Perioperative Factors Associated With Allogeneic Blood Transfusion in Elective THA

Abbreviations

ASA

American Society of Anesthesiologists

AUC

Area Under the Curve

BMI

Body Mass Index

CI

Confidence Interval

CS

Cannulated Screw

DHS

Dynamix Hip Screw

DVT

Deep Vein Thrombosis

EPZ

Endoprosthetic Center

FPR

Wrong Positive Rate

GT

Great Trochanter

HBL

Hidden Blood Loss

Hb

hemoglobin

Hct

hematocrit

HHS

Harris Hips Score

MCH

mean corpuscular hemoglobin

MCHC

mean corpuscular hemoglobin concentration

MCV

mean corpuscular volume

NPV

Negative Predictive Value

OR

Odds Ratio

PBM

Patient Blood Management

PE

Pulmonary Embolism

PPV

Positive Predictive Value

PRBC

Packed Red Blood Cells

RBC

red blood cells

RCT

Randomized Controlled Trial

ROC

Receiver Operating Characteristic

SE

Standard Error

STROBE

Strengthening the Reporting of Observational Epidemiology

THA

total hip arthroplasty

TRALI

transfusion‐related acute lung injury

TXA

tranxeamic acid

Study Context

This retrospective analysis was conducted in accordance with STROBE guidelines¹¹ and forms part of the Brandenburg THA Blood Management Series. All elective primary THAs performed at the Endoprosthetic Center (EPZ) of the University Hospital Brandenburg/Havel between 1 January 2016 and 31 December 2023 were identified through the institutional registry (> 1400 cases). Ethical approval for all retrospective evaluations was granted by the Ethics Committee of the University of Brandenburg (No. 292032025‐BO‐E‐RETRO).

Data Extraction

Two investigators (NR, DF) independently extracted data from the electronic hospital information system. Operative reports, anesthesia records, discharge summaries, and internal quality assurance forms were reviewed. Extracted variables included demographics (age, sex, height, weight), BMI (calculated), ASA class, surgical indication, surgeon experience, operative time, cement use, intra‐ and postoperative complications, reoperations, transfusion events (yes/no, number of RBC units), and laboratory values (RBC, Hb, Hct, MCV, MCH, MCHC). Postoperative labs were captured at predefined intervals (≤ 48 h, postoperative days 3–6, ≥ 7 days) and immediately before/after transfusion. During the entire study period (2016–2023), transfusion decisions followed institutional Patient Blood Management–oriented standards, primarily based on hemoglobin thresholds in combination with clinical symptoms (e.g., hemodynamic instability, signs of hypoxia), rather than isolated laboratory values. These criteria were applied consistently across treating physicians as part of routine clinical practice. Cup inclination was measured on standardized postoperative anteroposterior pelvic radiographs according to institutional routine. All procedures were performed within a certified Endoprosthetic Center following standardized perioperative workflows by experienced surgeons. Surgeon experience was operationalized as a categorical binary variable based on institutional certified Endoprosthetic Center classification, distinguishing procedures performed by the designated primary consultant surgeon (“Hauptoperateur”) from those performed by non‐primary surgeons. Surgeon experience was not modeled as a continuous variable, nor quantified by years of practice or cumulative case volume. Disagreements in data extraction were resolved by consensus. Missing predictor values were handled by complete‐case analysis. Cases with missing data required for the calculation of hidden blood loss (HBL) were excluded, resulting in a complete‐case analysis for all variables included in the regression model.

Blood Loss Calculations

Individual blood volume was calculated using the Nadler formula¹². Total blood loss was estimated using the Gross/Sehat dilution method. Measured blood loss was taken from the intraoperative suction volume. HBL was defined as:

Total blood loss–(suction loss+transfused volume).

One packed red blood cell (PRBC) unit was assumed to equal 260 mL unless documented otherwise.

Logistic Regression Analysis

A multivariable logistic regression model was fitted to identify independent factors associated with allogeneic red blood cell transfusion. All available demographic, laboratory, and procedural variables were entered simultaneously without prior feature reduction. The dependent variable was transfusion (yes/no). Continuous variables were used as linear terms; categorical variables were dummy‐coded. Odds ratios (OR) with 95% confidence intervals (CI) were computed. Multicollinearity diagnostics were not formally performed. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC). Formal multicollinearity diagnostics (e.g., variance inflation factors) and alternative regression strategies were not applied, as the model was intended for exploratory risk discrimination rather than coefficient optimization or causal inference. Given the inclusion of multiple biologically related hematologic variables, some degree of multicollinearity was expected and individual coefficient estimates were therefore interpreted with caution.

