Internal Validation of Preoperative Inflammation–Nutrition Markers for Predicting Complications After Complex Lower-Extremity Wound Reconstruction

Abstract

Background

Complications after complex lower-extremity wound reconstruction remain common in clinically heterogeneous adult patients, many of whom are older and multimorbid. We evaluated whether adding the C-reactive protein-to-albumin ratio and prognostic nutritional index to a prespecified clinical baseline model improves internally validated prediction of 30-day clinically significant reconstruction-related complications.

Methods

In a prospective single-center cohort from 2021 to 2024, consecutive adults undergoing complex lower-extremity reconstruction were enrolled. The C-reactive protein-to-albumin ratio and prognostic nutritional index were calculated from routine preoperative laboratories. A ridge-penalized logistic regression clinical baseline model was compared with an augmented model adding log-transformed C-reactive protein-to-albumin ratio and prognostic nutritional index. Internal validation used 5-fold cross-validation with out-of-fold predictions. Performance was assessed by area under the receiver operating characteristic curve (AUC) and Brier score; calibration and decision-curve analyses were interpreted descriptively from the internally validated predictions. Log-transformed C-reactive protein-to-albumin ratio was used as a monotonic skewness-reducing transformation rather than as a broader nonlinear biologic model.

Results

Among 693 patients (median age 70.7 years; 62.3% male), 139 (20.1%) met the composite endpoint. Infection, necrosis, and additional unplanned interventions occurred in 12.1%, 7.6%, and 10.2%, respectively (non–mutually exclusive). The C-reactive protein-to-albumin ratio was higher (median 0.55 vs 0.19) and the prognostic nutritional index lower (43.2±4.2 vs 46.1±4.5) in patients with complications. With the augmented model, higher log-transformed C-reactive protein-to-albumin ratio and lower prognostic nutritional index shifted predicted risk in the expected direction. The augmented model improved discrimination from AUC 0.71 (0.65–0.75) to 0.79 (0.75–0.83), reduced Brier score (0.132 vs 0.147), and showed higher net benefit than the baseline model across an approximate 10–35% threshold range in exploratory decision-curve analysis.

Conclusion

Routine preoperative C-reactive protein-to-albumin ratio and prognostic nutritional index may provide potentially useful incremental prognostic information for early complications after complex lower-extremity wound reconstruction, should be interpreted alongside clinical judgment, and external validation is warranted.

Introduction

Complex lower-extremity wounds often affect older and multimorbid adults and present a significant clinical challenge because multimorbidity is common, including diabetes, peripheral artery disease, and chronic venous disease, all of which can impair tissue perfusion and immune defense and increase the likelihood of infection and tissue loss.^1,2 These wounds, including venous, arterial, and neuropathic ulcers, are associated with substantial morbidity, including necrosis, reoperation, prolonged hospitalization, and in some cases major amputation.^1,3 Age-related physiological changes further compromise healing and increase vulnerability to complications.² Although contemporary reconstructive strategies, including free flaps, regional flaps, and adjunctive technologies, can improve limb salvage, early complications remain frequent and clinically important.^4,5 Interdisciplinary care integrating trauma/orthopedic, vascular, plastic surgery, and geriatric principles is often required to optimize comorbidities, wound bed preparation, and perioperative planning.^4,6,7 However, pragmatic preoperative tools to identify adult patients at highest risk for early reconstruction-related complications are limited.

Wound healing may also be shaped by systemic factors such as inflammaging, sarcopenia, and protein-energy malnutrition, which may slow healing and reduce tolerance of surgical stress.^8–10 This relationship may be particularly relevant in lower-extremity reconstruction because distal perfusion limitations, neuropathy, contamination, exposed hardware or bone, and weight-bearing demands can narrow the margin for successful flap or graft healing.^4–7 Objective, scalable markers derived from routine laboratory tests could therefore improve perioperative risk stratification beyond standard clinical variables. The C-reactive protein-to-albumin ratio (CAR) integrates systemic inflammatory burden with protein reserve, while the prognostic nutritional index (PNI) combines nutritional and immune competence using albumin and lymphocyte count.¹¹ Both indices are inexpensive and widely available and have been associated with postoperative complications in other surgical contexts, including wound-related outcomes.¹¹ However, prior evidence has largely come from other surgical or mixed reconstructive settings, and lower-extremity wound reconstruction-specific prediction evidence remains limited.^11–14

