Clinical and hematological factors affecting perioperative blood loss following total knee arthroplasty: A new clinical prediction model

As an effective treatment for end-stage knee osteoarthritis, total knee arthroplasty (TKA) has been extensively performed globally over the past years. Despite its effectiveness, patients who undergo TKA are at risk of a series of complications, from infection to thromboembolism,[1] and substantial blood loss, which remains a concern among orthopedic surgeons, causing increased possibility for blood transfusion, prolonged hospitalization, and higher cost.[2,3] Postoperative TKA blood loss is now partially reduced due to the progress of surgical techniques and adjuvant medications. And many efforts on preoperative estimation of postoperative blood loss have been made over the past decade. However, there are still no systemic ways to assess the independent predictive value of potentially correlated risk factors; therefore, it is clinically critical to identify patients at an increased risk of blood loss and establish an effective blood management strategy.[4] This study analyzed total blood loss in patients receiving primary TKA, aiming to (i) identify influence factors for blood loss after TKA, especially among hematological parameters; (ii) develop a model for individualized prediction of blood loss following TKA using readily available clinical variables; and (iii) propose potential management strategies for adverse outcomes. This study was approved by the Institutional Review Board of the Peking Union Medical College Hospital (No. S-K2005), and because this was a retrospective study, the patients' informed consent was waived. The flowchart for experimental design is shown in Supplementary Figure 1, https://links.lww.com/CM9/C355. For all 1587 patients, the age, gender, body mass index (BMI), tobacco use (a positive status was defined as current smoker within 1 year, 1+ pack/day), alcohol use (a positive status was defined as current drinker within 1 year, 1+ drink/day) were recorded.[5] In addition, perioperative laboratory parameters were recorded. Total blood loss was calculated using Rosencher et al's[6] formula: Total red blood cell (RBC) loss (mL) = [Uncompensated RBC loss (mL)] + [Compensated RBC loss (mL)] Uncompensated RBC loss (mL) = [Initial RBC (mL)] – [Final RBC (mL)] Compensated RBC loss = [Sum of RBCs received from the various sources of transfusion] Initial RBC = [Estimated blood volume (mL)] × [Initial Hematocrit (HCT) level (%)] at Day – 1 Final RBC (mL) = [Estimated blood volume (mL)] × [Final HCT level (%)] at Day + 3 The estimated blood volume in milliliters was calculated for women and men separately with the use of the formulas[7]: Women: [Body surface area (m2)] × 2430 Men: [Body surface area (m2)] × 2530 where body surface area was calculated as: 0.0235 × [height (cm)]0.42246 × [weight (kg)]0.51456 Finally, the total blood loss was obtained at a HCT level of 35%: Total blood loss (mL) = [Total RBC loss (mL)]/0.35 Analyses were performed using SPSS software version 22.0 (SPSS Inc, Chicago, USA) and RStudio version 1.1.461 (RStudio, Boston, USA). Multivariable stepwise linear regression were established to determine potential independent influence factors for blood loss. P <0.05 was considered statistically significant. Considering that when blood loss is over 1500 mL, patients' physical condition and prognosis will be greatly reduced, so we converted the dependent variables into binary categorical variables at the threshold of 1500. The association between the pre-specified predictors and the primary outcome was assessed using LASSO Cox regression model analysis. And we used the Cox regression coefficients to generate nomograms. The model was internally validated using 1000 bootstrap samples. At the same time, the model was externally validated using another cohort of patients from January 2021 to December 2023. Agreement between predicted and observed outcomes was also evaluated graphically using receiver operating characteristic (ROC) analysis and the area under curve (AUC). The calibration curve was plotted to evaluate the agreement between nomogram-derived probability and actual observations of the model. A decision curve analysis (DCA) was also performed to evaluate the clinical benefit of our model. In this study, the total blood loss on the third day after the TKA surgery reached 961.29 ± 489.40 mL. Male patients lost an average of 1218.33 ± 549.67 mL of blood, whereas female patients lost an average of 898.89 ± 452.53 mL. We conducted a multivariate analysis of demographics, clinical, surgical variables, peroperative hematological parameters. Results were summarized in Supplementary Table 1, https://links.lww.com/CM9/C355. For these variables showing a correlation with postoperative blood loss, they were used for LASSO Cox regression model analysis, and the results are presented in Figure 1 and Supplementary Table 2, https://links.lww.com/CM9/C355.Figure 1: Development and performance of the nomogram. (A) Tuning parameter (λ) selection in the LASSO model used 10-fold cross-validation via minimum criteria. The AUC curve was plotted versus log (λ). Dotted vertical lines were drawn at the optimal values by using the minimum criteria and the 1 standard error of the minimum criteria (the 1-SE criteria). A λ value of 0.01623, with log (λ), 10 was chosen (1-SE criteria) according to 10-fold cross-validation. (B) LASSO coefficient profiles of the 13 features. A coefficient profile plot was produced against the log (λ) sequence. Vertical line was drawn at the value selected using 10-fold cross-validation, where optimal λ resulted in 10 non-zero coefficients. (C) A nomogram was created to evaluate the possibility of postoperative hemorrhage (total blood loss >1500 mL). To use the nomogram, find the position of each variable on the corresponding