Abstract
Background
Postoperative nausea and vomiting (PONV) is a common complication following anesthesia. Most existing studies have focused on patients undergoing non-orthopedic surgeries with general anesthesia, neglecting the occurrence of PONV in patients receiving spinal anesthesia. This study aims to investigate the potential risk factors for PONV in patients undergoing orthopedic surgery with spinal anesthesia through a prospective cohort study and to construct and evaluate a clinical prediction model.
Methods
Relevant factors associated with PONV in patients undergoing elective orthopedic surgery with spinal anesthesia from March11,2024, to October5,2024, were collected based on a literature review and clinical experience. Univariate and multivariate logistic regression analyses were conducted to identify potential predictive factors and construct a clinical prediction model. The model was evaluated using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA), and was ultimately visualized using a nomogram.
Results
A total of 765 patients were ultimately included in this study, with a PONV incidence of 32.8% (251 cases). Multivariate logistic regression analysis revealed that female gender (OR = 2.51,95%CI:1.55–4.06, P < 0.001), a history of PONV (OR = 3.13,95% CI:1.73–5.67, P < 0.001), and preoperative pain scores (OR = 1.46,95% CI:1.08–1.98, P = 0.013) were independent predictive factors for PONV in patients undergoing orthopedic surgery with spinal anesthesia. Although non-smoking history, motion sickness history, preoperative high anxiety, and prolonged fasting did not demonstrate statistical significance, the study integrated these seven indicators based on clinical relevance to successfully construct a nomogram.
Conclusion
The incidence of PONV is relatively high in patients undergoing orthopedic surgery with spinal anesthesia. Female gender, a history of PONV, and preoperative pain scores are identified as independent risk factors for the occurrence of PONV. The nomogram model constructed based on these factors demonstrates good predictive performance and clinical utility, aiding clinicians in the early identification of high-risk patients and the implementation of preventive strategies.
Trial registration
This study was registered on April 23, 2024, in the Chinese Clinical Trial Registry (Registration Number ChiCTR2400083373).This trial is retrospectively registered.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12871-026-03657-7.
Keywords: Spinal anesthesia, Orthopedics, Postoperative nausea and vomiting, PONV, Prediction model, Nomogram
Background
Postoperative nausea and vomiting (PONV) is a prevalent complication following surgery, characterized by symptoms of nausea, retching, or vomiting in the postoperative period. Typically, PONV manifests within 24–48 h post-operation, with some patients experiencing symptoms for 3–5 days [1]. PONV not only causes discomfort for patients but also elevates the risk of additional postoperative complications, including wound dehiscence, aspiration pneumonia, and electrolyte imbalances [2]. Studies indicate that the incidence of PONV under general anesthesia ranges from 30% to 50%, escalating to 70%−80% in high-risk patient populations [3]. However, the prevalence of PONV in spinal anesthesia surgeries remains uncertain, necessitating vigilance from anesthesiologists.
Studies have indicated that orthopedic surgeries, characterized by extensive surgical procedures, heightened opioid utilization, a high Apfel risk score, and prolonged general anesthesia duration, are considered high-risk factors for PONV [4]. As society continues to evolve and the population ages, the incidence of fractures and accidental injuries resulting from osteoporosis is projected to rise. Addressing concerns related to the elevated complication rates associated with general anesthesia in major trauma or elderly patients, such as hemodynamic instability, challenging extubation, and postoperative cognitive dysfunction [5, 6], spinal anesthesia emerges as a favorable option for such surgical cases, offering notable advantages [7]. Furthermore, spinal anesthesia can mitigate acute pain during the emergence from general anesthesia, decrease opioid requirements, lower hospitalization costs, and shorten the length of hospital stays. Nevertheless, the potential issue of spinal anesthesia triggering the vagal response and thereby increasing the risk of PONV through sympathetic nerve inhibition remains a significant consideration [8], substantially impacting patient satisfaction with anesthesia.