ROC Analysis and Cutoff Selection

Univariate receiver operating characteristic (ROC) analyzes were conducted for two single predictors: cup inclination (degrees) and 48‐h hematocrit (l/l). Optimal cutoffs were identified using Youden's index and then clinically rounded. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were computed for each threshold.

Two combined decision rules were evaluated using the rounded thresholds:

• AND rule: inclination < tIncl and Hct < tHct
• OR rule: inclination < tIncl or Hct < tHct

Cohort and Descriptive Characteristics

A total of 648 primary THA were included in the final analysis after exclusion of cases lacking sufficient data to calculate HBL, resulting in a complete‐case analytic cohort (Figure 1). Overall, 104 patients (16.0%) received at least one allogeneic red blood cell transfusion during the index hospital stay, while 544 patients (84.0%) were not transfused. The cohort had a mean age of 70.8 ± 10.3 years (median 71.5), with women accounting for 57.6% of all patients. Uncemented implants were used in 84.6% of procedures, and standard straight stems in 66.5%. Primary osteoarthritis was the dominant indication, accounting for 91.4% of cases. The mean duration of surgery was 72.0 ± 22.9 min, and mean intraoperative measured blood loss was 529.8 ± 311.1 mL. Calculated total blood loss averaged 1372 ± 504 mL (median 1335), while mean HBL amounted to 854.9 ± 443 mL (median 824). Postoperative complications were infrequent, with a reoperation rate of 2.0%. A detailed summary of baseline variables is presented in Table 1.

Note: Continuous variables are presented as mean ± standard deviation or median (range), while categorical variables are shown as absolute numbers and percentages. Total perioperative blood loss averaged 1372 ± 504 mL, and calculated HBL averaged 854.9 ± 443 mL (median 824 mL). Uncemented implant fixation accounted for 84.6% of procedures, and the overall reoperation rate was low at 2.0%.

Abbreviations: ANFH, avascular necrosis of the femoral head; ASA, American Society of Anesthesiologists; BMI, body mass index; DDH, developmental dysplasia of the hip; DVT, deep vein thrombosis; GT, greater trochanter; HBL, hidden blood loss; OA, osteoarthritis; PE, pulmonary embolism; THA, total hip arthroplasty.

Results of Multivariable Logistic Regression

The multivariable logistic regression model included 39 predictors (28 continuous and 11 categorical variables) and was based on 104 transfusion events, corresponding to an approximate events‐per‐variable ratio of 2.7. Accordingly, the model was intended for exploratory risk discrimination rather than precise coefficient estimation or causal inference (Table 2). Cup inclination showed a protective effect: each additional degree of inclination was associated with an 11% reduction in the odds of receiving a transfusion (OR = 0.89, 95% CI 0.84–0.95; p = 0.0003). However, when considered as a standalone predictor, cup inclination demonstrated poor discriminatory performance (AUC = 0.393), indicating limited clinical utility despite statistical significance in the multivariable model. In contrast, the 48‐h hematocrit demonstrated a very strong inverse relationship with transfusion risk, reflected by an extremely small odds ratio (OR ≈5.17 × 10⁻⁴⁸, 95% CI 6.38 × 10⁻⁸⁹–4.20 × 10⁻⁷; p = 0.023), indicating that lower hematocrit values at 48 h were strongly predictive of transfusion. Reoperation during the hospital stay was also significantly associated with transfusion, increasing the odds more than 13‐fold (OR = 13.19, 95% CI 1.01–171.40; p = 0.049). Overall, the model demonstrated excellent discriminatory performance (AUC = 0.931), which should be interpreted as exploratory (Figure 2).

Note: This table presents all 39 predictors included in the final model, along with logistic regression coefficients (β), standard errors (SE), Wald z‐statistics, p‐values, OR, and 95% CI. Significant predictors (cup inclination, 48‐h postoperative hematocrit, and reoperation during the hospital stay) demonstrated independent associations with transfusion risk, while remaining variables did not reach statistical significance but are shown here for transparency and completeness.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; CI, confidence interval; DVT, deep vein thrombosis; Hct, hematocrit; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; OR, odds ratio; PE, pulmonary embolism; SE, standard error.