We aimed to determine whether a simple combined assessment of systemic inflammation and immunonutritional reserve, using CAR and PNI derived from routine preoperative laboratory testing, improves prediction of early clinically significant complications after complex lower-extremity wound reconstruction beyond what can be achieved with standard clinical variables alone. We hypothesized that adding log-transformed CAR and PNI to a prespecified clinical baseline model would yield clinically meaningful gains in discrimination, and overall internal predictive performance for a pragmatic 30-day composite endpoint.

Methods

Study Design and Reporting

We conducted a non-interventional, single-center prospective observational cohort study at the Affiliated Taihe Hospital of Hubei University of Medicine to evaluate whether routine preoperative inflammation–nutrition markers improve prediction of early postoperative complications after complex lower-extremity wound reconstruction. The protocol conformed to the Declaration of Helsinki and was approved by the ethics committee of the Affiliated Taihe Hospital of Hubei University of Medicine. All participants provided written informed consent prior to any study procedures. The study was reported in accordance with the STROBE guideline, the RECORD guideline for routinely collected health data, and applicable TRIPOD reporting items for prediction model studies.^15–17

Setting and Participants

Consecutive adult patients undergoing complex lower-extremity wound reconstruction between January 2021 and December 2024 were screened for inclusion. Complex wounds were defined as lower-extremity soft-tissue defects not amenable to primary closure after definitive debridement and requiring reconstructive coverage, including defects with exposed bone, tendon, hardware, or neurovascular structures and/or substantial contamination.

Patients were included if preoperative serum C-reactive protein (CRP), serum albumin, and absolute lymphocyte count were available within a prespecified preoperative window (closest measurement within 7 days before surgery) and if 30-day outcome ascertainment was available. Patients lacking laboratory values needed to compute CAR and PNI or lacking follow-up for the primary endpoint were excluded.

Baseline Clinical Assessment and Covariates

Baseline variables were collected prospectively from standardized clinical records and operative notes, including age, sex, body mass index (BMI), diabetes mellitus, peripheral artery disease (PAD, including chronic limb-threatening ischemia [CLTI]), chronic kidney disease (CKD), current smoking status, American Society of Anesthesiologists (ASA) physical status class, wound etiology, and reconstruction type.

Diabetes and CKD were defined based on a documented diagnosis and/or ongoing disease-specific treatment at the time of surgery. PAD (including CLTI, the most severe PAD phenotype) was recorded as a comorbidity based on a documented diagnosis supported by clinical evaluation and objective vascular testing and/or imaging when clinically indicated. Wound etiology was categorized a priori as diabetic/ischemic, trauma, infection-related, or other based on the dominant presenting phenotype documented by the treating team. These categories were pragmatic case-mix descriptors and were not intended as mutually exclusive etiologic adjudications. Reconstruction type was classified as free flap, pedicled/local flap, or graft/other based on the definitive reconstructive procedure.

Laboratory Measurements and Biomarker Indices

Preoperative CRP, albumin, and absolute lymphocyte count were obtained from routine clinical laboratory testing using standardized automated assays. CAR and PNI were calculated from the same preoperative blood draw. When multiple eligible preoperative draws existed, the sample closest to surgery within 7 days was used.

CAR was calculated as CRP (mg/L) divided by albumin (g/L). Because CAR is right-skewed, it was natural log-transformed for modeling (log(CAR)) to improve linearity on the log-odds scale.

PNI was calculated as albumin (g/L) + 5×lymphocyte count (×10⁹/L), which is algebraically equivalent to the conventional Onodera formulation (10×albumin [g/dL] + 0.005×lymphocyte count [/mm³]). This formula follows the Onodera prognostic nutritional index convention; the unit conversion is shown to make the calculation transparent.¹⁸

Outcome Definitions and Follow-Up

The primary endpoint was a 30-day pragmatic composite of clinically significant reconstruction-related complications, defined as the occurrence of any of the following within 30 days after surgery:

1. Postoperative wound infection (including surgical site infection) diagnosed clinically and requiring initiation or escalation of therapeutic systemic antibiotics beyond routine prophylaxis and/or procedural source control (eg, drainage or debridement);

2. Partial or total necrosis of reconstructed tissues requiring intervention (eg, debridement, revision, or re-coverage);

3. Additional postoperative surgical intervention to address reconstruction-related complications, including an unplanned return to the operating room.

