We designed a prospective longitudinal observational study with a 28-day follow-up period. The study followed the ethical standards outlined in the Declaration of Helsinki and was approved by the local ethics committee of our university hospital. Consecutive sampling was performed among adults eligible for skin cancer excision on the face, scalp, or limbs, followed by closure with a full-thickness or split-thickness skin graft at our dermatologic surgery unit.

The exclusion criteria were axillary temperature greater than 37.5°C, inability to remain still during LSCI recording, grafts with donor or recipient sites involving the eyelids or nose, and postoperative bacterial graft infection or hematoma. These criteria were established to avoid measurement bias with LSCI, including bias caused by involuntary movement of underlying structures, such as the eyelids or nasal valve, and inability to assess the entire graft surface in a single recording, for example, in a pannasal graft.

The inclusion period was from April 2023 to January 2024. Anticipating a 10% dropout rate and accepting an α risk of 0.05 and a β risk of 0.20 in a 2-sided test, the required sample size for detecting a correlation coefficient of 0.4 was 52 participants.

The outcome was skin graft necrosis on day 28, defined as any loss of surface integrity, ranging from epidermolysis progressing to scab formation to full-thickness necrosis. Necrosis extent corresponded to the percentage of necrotic graft tissue. Clinically significant graft necrosis was defined as 20% or greater necrosis extent, as adopted by other study groups and implying the need for ongoing wound care.15,16 The studied factor was perfusion assessed by LSCI on the graft bed on day 0 and on the graft surface on day 7.

Participant characteristics included in the analysis, such as age, sex, history of arterial hypertension, dyslipidemia, or diabetes, current smoking, and use of anticoagulant or antiplatelet drugs, were collected during the preoperative evaluation. There were no changes to the reconstructive algorithms of our department or to standard postoperative wound care for this study.

Lesions were removed after infiltration with lidocaine 2% and epinephrine 1:100,000. The depth of dissection was the galea aponeurotica on the scalp, the superficial musculoaponeurotic system on the face, and the subcutaneous tissue or deep fascia on the limbs. Full-thickness skin grafts were harvested from the infraclavicular area for facial and scalp defects. Split-thickness skin grafts measuring 0.4mm were harvested from the inner arm for upper limb defects and from the anterior thigh for lower limb defects. A tie-over dressing was applied for 1 week, and a new dressing with petrolatum gauze was used thereafter.

During surgery, graft bed perfusion, expressed in arbitrary perfusion units (APU), was assessed with the PeriCam PSI NR® LSCI system (Perimed, Järfälla, Sweden) at least 15min after skin cancer removal. On day 7, at the first postoperative graft dressing change, graft perfusion was also evaluated. Each LSCI recording lasted 60s and was performed at a distance of 25cm from the skin. The studied area was immobilized with a vacuum pillow (AB Germa, Kristianstad, Sweden) to avoid movement artifacts.

To obtain a more physiological and comparable evaluation, perfusion in APU was divided by each participant's mean arterial pressure (MAP), yielding cutaneous vascular conductance (CVC).14 CVC reflects the ease with which blood flows through the skin microcirculation at a given pressure difference.17

Skin graft evolution was recorded using sequential clinical photographs, which included a surgical ruler placed close to the graft. Necrosis extent was calculated from the clinical photographs obtained on day 28 by dividing the necrotic area by the total graft area. These areas were delimited and calculated with SketchAndCalc™ version 6.3.0 because the clinical photographs contained a surgical ruler that served as a scale. Fig. 1 presents examples of clinical photographs and LSCI recordings on days 0, 7, and 28 for grafts on the face and scalp.

Fig. 1.

Clinical photographs and LSCI recordings of full-thickness skin grafts. Clinical photographs (upper panels) and LSCI recordings (lower panels) of full-thickness skin grafts on the face, without clinically significant necrosis, and scalp, with clinically significant necrosis. Upper and lower panels refer to the same evaluation point: graft bed (A, B and G, H), graft on day 7 (C, D and I, J), and graft on day 28 (E, F and K, L). The LSCI software generates blood perfusion images of the scanned area using a semiquantitative color scale, ranging from black, indicating no perfusion, to red, indicating intense perfusion. Perfusion is also provided in arbitrary perfusion units (APU), not shown. Cutaneous vascular conductance is obtained by normalizing APU to the patient's mean arterial pressure.

The sample underwent descriptive statistical analysis, and variable normality was assessed using the Shapiro–Wilk test. Correlation analysis was used to determine the relationship between perfusion and necrosis extent. Univariable linear regression was used to identify significant predictors of necrosis, and multivariable models incorporated statistically significant predictors.

Comparative analyses between groups, with and without necrosis extent of 20% or greater, were performed using the χ2 test, Fisher exact test, and Mann–Whitney U test, as appropriate. A binary logistic regression model was used to estimate the odds ratio for necrosis extent of 20% or greater. Receiver operating characteristic curves were generated to predict necrosis extent of 20% or greater and to determine the perfusion cutoff points with the highest accuracy using the Youden index. Positive and negative predictive values were calculated.