Currently, there is a considerable amount of research and prevention guidelines regarding PONV following general anesthesia; however, studies on PONV in patients undergoing orthopedic surgery with spinal anesthesia are limited and primarily retrospective. In clinical practice, anesthesiologists often add adjuncts (such as morphine, fentanyl, dexmedetomidine, etc.) during spinal anesthesia to enhance the efficacy of local anesthetics and maintain hemodynamic stability, which has led to a significant increase in the incidence of PONV in these patients [9, 10]. This study aims to employ a prospective cohort design to select patients undergoing orthopedic surgery with spinal anesthesia, statistically analyze and identify high-risk factors for PONV, and construct a predictive model for PONV in this population. The goal is to facilitate the early identification of high-risk patients and implement preventive strategies to reduce the occurrence of PONV.
Materials and methods
Study design and patients
This study strictly adheres to the TRIPOD statement. The prospective cohort study collected patient data from March11,2024, to October15,2024. Inclusion criteria were as follows: (a) patients undergoing elective orthopedic surgery; (b) patients clinically evaluated as suitable for undergoing spinal anesthesia; (c) patients aged ≥ 12 years with American Society of Anesthesiologists (ASA) grade I-III; (d) patients who provided informed consent for participation in the research. Exclusion criteria included: (a) patients who refused to participate in the study; (b) patients who declined spinal anesthesia; (c) patients unable to complete the assessment due to conditions such as deafness or cognitive impairment. Additional exclusion criteria were: (a) patients who experienced severe adverse events during surgery, such as massive hemorrhage, cardiac arrest, or pulmonary embolism; (b) patients transferred to the ICU postoperatively; and (c) patients who were converted to general anesthesia mid-procedure. This study has been approved by the Ethics Committee of Ningxia Hui Autonomous Region People’s Hospital (Ethics No. [2024]--LL-045) and registered with the Chinese Clinical Trial Registry (ChiCTR2400083373).
Outcome definition
Outcome Assessment: The incidence of PONV was assessed at 0–2 h, 2–6 h, 6–24 h, and 24–48 h postoperatively. Nausea was defined as a subjective sensation of the urge to vomit, assessed using a Visual Analog Scale (VAS, 0–10). Vomiting was defined as the forceful expulsion of gastric contents or retching. Rescue Protocol: Patients reporting a nausea score ≥ 4 or experiencing vomiting received rescue antiemetics (e.g., Ondansetron 4 mg IV). Definition of Endpoint: The primary endpoint, PONV, was defined as a composite outcome: the occurrence of any nausea, vomiting, or the requirement for rescue antiemetics within 48 h after surgery.
Anesthesia management
Standardized Anesthesia Protocol: Spinal anesthesia was performed at the L3-4 or L2-3 interspace. The standard intrathecal injection consisted of 0.5% bupivacaine (2–3 ml). Intrathecal Opioids: Based on the anesthesiologist’s discretion for postoperative analgesia and surgical duration and fentanyl (50 ug) was added to the intrathecal mixture in specific cases, which was recorded as the variable “Intraspinal opioid injection”.Sedation: Intraoperative sedation (e.g.,Dexmedetomidine) was administered to maintain a Ramsay Sedation Scale of 2–3. This was recorded as “Intraoperative sedation”. Prophylaxis: To observe the baseline incidence of PONV, prophylactic antiemetics were not routinely administered preoperatively unless the patient had specific contraindications or extreme risk, adhering to the pragmatic nature of this observational study.
Data acquisition
Relevant factors and baseline data associated with PONV in patients undergoing orthopedic surgery with spinal anesthesia were collected through a review of the literature and clinical experience. The collected data included: (1) Demographics: age, gender, ethnicity, body mass index (BMI), etc.; (2) Anamnestic Data: history of smoking, motion sickness, PONV, etc.; (3) Preoperative Assessments: APAIS score, preoperative pain intensity, fasting duration; and (4) Intraoperative and Postoperative Details: surgical type, anesthesia level, and analgesic usage. Detailed variables are presented in Table 1.
