PerfSCORE brings several intraoperative measurements into a single assessment of perfusion quality. In this prospective observational study, Koçarslan and colleagues compared perfusion-related variables with near-infrared spectroscopy (NIRS) and serum S100B in adults undergoing open-heart surgery. The investigators reported that PerfSCORE was associated with early mortality, whereas S100B did not show a significant association with mortality or neurological complications. The findings are relevant to perfusion practice, but the small number of deaths and inconsistencies in the published report limit how confidently the model can be applied.
The study enrolled 60 adults at Antalya Training and Research Hospital in Turkey between January and June 2024. Mean age was 62.05 ± 9.60 years, and the cohort included 39 men and 21 women. Patients with chronic kidney disease, congestive heart failure, recent cerebrovascular accident, or recent myocardial infarction were excluded. The surgical case mix included 31 coronary artery bypass grafting operations, five mitral valve replacements, eight aortic valve replacements, one aortic procedure, and 15 combined operations. The abstract counts 73 procedures across these 60 patients, reflecting the inclusion of combined surgery.
Cerebral oxygen saturation was monitored bilaterally with NIRS. Measurements were recorded at the start of surgery, the beginning of cardiopulmonary bypass (CPB), the end of CPB, and the end of surgery. S100B was measured before surgery and at approximately six and 24 hours afterward using an ELISA assay. The authors describe CPB flow adjustment guided by NIRS, but the study was not a randomized comparison of NIRS-guided management versus another approach. Consequently, it cannot establish whether that management strategy itself improved patient outcomes.
PerfSCORE incorporated ten physiological and procedural variables: pH, base excess, arterial carbon dioxide tension, arterial oxygen tension, activated clotting time, fluid balance, donor blood transfusion, hematocrit, potassium, and CPB duration. Table 4 assigns each variable a score of zero, one, or two according to specified ranges. These are study scoring thresholds, not universal treatment targets. The reported mean PerfSCORE was 6.71 ± 2.30, with observed values ranging from two to 12.
The principal clinical outcomes included six early deaths and three early cerebrovascular accidents. Six deaths among 60 patients represent a 10% observed mortality rate. Table 1 notes that three of those who died had been discharged from hospital, but the report does not clearly define the precise early-mortality follow-up window. This uncertainty matters when comparing the findings with studies that use explicitly defined in-hospital or 30-day endpoints.
The NIRS–S100B relationship was generally weak in the detailed correlation table. For example, NIRS at the start of surgery correlated negatively with S100B at 24 hours (r = −0.282, p = 0.029). Several other relationships were not statistically significant. Table 3 does not support a consistent, strong relationship across the different measurement times. It also conflicts with a results paragraph describing moderate positive correlations of 0.42 and 0.47. The abstract’s description of weak correlations is more consistent with the detailed table.
Individual perfusion-related measures showed modest discrimination for mortality. PerfSCORE had an area under the receiver operating characteristic curve of 0.708, with a 95% confidence interval of 0.517–0.900 and p = 0.096. Hematocrit score had an AUC of 0.747 and p = 0.049. The AUCs for carbon dioxide score, pH score, and fluid balance were 0.724, 0.716, and 0.711, respectively. These results should be distinguished from the classification performance of the combined regression model.
In the multivariable logistic regression table, PerfSCORE had an odds ratio of 1.42 (95% CI 1.18–1.71; p < 0.001). This indicates a reported 42% increase in the odds of the modeled outcome per unit increase in PerfSCORE, not a 42-percentage-point increase in absolute mortality risk. Hematocrit score, fluid balance, and pH score also reached statistical significance, whereas carbon dioxide score did not. The model’s Nagelkerke pseudo-R² was 0.765; this is a model-fit statistic and should not be interpreted as a literal percentage of outcome variance explained.
Table 7 shows that the combined model correctly classified five of six deaths and 53 of 54 survivors. Those figures correspond to 83.3% sensitivity and 98.1% specificity in this dataset, with 58 of 60 patients classified correctly overall. Some narrative passages reverse the death and survivor percentages. The table provides the clearest basis for interpreting the reported performance, and these values describe classification within the study sample rather than accuracy established in an independent validation cohort.
Several methodological issues argue against treating the model as ready for routine prognostic use. The regression includes five predictors despite only six deaths, creating a substantial risk of overfitting and unstable estimates. Components of PerfSCORE also appear separately in the model, potentially introducing overlapping predictor information. The report does not establish performance in another hospital or patient population. Surgical heterogeneity, excluded high-risk groups, and possible differences in intraoperative management further limit generalizability.
The publication also contains a scoring inconsistency. Its methods describe risk bands spanning seven to 21 points, while Table 4 uses ten components scored zero to two and Table 1 reports values as low as two. That discrepancy should be clarified before attempting to reproduce the score or assign clinical risk categories. These limitations do not erase the observed associations, but they reduce confidence in the specific model estimates and in the claim of robust independent prediction.
For perfusionists, the useful message is that integrated assessment of bypass exposure and physiological conditions may be more informative than relying on an isolated biomarker. However, this study does not show that NIRS monitoring lacks value, nor does failure to detect an S100B association prove that the biomarker is never useful. Its results support further investigation of perfusion-based risk assessment with clearer reporting, adequate event numbers, prespecified modeling, and external validation. Until then, PerfSCORE should be viewed as a promising research approach rather than a validated replacement for established perioperative risk assessment.
Source: Koçarslan A, Onuk ZA, Kulaksızoğlu S, Özerdem G, Yılmaz H, Aydın MS. Nigerian Journal of Clinical Practice. 2026;29:982–989. DOI: 10.4103/njcp.njcp_86_26.