All analyses were conducted using IBM SPSS Statistics, version 26. Statistical significance was set at P<.05.

Fig. 2 presents scatterplots showing the distribution of necrosis extent on day 28 and perfusion assessed on the graft bed and graft surface, revealing nonlinear negative relationships. A curve estimation approach showed that the logarithmic model provided the best fit for these relationships. Subsequently, the CVC variable underwent logarithmic transformation, allowing linearization of the correlations and linear regression analysis. The following perfusion results are presented as the natural logarithm of CVC [ln(CVC)].

Fig. 2.

Relationship between perfusion and necrosis extent. Scatterplots of perfusion, measured as cutaneous vascular conductance (CVC), in arbitrary perfusion units (APU)/mm Hg, on the graft bed intraoperatively and on the graft on day 7, and necrosis extent on day 28.

Perfusion assessed on the graft bed was very strongly correlated with necrosis extent on day 28 (r=−0.86; P<.001). On day 7, the correlation between graft perfusion and necrosis extent was strong (r=−0.69; P<.001). These correlations differed significantly (z=2.2; P=.012).

Table 2 shows the univariable linear regression models based on perfusion. The highest R2 value, 0.75, was observed for graft bed perfusion, meaning that 75% of the variability in necrosis extent could be explained by this variable. Specifically, each unit increase in graft bed perfusion reduced necrosis extent by 41 percentage points. The predictive capacity for necrosis was lower on day 7, with an R2 value of 0.46, and each unit increase in graft perfusion reduced necrosis extent by 21 percentage points. Because the 95%CIs for the linear regression β coefficients did not overlap, the differences in the mean estimated effects were significant.

Table 3 describes the correlations between patient, defect, and graft characteristics and necrosis extent. Only age (r=−0.28), defect location on the face (r=−0.29), and defect location on the lower limb (r=0.35) were significantly correlated with necrosis extent. None of these variables explained more than 11% of the variation in necrosis extent (Table 4).

On day 0, adjustment of the model combining graft bed perfusion with age and defect location (F=41.0; P<.001) was worse than that of the model with graft bed perfusion alone (F=158.9; P<.001), despite the same R2 value (Table 5). On day 0, the best predictive model for necrosis extent can be expressed as follows: necrosis extent=0.12−0.41×ln(CVC graft bed).

A similar finding was observed for graft perfusion on day 7, with the univariable model showing better adjustment (F=46.8) than the multivariable model (F=13.0) (Table 5). On day 7, necrosis extent can be predicted by graft perfusion according to the following formula: necrosis extent=0.10−0.21×ln(CVC graft day 7).

Clinically significant necrosis, defined as necrosis extent of 20% or greater, developed in 18 of 55 grafts (32.7%); 12 of these grafts (66.7%) were located on the lower limb, and all were split-thickness grafts. Graft bed perfusion in the group with necrosis extent of 20% or greater was approximately half that measured in the group with necrosis extent less than 20% (0.55 vs 1.06APU/mm Hg; P<.001).

Binary logistic regression models were created using graft bed perfusion alone or combined with lower limb location, which had been identified as an independent predictor of clinically significant necrosis. To prevent redundancy, graft type was not included in the regression model because all split-thickness grafts with necrosis extent of 20% or greater corresponded to lower limb grafts. Both models were statistically significant, but the accuracy of the univariable model (90.9%) was higher than that of the multivariable model (67.3%), indicating that perfusion alone was a better predictor than perfusion combined with lower limb location.

For each 0.1APU/mm Hg increase in graft bed CVC, the odds of clinically significant necrosis decreased by 57% (OR, 0.43; 95%CI, 0.28–0.68). This interpretation was based on rescaling CVC, initially ranging from 0.65 to 1.73APU/mm Hg, by multiplying it by 10 to obtain a new range.

ROC curves assessing the predictive capacity for necrosis extent of 20% or greater based on perfusion on days 0 and 7 are presented in Fig. 3. The area under the curve was 0.92 (95%CI, 0.83–1.00) on day 0 and 0.81 (95%CI, 0.69–0.94) on day 7. The best perfusion threshold values for necrosis extent of 20% or greater were 0.64 and 0.48APU/mm Hg on days 0 and 7, respectively. The day 0 threshold showed the highest positive predictive value (100%) and negative predictive value (93%). On day 7, the positive predictive value was 68%, and the negative predictive value was 86%.

Fig. 3.

ROC curves for clinically significant necrosis. Receiver operating characteristic (ROC) curves for clinically significant necrosis based on day 0 graft bed perfusion and day 7 graft perfusion. The area under the curve on days 0 and 7 was 0.92 and 0.81, respectively.