Component complications were not treated as mutually exclusive because they can co-occur in the same patient. Outcomes were ascertained through review of inpatient notes, operative records, antibiotic and procedure documentation, and post-discharge encounters within the 30-day window, including readmissions when applicable.

Statistical Analysis

Baseline characteristics were summarized overall and by primary endpoint status using means with standard deviations (SD) or medians with interquartile ranges (IQR) for continuous variables and counts with percentages for categorical variables. Between-group comparisons were used descriptively (not for model selection): Welch’s t test for approximately normally distributed variables, the Mann–Whitney U-test for skewed variables, and χ² tests for categorical variables. False discovery rate (FDR) control across baseline comparisons used the Benjamini–Hochberg procedure.

Prognostic modeling used ridge-penalized logistic regression to estimate the probability of the 30-day composite endpoint and to improve coefficient stability in the presence of correlated predictors, particularly because albumin contributes to both CAR and PNI. A prespecified clinical baseline model included age, sex, BMI, diabetes, PAD (including CLTI), CKD, current smoking, ASA class (operationalized as ASA IV vs ASA II–III for parsimony), wound etiology (infection-related vs others), and reconstruction type (free flap vs others). An augmented model added log(CAR) and PNI.

Internal validation of predictive performance was based on out-of-fold predictions from 5-fold cross-validation. Five folds were chosen a priori to preserve a meaningful validation sample in each fold while maintaining approximately 28 events per held-out fold. To avoid information leakage, preprocessing (centering/scaling) and ridge penalty tuning were performed using training data within each fold, and the fitted model was applied to the held-out fold to generate out-of-fold predictions. The ridge penalty was tuned within the training data by inner cross-validation over a logarithmic grid of inverse regularization strengths (C=10⁻³ to 10³), minimizing mean log loss. Analyses were performed in Python 3.11 using scikit-learn for ridge logistic regression and cross-validation, with pandas, NumPy, SciPy/statsmodels, and matplotlib for data management, descriptive statistics, and figures. Discrimination was summarized by AUC, calibration by risk-decile calibration plots, overall accuracy by Brier score, and clinical utility by decision curve analysis. Confidence intervals for AUCs and ΔAUC were obtained by bootstrap resampling applied to the cross-validated predictions.

Because penalized regression does not naturally support conventional Wald-based hypothesis testing, adjusted odds ratios and 95% confidence intervals were reported as descriptive standardized effect-size summaries. Confidence intervals were generated using bootstrap percentile intervals from refitting the complete modeling workflow to bootstrap samples. We therefore focus on direction, magnitude, and predictive performance rather than coefficient-level p values.

Missingness was summarized by variable before and after eligibility exclusions. In the screened cohort, 87 patients were excluded: 61 had incomplete required prospective laboratory data within 7 days (missing C-reactive protein, n=24; missing albumin, n=18; missing lymphocyte count, n=11; available laboratory testing >7 days preoperatively, n=8) and 26 lacked 30-day outcome ascertainment. In the final analysis cohort, all candidate clinical predictors and required biomarker components were complete, so complete-case modeling used all 693 included patients.

Results

Study Population and 30-Day Complication Rates

A total of 693 adults undergoing complex lower-extremity wound reconstruction between January 2021 and December 2024 were included in the analysis (Figure 1). After eligibility exclusions, no candidate modeling variable had missing data in the analysis cohort. The cohort had a median age of 70.7 years (IQR 62.7–78.6), and 432 patients were male (62.3%) (Table 1). The 30-day composite endpoint occurred in 139 patients (20.1%), while 554 patients (79.9%) did not meet the composite endpoint (Figure 1).

Figure 1.

Study flow diagram for the prospective cohort of adults undergoing complex lower-extremity wound reconstruction.

Table 1.