axis, draw a line to the points axis, add the points for the ten variables to create a total point value, and draw a line to the possibility axis to determine the probability on the lower line of the nomogram. ALB: Albumin; APTT: Activated partial thromboplastin time; BMI: Body mass index; HCT: Hematocrit; MONO: Monocyte; PLT: Blood platelet; TBIL: Total bilirubin.These independently associated risk factors were used to form a blood loss risk estimation nomogram [Figure 1C]. The optimal cutoff value of the total nomogram scores was determined to be 200. The sensitivity, specificity, positive predictive value, and negative predictive value when used in predicting risk of postoperative hemorrhage were 81%, 65%, 24%, and 96%, respectively. The resulting model was internally validated using the bootstrap validation method. The nomogram demonstrated good accuracy in estimating the risk of blood loss, with an unadjusted C index of 0.78 and a bootstrap-corrected C index of 0.77. ROC analysis and AUC graphically showed good agreement on the presence between the risk estimation by the nomogram (the mean AUC of the nomogram was 0.784), with high sensitivity and specificity of 79% and 64%, respectively, which indicated good discrimination. The calibration curve suggested that the nomogram had good calibration and fit in the internal validation set. The DCA results demonstrated that the nomogram added more net benefit than the "treat all" strategy or the "treat none" strategy in the training cohort. Overall, 500 patients after TKA from January 2021 to December 2023 from another center were finally included for external validation of the nomogram. The area under the ROC curve remained stable in the external validation (0.774). The calibration plots of the nomogram showed high consistency between the ideal line and the observed survival probability in the external validation cohorts. In the external validation, the DCA also revealed that the nomogram had an overall net benefit. The results of intetnal and external validation are shown in Supplementary Figure 2, https://links.lww.com/CM9/C355. In this study, we developed a prediction model based on demographics and hematological characteristics to improve the prediction of risk of hemorrhage after surgery for patients with TKA. Our results showed that this model can predict patients with high risk of postoperative blood loss (>1500 mL). Few studies investigated the relationship between smoking and postoperative blood loss. Nordestgaard et al[8] reported two potential mechanisms of smoking's influence on surgical bleeding. The chemicals inside the tobacco increased the recruitment of inflammatory cells to the vessel wall, causing injury to endothelial cells and vessel walls to be more fragile, resulting in increased perioperative bleeding. The other mechanism is reduced oxygenation and altered function of inflammatory cells during the initial healing process, which would increase short-term bleeding risk (within 72 hours). McCunniff et al[9] found platelet membrane change and impaired function caused by tobacco use. Tobacco use was also found to alter the membrane of red blood cells to be more fragile and hemolytic, causing postoperative blood loss. Multivariate analysis showed a positive correlation between preoperative HCT and postoperative blood loss (P <0.001). In a postoperative bleeding study in cervical cancer, Zhao et al[10] reported a significant positive correlation between preoperative HCT and insidious postoperative blood loss (P = 0.003), because high HCT level could increase blood accumulation in interstitial space by participating in postoperative hyperfibrolysis, thus increasing total blood loss by adding to insidious loss. In addition, for the same volume of blood loss, a higher HCT level means more lost visible components in the blood, resulting in more relative blood loss. We also found a positive relationship between preoperative ALB and postoperative blood loss. Possible explanations include compromised coagulation and inhibited platelet aggregation by albumin, as reported in a previous study.[11] Second, a relatively higher level of albumin could maintain colloid osmotic pressure, which would help replenish circulation volume, resulting in a higher relative volume of blood loss. Moreover, elevated preoperative albumin would cause volume expansion secondary to increased colloid osmotic pressure and diluting clotting factors, resulting in more bleeding intraoperative and postoperatively bleeding. Herein, we have explored many demographic and hematological parameters for blood loss after TKA, and developed a prediction model to improve the prediction of risk of hemorrhage after surgery for patients with TKA. The use of the nomogram in estimating the risk of a patient harboring postoperative hemorrhage to direct clinical treatment is a new concept. For clinical use of the model, we summarized the sensitivity, specificity, negative predictive value, and positive predictive value in estimating the risk of postoperative hemorrhage using 200 as the cutoff value. Patients with a score of 200 or more are a high-risk subgroup of blood loss. Based on these preoperative and postoperative predictions, the nomogram might serve as a tool to select patients at an increased risk of blood loss and establish an effective blood management strategy. This study still had some limitations. There is a geographical limitation as the sample was sourced from a single center. And the study adopted a retrospective design and relied on past medical records. The completeness and accuracy of the data may be constrained by the standard of medical record keeping. A further prospective study is needed to verify some hypotheses we proposed. Conflicts of interest None. Funding This work was supported by CAMS Innovation Fund for Medical Sciences (CIFMS) (No. 2023-I2M-C&T-B-044).