Results of Univariate ROC Analyses

When considered as a single predictor, cup inclination yielded an AUC of 0.393, indicating poor discriminatory capability. The Youden‐derived cutoff of < 55° produced very high sensitivity (0.981) but almost negligible specificity (0.0097), resulting in a Youden index of 0.0097 and confirming that inclination alone does not meaningfully distinguish between patients who did and did not receive transfusions. For clinical interpretability, the cutoff was rounded to 55°, which corresponds to commonly accepted target ranges for acetabular cup inclination in THA and reflects routine institutional practice rather than a data‐driven optimization alone.

In contrast, the 48‐h hematocrit showed strong predictive ability with an AUC of 0.817. A threshold of < 0.28 L/L provided a reasonable balance of sensitivity (0.710) and specificity (0.866), yielding a Youden index of 0.575. Positive predictive value at this threshold was 0.535, while the negative predictive value was high at 0.928, indicating that a hematocrit ≥ 0.28 L/L strongly predicted absence of transfusion.

Results of Combined Decision Rules

Using clinically rounded thresholds, the AND rule—defined as cup inclination < 55° and 48‐h hematocrit < 0.28 L/L—demonstrated moderate sensitivity (0.573) and high specificity (0.887), yielding a Youden index of 0.460. The OR rule, which classified patients as high‐risk when either predictor was below threshold, resulted in high sensitivity but extremely poor specificity, producing a Youden index near zero. This indicates that combining the rules with an OR logic does not improve clinical discrimination, whereas the AND rule retains practical diagnostic value (Table 3).

Note: Inclination < 55° showed high sensitivity but nearly zero specificity, confirming poor discriminatory value as a standalone predictor. In contrast, hematocrit < 0.28 L/L achieved balanced sensitivity and specificity and a strong Youden index. The combined AND‐rule (inclination < 55° AND hematocrit < 0.28 L/L) yielded moderate sensitivity and high specificity, providing improved clinical discrimination compared with inclination alone.

Abbreviations: AND, logical conjunction; AUC, area under the curve; OR, logical disjunction; PPV, positive predictive value; NPV, negative predictive value.

Summary of Key Findings

In this retrospective evaluation of 8 years of elective primary THA at a certified EPZ, the 48‐h hematocrit emerged as the strongest and most clinically meaningful predictor of perioperative allogeneic transfusion. Cup inclination exhibited statistical significance only within the multivariable model but demonstrated negligible discriminatory ability when examined as an isolated predictor. Reoperation showed a strong association with transfusion, which likely reflects common underlying procedural complexity rather than a direct preoperative risk variable.

Interpretation of 48‐h Hematocrit as the Primary Transfusion Predictor

The 48‐h hematocrit emerged as the most powerful and clinically meaningful predictor of perioperative transfusion, outperforming both inclination and all other demographic and procedural variables. This strong predictive value is physiologically plausible: hematocrit reflects cumulative perioperative blood loss (visible and hidden), hemodilution, and individual patient reserve. The AUC of 0.817 and the optimal cutoff of < 0.28 L/L demonstrate that early postoperative hematocrit accurately captures the net result of intraoperative bleeding, fluid shifts, and ongoing postoperative loss. The high negative predictive value (0.928) further highlights its practical utility, as a hematocrit ≥ 0.28 L/L reliably excludes transfusion need in most cases. Unlike inclination, which operates as a surrogate of technical execution, or reoperation, which reflects postoperative course, hematocrit directly reflects the patient's physiologic state. Its superior discriminative performance confirms that transfusion decisions should be guided primarily by early postoperative hematocrit values rather than radiographic or purely procedural metrics.

Current international transfusion and Patient Blood Management (PBM) guidelines¹³ primarily base transfusion decisions on hemoglobin thresholds in combination with clinical symptoms rather than on hematocrit values alone. In this context, the 48‐h hematocrit identified in the present study should not be interpreted as an alternative transfusion trigger. Instead, it represents an early postoperative marker that integrates cumulative perioperative blood loss, hemodilution, and individual patient reserve. As such, it complements guideline‐based decision‐making by aiding early risk stratification, while definitive transfusion decisions should remain aligned with established guideline recommendations.