Baseline Characteristics by 30-Day Composite Endpoint

Variable	No Complication (n=554)	Complication (n=139)	p value	FDR q
Age (years)	69.9 [61.6, 77.7]	73.8 [66.8, 80.7]	<0.001	<0.001
Male	350 (63.2%)	82 (59.0%)	0.360	0.360
BMI (kg/m2)	24.3 ± 3.4	24.1 ± 3.5	0.550	0.754
Diabetes mellitus	181 (32.7%)	68 (48.9%)	<0.001	<0.001
Peripheral arterial disease/CLTI	150 (27.1%)	60 (43.2%)	<0.001	<0.001
Chronic kidney disease	55 (9.9%)	16 (11.5%)	0.613	0.764
Current smoking	148 (26.7%)	45 (32.4%)	0.220	0.264
ASA class	—	—	0.029	0.032
ASA II	162 (29.2%)	27 (19.4%)	—	—
ASA III	316 (57.0%)	84 (60.4%)	—	—
ASA IV	76 (13.7%)	28 (20.1%)	—	—
Wound etiology	—	—	0.024	0.037
Diabetic/ischemic	241 (43.5%)	44 (31.7%)	—	—
Trauma	135 (24.4%)	36 (25.9%)	—	—
Infection-related	94 (17.0%)	37 (26.6%)	—	—
Other	84 (15.2%)	22 (15.8%)	—	—
Reconstruction type	—	—	0.147	0.360
Free flap	233 (42.1%)	68 (48.9%)	—	—
Pedicled/local flap	196 (35.4%)	37 (26.6%)	—	—
Graft/other	125 (22.6%)	34 (24.5%)	—	—
CRP (mg/L)	7.36 [2.83, 18.95]	19.28 [7.86, 43.70]	<0.001	<0.001
Albumin (g/L)	37.9 ± 3.6	35.8 ± 3.5	<0.001	<0.001
Lymphocytes (×109/L)	1.6 ± 0.3	1.5 ± 0.3	<0.001	<0.001
CAR (CRP/albumin)	0.19 [0.07, 0.50]	0.55 [0.23, 1.24]	<0.001	<0.001
PNI	46.1 ± 4.5	43.2 ± 4.2	<0.001	<0.001

Within 30 days, postoperative infection occurred in 84 patients (12.1%), necrosis requiring intervention occurred in 53 patients (7.6%), and additional postoperative surgical intervention to address complications occurred in 71 patients (10.2%).

Baseline Characteristics by Complication Status

Compared with patients without complications, those who experienced the 30-day composite endpoint were older (median 73.8 vs 69.9 years; Table 1) and had higher prevalences of diabetes mellitus (48.9% vs 32.7%) and PAD (including CLTI) (43.2% vs 27.1%) (Table 1). ASA class distribution also differed, with a higher proportion of ASA IV among patients with complications (20.1% vs 13.7%) (Table 1). Wound etiology differed between groups, driven by a higher proportion of infection-related wounds among patients with complications (26.6% vs 17.0%) (Table 1).

Preoperative Inflammatory and Immunonutritional Markers

Preoperative inflammatory burden and immunonutritional reserve differed by outcome status (Table 1). Patients with complications had higher CRP (median 19.28 vs 7.36 mg/L) and lower albumin (35.8±3.5 vs 37.9±3.6 g/L) and lymphocyte counts (1.5±0.3 vs 1.6±0.3 ×10⁹/L) (Table 1). Consistent with these component shifts, CAR was higher in the complication group (median 0.55 [IQR 0.23–1.24] vs 0.19 [0.07–0.50]) and PNI was lower (43.2±4.2 vs 46.1±4.5) (Table 1).

Multivariable Ridge-Penalized Prognostic Modeling

In ridge-penalized logistic regression adjusted for prespecified clinical predictors, both biomarkers contributed materially to prediction of the 30-day composite endpoint (Table 2 and Figure 2). Within the augmented model, higher log(CAR) and lower PNI shifted predicted risk in the expected direction (Table 2 and Figure 2). Among clinical predictors, larger effect estimates in the augmented model were observed for age, diabetes, PAD (including CLTI), and infection-related wound etiology (Table 2).

Table 2.

