Construction of a risk prediction model for postoperative cognitive dysfunction in colorectal cancer patients

Abstract

Background: Colorectal cancer (CRC) is one of the most prevalent and lethal malignant tumors worldwide. Currently, surgical intervention was the primary treatment modality for CRC. However, increasing studies have revealed that CRC patients may experience postoperative cognitive dysfunction (POCD).

Aim: To establish a risk prediction model for POCD in CRC patients and investigate the preventive value of dexmedetomidine (DEX).

Methods: A retrospective analysis was conducted on clinical data from 140 CRC patients who underwent surgery at the People's Hospital of Qian Nan from February 2020 to May 2024. Patients were allocated into a modeling group (n = 98) and a validation group (n = 42) in a 7:3 ratio. General clinical data were collected. Additionally, in the modeling group, patients who received DEX preoperatively were incorporated into the observation group (n = 54), while those who did not were placed in the control group (n = 44). The incidence of POCD was recorded for both cohorts. Data analysis was performed using statistical product and service solutions 20.0, with t-tests or χ ² tests employed for group comparisons based on the data type. Least absolute shrinkage and selection operator regression was applied to identify influencing factors and reduce the impact of multicollinear predictors among variables. Multivariate analysis was carried out using Logistic regression. Based on the identified risk factors, a risk prediction model for POCD in CRC patients was developed, and the predictive value of these risk factors was evaluated.

Results: Significant differences were observed between the cognitive dysfunction group and the non-cognitive dysfunction group in diabetes status, alcohol consumption, years of education, anesthesia duration, intraoperative blood loss, intraoperative hypoxemia, use of DEX during surgery, intraoperative use of vasoactive drugs, surgical time, systemic inflammatory response syndrome (SIRS) score (P < 0.05). Multivariate Logistic regression analysis identified that diabetes [odds ratio (OR) = 4.679, 95% confidence interval (CI) = 1.382-15.833], alcohol consumption (OR = 5.058, 95%CI: 1.255-20.380), intraoperative hypoxemia (OR = 4.697, 95%CI: 1.380-15.991), no use of DEX during surgery (OR = 3.931, 95%CI: 1.383-11.175), surgery duration ≥ 90 minutes (OR = 4.894, 95%CI: 1.377-17.394), and a SIRS score ≥ 3 (OR = 4.133, 95%CI: 1.323-12.907) were independent risk factors for POCD in CRC patients (P < 0.05). A risk prediction model for POCD was constructed using diabetes, alcohol consumption, intraoperative hypoxemia, non-use of DEX during surgery, surgery duration, and SIRS score as factors. A receiver operator characteristic curve analysis of these factors revealed the model's predictive sensitivity (88.56%), specificity (70.64%), and area under the curve (AUC) (AUC = 0.852, 95%CI: 0.773-0.919). The model was validated using 42 CRC patients who met the inclusion criteria, demonstrating sensitivity (80.77%), specificity (81.25%), and accuracy (80.95%), and AUC (0.805) in diagnosing cognitive impairment, with a 95%CI: 0.635-0.896.

Conclusion: Logistic regression analysis identified that diabetes, alcohol consumption, intraoperative hypoxemia, non-use of DEX during surgery, surgery duration, and SIRS score vigorously influenced the occurrence of POCD. The risk prediction model based on these factors demonstrated good predictive performance for POCD in CRC individuals. This study offers valuable insights for clinical practice and contributes to the prevention and management of POCD under CRC circumstances.

Keywords: Anesthesia; Cognitive dysfunction; Colorectal cancer; Dexmedetomidine; Postoperative; Preventive value; Risk prediction model.

Figure 1

Receiver operating characteristic curve. A: The risk prediction model for Receiver operating characteristic in colorectal cancer patients; B: Postoperative cognitive impairment in the validation group of colorectal cancer patients using a risk prediction model. AUC: Area under the curve; CI: Confidence interval.

Figure 2

Calibration curve. A: The predictive model for postoperative cognitive impairment in colorectal cancer patients in the modeling group; B: The predictive model for postoperative cognitive impairment in colorectal cancer patients in the validation group.

Figure 3

Decision curve analysis curve of the risk prediction model for predicting postoperative cognitive impairment in colorectal cancer patients in the modeling group.