically oriented frostbite classification systems, such as the Cauchy classification, incorporate the anatomical extent of injury and may be complemented by radionuclide imaging to improve prognostic assessment and guide treatment decisions. The Cauchy classification was not routinely used at our institution during the study period; therefore, patients were classified using the conventional four-degree clinical classification. In addition, radionuclide imaging was not routinely available, and imaging-assisted prognostic assessment could not be evaluated. Consequently, the applicability of our findings to centers routinely using Cauchy grading and imaging-assisted prognostic assessment should be interpreted with appropriate caution [11–13]. An important characteristic of the present cohort was that all patients presented more than 24 h after cold exposure. Consequently, none of the patients were eligible for early thrombolytic therapy, and the findings of the present study primarily reflect outcomes among delayed-presenting frostbite patients. Despite the relatively high amputation rate, no in-hospital deaths occurred. These findings suggest that delayed presentation may substantially limit opportunities for limb salvage, whereas mortality appears to be low in this predominantly young population.

Despite these limitations, the study possesses several important strengths. It represents one of the largest contemporary frostbite cohorts reported from a migrant population, includes a substantial number of amputation events, and evaluates both traditional clinical variables and novel inflammatory–nutritional biomarkers. Furthermore, the use of ROC analysis, multivariable modeling, and Firth penalized regression allowed robust assessment of predictors despite the strong association between frostbite degree and outcome. External validation using cohorts in which Cauchy grading and imaging-based prognostic assessment are routinely applied would further strengthen the generalizability of the present findings.

Conclusion

In conclusion, amputation risk in frostbite is determined primarily by injury depth and extent. Frostbite degree is the most powerful admission predictor of tissue loss, while frostbite extent provides additional prognostic discrimination. Although inflammatory and nutritional biomarkers such as CRP, NLR, albumin, and PNI are associated with amputation in univariate analyses, they do not independently predict tissue loss after accounting for injury severity. These findings support the use of simple clinical assessment for early risk stratification and highlight the dominant prognostic importance of frostbite depth and extent in patients with severe cold injury. These findings provide a simple admission-based framework for early risk stratification in patients with frostbite, particularly in resource-limited settings where advanced imaging is unavailable.

Author contributions

Conceptualization: İbrahim Doğan, Sebahattin Çelik Methodology: İbrahim Doğan, Sebahattin Çelik Data Curation: İbrahim Doğan, Ali Rıza Karayıl, Sıdkı Arşad Narçin, Berk Topaloğlu, Ezgi Sönmez, Mehmet Kadir Bartın Formal Analysis: Serpil Sevimli Deniz, Sebahattin Çelik Investigation: İbrahim Doğan, Ali Rıza Karayıl, Serpil Sevimli Deniz, Sıdkı Arşad Narçin, Berk Topaloğlu, Serdar Çoban, Ezgi Sönmez, Mehmet Kadir Bartın, Sebahattin Çelik, Meltem Ceylan Delice Writing – Original Draft: İbrahim Doğan, Serdar Çoban Writing – Review & Editing: All authors Supervision: Sebahattin Çelik All authors read and approved the final manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical approval

The study protocol was approved by the Non-Interventional Clinical Research Ethics Committee of University of Health Sciences Van Training & Research Hospital (Decision No: GOKAEK/2026-02/14; Date: 06 February 2026). The requirement for informed consent was waived because of the retrospective study design.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.