Multivariable Prognostic Models (Ridge Logistic Regression)

Predictor	Baseline Model OR (95% CI)	Augmented Model OR (95% CI)
Age (per SD)	1.41 (1.23–1.64)	1.29 (1.14–1.44)
Male	1.10 (0.88–1.40)	1.05 (0.94–1.19)
BMI (per SD)	1.13 (0.96–1.35)	1.13 (0.99–1.28)
Diabetes mellitus	1.51 (1.23–1.88)	1.28 (1.15–1.44)
Peripheral arterial disease/CLTI	1.80 (1.47–2.17)	1.43 (1.25–1.61)
Chronic kidney disease	1.14 (0.93–1.40)	1.07 (0.98–1.18)
Current smoking	1.14 (0.93–1.40)	1.07 (0.97–1.21)
ASA class IV (vs II–III)	1.17 (0.94–1.47)	1.09 (0.98–1.23)
Infection-related etiology (vs others)	1.35 (1.09–1.68)	1.19 (1.06–1.34)
Free flap reconstruction (vs others)	1.12 (0.89–1.39)	1.07 (0.93–1.20)
log(CAR) (per SD)	—	1.49 (1.33–1.69)
PNI (per SD)	—	0.74 (0.65–0.83)

Notes: Odd ratios are shown per 1 SD increase for continuous predictors where applicable. Confidence intervals are bootstrap percentile intervals from refitting the complete ridge-modeling workflow.

Figure 2.

Adjusted associations of preoperative biomarkers with the 30-day composite complication endpoint from the augmented prognostic model. Odds ratios are shown per 1 SD increase in log(CAR) and PNI with 95% CIs. The x-axis is odds ratio on a logarithmic scale, and the y-axis identifies the biomarker index.

Model Discrimination, Calibration, and Decision-Curve Utility

Using out-of-fold predictions from internal cross-validation, the clinical baseline model achieved an AUC of 0.71 (95% CI 0.65–0.75), and adding CAR and PNI improved discrimination to an AUC of 0.79 (95% CI 0.75–0.83), corresponding to an absolute ΔAUC of 0.08 (95% CI 0.04–0.12; p<0.001) (Table 3 and Figure 3). Overall predictive accuracy improved with biomarker inclusion, with the Brier score decreasing from 0.147 for the baseline model to 0.132 for the augmented model (Table 3).

Table 3.

Model Performance and Clinical Utility

Model	AUC (95% CI)	Brier Score	ΔAUC vs Baseline (95% CI)	p value	DCA: Threshold Range with Higher Net Benefit
Clinical baseline	0.71 (0.65–0.75)	0.147	—	—	—
Baseline + CAR + PNI	0.79 (0.75–0.83)	0.132	0.08 (0.04–0.12)	<0.001	0.10–0.35

Figure 3.

Receiver operating characteristic (ROC) curves comparing discrimination of the clinical baseline model versus the augmented model (clinical predictors + CAR + PNI). The green dashed diagonal line denotes the no-discrimination reference line (AUC=0.50).

Calibration assessed by risk-decile plots showed that observed event rates increased across deciles of predicted risk for the augmented model, with slight overestimation around the fifth decile and underestimation around the sixth decile, whereas the highest-risk deciles tracked the ideal line reasonably closely with slight overestimation at the top decile (Figure 4). In decision curve analysis, the augmented CAR+PNI model provided higher net benefit than the clinical baseline model and default treat-all/treat-none strategies across clinically relevant threshold probabilities, particularly in the approximate 0.10–0.35 range (Figure 5).

Figure 4.

Calibration plot for the augmented model using deciles of predicted risk.

Figure 5.

Decision curve analysis (DCA) comparing net benefit of the baseline and augmented models across clinically relevant threshold probabilities. Treat-all and treat-none strategies are shown for reference.

Discussion

In this prospective cohort of adults undergoing complex lower-extremity wound reconstruction, preoperative inflammation–nutrition markers improved internal predictive performance when added to prespecified clinical predictors. Importantly, these findings arise from a single-center model-development study with internal cross-validation only. Together, they suggest that routinely available CRP, albumin, and lymphocyte information may complement pragmatic preoperative risk stratification in this older-skewed adult cohort, but they do not establish that the specific CAR+PNI parameterization is uniquely optimal or superior to models based on the individual CRP, albumin, and lymphocyte components, or ready for broad clinical deployment.