Interpretation of Inclination and Surgical Quality Markers

Despite statistical significance in the multivariable model, cup inclination demonstrated negligible standalone discriminative ability (AUC 0.393) and should therefore not be interpreted as a clinically useful predictor when evaluated in isolation. The conflicting statistical behavior of cup inclination, significant in multivariable analysis yet uninformative as a stand‐alone threshold, suggests that inclination functions primarily as a surrogate indicator of surgical technique rather than a biomechanically independent determinant of blood loss. A flatter, controlled inclination, as long as it is not pathologically shallow, typically corresponds with gentler exposure, balanced soft tissue tension, and more refined acetabular preparation. These technical nuances can reduce intraoperative bleeding. Thus, inclination contributes only in the context of multivariable relationships, reflecting operative quality rather than representing a direct causal predictor.

Interpretation of Reoperation as Marker of Operative Difficulty

Reoperation represents a downstream postoperative outcome and should be interpreted as a marker of underlying perioperative complexity rather than a preoperative predictor. Reoperation in this dataset refers to subsequent revision surgery or hematoma evacuation, often occurring days to months after the primary procedure. Its strong association with transfusion likely arises from shared underlying intraoperative difficulty. Cases that later required revision frequently represent technically demanding or complication‐prone surgeries characterized by increased preoperative deformity, greater implant manipulation, abnormal anatomy, or surgeon‐related complexity. Consequently, transfusion and future revision surgery may represent parallel outcomes of the same initial surgical stressors rather than an independent predictive pathway.

Value of Cutoff‐Based Decision Rules

Cutoff‐based decision rules, although attractive for clinical simplicity, demonstrated clear limitations in this study. The poor performance of the OR‐based decision rule is explained by the opposing discriminative properties of the two variables. Cup inclination showed extremely high sensitivity but almost no specificity, whereas 48‐h hematocrit demonstrated balanced sensitivity and specificity. When combined using an OR logic, the low‐specificity component dominates, resulting in a high false‐positive rate and a Youden index close to zero. In contrast, the AND rule mitigates this effect by requiring concordance of both criteria, thereby improving specificity and overall clinical usefulness. The inclination threshold (< 55°) exhibited excellent sensitivity but essentially zero specificity, confirming that cup position alone cannot be used to guide perioperative blood management decisions. In contrast, the 48‐h hematocrit threshold (< 0.28 L/L) preserved both sensitivity and specificity, producing a Youden index of 0.575 and demonstrating strong negative predictive ability. Combined decision rules revealed that an AND logic could modestly improve clinical precision, whereas an OR logic offered no meaningful benefit. These observations highlight that only hematocrit retains clinical utility as a threshold‐based predictor, whereas inclination should not be reduced to a binary risk rule.

Relationship to Relevant Evidence

Sehat et al.¹⁴ showed that “hidden” blood loss constitutes a substantial proportion of total perioperative blood loss after arthroplasty and should be considered when estimating true blood loss. Carson et al.¹⁵ summarized randomized trial evidence supporting restrictive red blood cell transfusion thresholds (including around 8 g/dL in orthopedic surgery) and emphasized integrating hemoglobin with clinical context. Rosencher et al.¹⁶ reported wide variability in European blood management practices and demonstrated that computed blood loss markedly exceeded estimated intraoperative loss, with baseline hemoglobin strongly influencing allogeneic transfusion probability. Goodnough and Shander¹⁷ outlined patient blood management as a multidisciplinary, evidence‐based strategy to optimize anemia management and minimize transfusion exposure and related risks. Patel et al.¹⁸ found that topical TXA achieved similar reductions in perioperative hemoglobin decline as intravenous TXA in a randomized arthroplasty setting, supporting pharmacologic blood‐sparing approaches.

For primary THA approaches, a 20‐RCT analysis (n = 1501) found SuperPATH to bleed ~82 mL less than posterior approach (MD −81.75 mL), while direct anterior approach bled ~92 mL more than posterior approach (MD +91.87 mL) and ~174 mL more than SuperPATH in indirect comparison (MD +173.62 mL); reported study‐level means spanned 89–1108 mL for SuperPATH, 166–391 mL for direct anterior approach, and 123.8–844.6 mL for posterior approach¹⁹. Beyond approach and procedure, meta‐regression across RCTs of SuperPATH vs. conventional hemiarthroplasty identified intraoperative blood loss (alongside age, mobilization timing, and incision length) as a significant predictor of early Harris Hip Score (HHS), reinforcing the causal plausibility that greater bleeding degrades early function²⁰. Taken together, these quantitative gradients (tens to hundreds of milliliters across techniques) are consistent with our finding that early postoperative hematocrit is an effective proxy for cumulative perioperative bleeding, whereas a simple inclination cutoff adds little discrimination unless interpreted as a surrogate of operative technique quality.