In our cohort, early clinically significant reconstruction-related complications were frequent, including postoperative infection, necrosis requiring intervention, and additional unplanned surgical intervention. Consistent with a biologically plausible pathway in complex lower-extremity wounds such as diabetic foot ulcer (DFU) with diabetic foot infection (DFI) and chronic limb-threatening ischemia (CLTI) due to peripheral arterial disease (PAD), as well as contaminated traumatic wounds (eg, open tibial fractures) or chronic osteomyelitis, higher preoperative inflammation relative to protein reserve and lower immunonutritional reserve identified patients at elevated risk. Specifically, CAR was higher and PNI lower in those with complications, and both remained independently associated with the composite endpoint after clinical adjustment. Mechanistically, this pattern aligns with the concept that systemic inflammation coupled with diminished protein reserve can impair tissue repair, microvascular integrity, and infection control, while reduced lymphocyte reserve can weaken host defense against bacterial burden. In older adults, these associations are likely amplified by inflammaging and immunosenescence, which contribute to chronic low-grade inflammatory signaling and catabolic stress,¹⁹ and overlap clinically with the frailty phenotype even when frailty is not directly measured.²⁰ Meta-analytic evidence also supports a robust cross-sectional relationship between frailty/pre-frailty and higher inflammatory biomarkers including CRP.²¹ Importantly, contemporary nutrition frameworks emphasize that albumin should not be interpreted as a stand-alone malnutrition diagnosis. Rather, it reflects illness/inflammatory physiology and must be contextualized within etiologic and phenotypic criteria,²² supporting our use of CAR/PNI as pragmatic risk markers rather than diagnostic labels. Finally, the immune component of PNI is supported by surgical evidence showing that preoperative lymphopenia is associated with higher postoperative complications and infections across procedures.²³

Our findings align with and extend prior work evaluating inflammation–nutrition markers in reconstructive surgery. PNI has previously been linked to microvascular outcomes in reconstructive surgery, including evidence that lower PNI predicts extremity free flap failure,¹² and related composite nutrition indices have shown discrimination for flap complications in microvascular reconstruction cohorts.¹³ Similarly, CAR has been associated with adverse outcomes in plastic and reconstructive surgical populations, supporting its role as an integrative marker of systemic stress and reserve.¹⁴ However, much of this literature focuses on heterogeneous reconstructive indications, oncologic reconstruction, or longer-term outcomes, whereas our study specifically addresses complex lower-extremity wound reconstruction with clinically relevant, early endpoints in an older, multimorbid population. This focus is clinically important because DFU/DFI and ischemic limb pathology dominate real-world geriatric limb salvage, and inflammatory activity varies widely across these phenotypes. Notably, recent diabetic foot research suggests that CRP correlates with DFI severity and may exhibit non-linear behavior across CRP ranges,²⁴ which supports the rationale for ratio-based and multicomponent markers that contextualize inflammation against reserve. In this setting, the joint CAR+PNI approach delivered measurable incremental value beyond clinical factors alone, improving discrimination, overall accuracy, and decision-analytic utility, suggesting that these routinely available labs can meaningfully shift preoperative risk estimates rather than simply recapitulate comorbidity burden.

Clinically, these findings support a pragmatic framework for preoperative stratification in older adults undergoing reconstruction for DFU/DFI, CLTI/PAD ulcers, traumatic defects, and osteomyelitis. First, higher-risk patients may justify intensified early infection surveillance and lower thresholds for escalation while remaining guideline-consistent. The IWGDF/IDSA guidance emphasizes that DFI diagnosis and severity staging are primarily clinical and should guide decisions on hospitalization, antibiotics, and source control.²⁵ CAR/PNI can complement this clinical classification by quantifying host vulnerability that is not fully captured by wound appearance alone. Second, for ischemic disease, CLTI guidelines stress structured staging and objective hemodynamic assessment with multidisciplinary planning,²⁶ and diabetes-specific PAD guidance reinforces systematic evaluation of perfusion and limb threat in patients with DFU and suspected PAD.²⁷ Integrating CAR/PNI may help identify older CLTI patients who require closer postoperative monitoring intensity and anticipatory planning for revision risk in addition to limb-based staging. Third, nutrition optimization pathways should be triggered by risk, not by albumin alone. ESPEN surgical nutrition guidance supports early identification of nutrition risk and timely intervention,²⁸ and GLIM provides a structured approach to diagnosing malnutrition using phenotypic and etiologic criteria,²² allowing CAR/PNI outliers to prompt formal assessment, dietitian involvement, and perioperative protein-energy optimization. Finally, individualized, explainable risk estimates may support shared decision-making by helping treating physicians align monitoring burden, operative intensity, and rehabilitation expectations with patients’ goals.^29,30

Nevertheless, important limitations should be acknowledged. The single-center setting and local practice patterns may limit generalizability to other institutions and case mixes. Residual confounding is possible because not all determinants of complications may be fully captured. CAR and PNI share albumin, creating collinearity that, while mitigated by ridge penalization, still warrants cautious interpretation of individual coefficients. Accordingly, the present data should be interpreted as supporting the combined inflammation-nutrition construct rather than providing that CAR and PNI outperform their individual components. The composite endpoint is heterogeneous and components can co-occur, which can obscure whether predictors are driven predominantly by infection, tissue loss, or reoperation pathways, and the absence of external validation precludes claims of transportability or readiness for broad implementation. Future work should prioritize multicenter external validation with recalibration across diverse patient populations with diabetic foot ulcer/diabetic foot infection, CLTI/PAD-related ulcers, traumatic defects, and osteomyelitis, and should integrate standardized geriatric vulnerability measures and component-level comparisons to test whether CAR/PNI adds value beyond individual laboratory components, frailty and limb-threat staging. Finally, prospective interventional or implementation studies are needed to determine whether CAR/PNI–guided pathways can meaningfully reduce complications and improve patient-centered outcomes in older adults undergoing limb-salvage reconstruction.

In conclusion, preoperative C-reactive protein-to-albumin ratio and prognostic nutritional index, derived from standard C-reactive protein, albumin, and lymphocyte measurements, may provide practical and objective incremental prognostic information when added to clinical predictors in patients undergoing complex lower-extremity wound reconstruction. In this cohort, adding CAR and PNI to clinical predictors improved discrimination and demonstrated potentially clinically relevant net benefit across plausible thresholds. Feasibility and use of these tests for individual patients should be confirmed by the treating physician and interpreted alongside clinical examination, perfusion assessment, infection evaluation, and patient goals. Multicenter external validation and implementation studies are needed before broad clinical adoption.

Funding Statement

Chronic Disease Management Research Project of National Health Commission Capacity Building and Continuing Education Center (GJJJMB202510010097); Graduate Education and Teaching Research Project of Hubei University of Medicine (YJ2025008); Teaching Research Project of the First Clinical College, Hubei University of Medicine (DYLC202502); College Students’ Innovation and Training Program Project of Hubei University of Medicine (X202510929087).

Data Sharing Statement

The study protocol, de-identified analytic dataset, and statistical programming code used for the main analyses are available from the corresponding author upon reasonable request. Raw patient-level data are not publicly deposited because they contain potentially identifiable clinical information. Any data sharing will be subject to institutional policies and ethics approval where applicable.

Ethics Approval and Consent to Participate

The protocol conformed to the Declaration of Helsinki and was approved by the ethics committee of the Affiliated Taihe Hospital of Hubei University of Medicine (XM2025034). All participants provided written informed consent prior to any study procedures.

Author Contributions

Zhengyi Hu: Conceptualization, Methodology, Validation, Writing – original draft.

Bitao Zhu: Data curation, Investigation, Methodology, Validation.

Xiaoxiao Wu: Data curation, Investigation, Validation, Writing – review and editing.

Xiaolong Zhang: Investigation, Resources.

Ke Wen: Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Visualization.

Qi Yang: Validation, Visualization, Writing – review and editing.

Xiabing Qin: Conceptualization, Funding acquisition, Project administration, Supervision, Writing – review and editing.

All authors took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors have no conflicts of interest to declare.