Clinical Implications

The findings have immediate relevance for perioperative blood management in THA performed within EPZ‐certified environments: (1) routine early postoperative hematocrit measurement (≤ 48 h) should be emphasized as the primary indicator for transfusion risk assessment; (2) cup inclination should not be used as an independent trigger for transfusion decisions; its predictive value only emerges when interpreted in relation to overall surgical technique quality; (3) enhanced intraoperative vigilance is warranted in technically challenging cases, as both transfusion and future reoperation appear to originate from underlying surgical difficulty rather than isolated risk variables; (4) clinical decision‐making may benefit from a pragmatic AND‐based combined rule (inclination + Hct), although hematocrit alone remains more efficient (Figure 3); (5) reoperation should be viewed as a retrospective marker of case complexity, not as a prospective stratification factor. As this was a single‐center study conducted in a certified Endoprosthetic Center with standardized perioperative workflows, the findings reflect a controlled high‐volume setting and may not be fully generalizable to institutions with different structures or surgical volumes.

Limitations

Several limitations must be acknowledged: (1) The retrospective design does not allow causal inference and may introduce documentation‐related biases. (2) Missing data were managed by complete‐case analysis, which may reduce power or introduce selection bias. (3) Multicollinearity was not formally assessed despite the inclusion of multiple correlated hematologic parameters and consequently, individual odds ratios—particularly for laboratory variables—should not be interpreted causally, and the regression results should be viewed as exploratory; (4) No internal validation (e.g., bootstrapping or cross‐validation) was performed. Consequently, the reported AUC may overestimate true predictive performance due to model complexity, the high number of predictors relative to transfusion events, and potential overfitting. The discriminative performance should therefore be interpreted as exploratory. (5) Reoperation is a postoperative event and cannot be used for prospective risk stratification; its association with transfusion reflects shared perioperative complexity rather than predictive utility; (6) As this was a single‐center EPZ study, external generalizability may be limited but reflects real‐world standardized workflows in a high‐volume certified environment; (7) Statistical significance of variables with low discriminative performance does not imply meaningful clinical discrimination and should be interpreted with caution; (8) The extremely strong association of the 48‐h hematocrit with transfusion suggests quasi‐separation within the logistic regression model, resulting in numerically extreme odds ratios; therefore, effect sizes should not be interpreted literally; (9) The single‐center design limits external validity, and the findings may not be directly generalizable to non‐certified centers or institutions with different perioperative workflows; (10) Cup inclination measurements were derived from routine postoperative radiographs, and no formal inter‐ or intra‐observer reliability analysis was performed, which may have introduced measurement variability; (11) Although institutional transfusion standards were applied throughout the study period, temporal or physician‐related variability in transfusion decisions cannot be completely excluded due to the retrospective design; (12) All associations identified in the multivariable analysis should be interpreted cautiously, as the model was explicitly exploratory, non‐validated, and not designed for causal inference. Statistically significant predictors are therefore discussed as signals of association rather than definitive determinants of transfusion risk.

Postoperative 48‐h hematocrit is a strong and clinically actionable marker associated with perioperative transfusion in primary THA. Cup inclination should be interpreted as an indirect proxy of surgical quality rather than an isolated predictor, while reoperation reflects underlying complexity rather than a preoperative risk factor. Hematocrit‐based thresholds may support early postoperative risk stratification for transfusion, whereas inclination‐based cutoffs lack discriminative power. These findings support the use of early postoperative laboratory markers for transfusion risk assessment, rather than isolated radiographic or intraoperative parameters.

Nikolai Ramadanov: conceptualization, investigation, funding acquisition, writing – original draft, writing – review and editing, visualization, validation, methodology, software, formal analysis, supervision, data curation, project administration, resources. Dakota Fuchs: investigation, data curation, methodology. Maximilian Heinz: conceptualization, writing – original draft. Robert Prill: writing – review and editing, supervision. Roland Becker: writing – review and editing, supervision.

The authors have nothing to report.

The authors declare no conflicts of interest.

Open Access funding enabled and organized by Projekt DEAL.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.