Table 1.
Comparison of baseline data from two datasets
| Characteristic | TotalN=765) | Training dataset (n=612) | Validation dataset (n=153) | χ2 | P |
|---|---|---|---|---|---|
| Preoperative variables | |||||
| Age (years) | 0.596 | 0.742 | |||
| ＜18 | 28 (3.7%) | 21 (3.4%) | 7 (4.6%) | ||
| 18～65 | 490 (64.1%) | 391 (63.9%) | 99 (64.7%) | ||
| ≥65 | 247 (32.3%) | 200 (32.7%) | 47 (30.7%) | ||
| Gender | 0.076 | 0.783 | |||
| Male | 320 (41.8%) | 254 (41.5%) | 66 (43.1%) | ||
| Female | 445 (58.2%) | 358 (58.5%) | 87 (56.9%) | ||
| BMI(Kg/m2) | 3.044 | 0.385 | |||
| ＜18.4 | 21 (2.7%) | 15 (2.5%) | 6 (3.9%) | ||
| 18.5～23.9 | 253 (33.1%) | 210 (34.3%) | 43 (28.1%) | ||
| 24～27.9 | 321 (42.0%) | 251 (41.0%) | 70 (45.8%) | ||
| ≥28 | 169 (22.1%) | 135 (22.1%) | 34 (22.2%) | ||
| Ethnicity | 1.961 | 0.375 | |||
| Han | 585 (76.5%) | 468 (76.5%) | 117 (76.5%) | ||
| Hui | 169 (22.1%) | 137 (22.4%) | 32 (20.9%) | ||
| Other | 11 (1.4%) | 7 (1.1%) | 4 (2.6%) | ||
| Educational level | 1.682 | 0.794 | |||
| None | 150 (19.6%) | 117 (19.1%) | 33 (21.6%) | ||
| Primary school | 117 (15.3%) | 98 (16.0%) | 19 (12.4%) | ||
| Junior high school | 217 (28.4%) | 171 (27.9%) | 46 (30.1%) | ||
| High school | 162 (21.2%) | 131 (21.4%) | 31 (20.3%) | ||
| Bachelor degree or above | 119 (15.6%) | 95 (15.5%) | 24 (15.7%) | ||
| ASA classification | 1.604 | 0.448 | |||
| Ⅰ | 18 (2.4%) | 14 (2.3%) | 4 (2.6%) | ||
| Ⅱ | 537 (70.2%) | 436 (71.2%) | 101 (66.0%) | ||
| Ⅲ | 210 (27.5%) | 162 (26.5%) | 48 (31.4%) | ||
| History of smoking | 0.354 | 0.552 | |||
| No | 601 (78.6%) | 484 (79.1%) | 117 (76.5%) | ||
| Yes | 164 (21.4%) | 128 (20.9%) | 36 (23.5%) | ||
| History of alcohol | 0.000 | 1.000 | |||
| No | 684 (89.4%) | 547 (89.4%) | 137 (89.5%) | ||
| Yes | 81 (10.6%) | 65 (10.6%) | 16 (10.5%) | ||
| History of motion sickness | 0.256 | 0.613 | |||
| No | 555 (72.5%) | 447 (73.0%) | 108 (70.6%) | ||
| Yes | 210 (27.5%) | 165 (27.0%) | 45 (29.4%) | ||
| History of surgical | 0.057 | 0.811 | |||
| No | 316 (41.3%) | 251 (41.0%) | 65 (42.5%) | ||
| Yes | 449 (58.7%) | 361 (59.0%) | 88 (57.5%) | ||
| History of PONV | 1.084 | 0.298 | |||
| No | 696 (91.0%) | 553 (90.4%) | 143 (93.5%) | ||
| Yes | 69 (9.0%) | 59 (9.6%) | 10 (6.5%) | ||
| Hypertension | 0.000 | 0.984 | |||
| No | 523 (68.4%) | 419 (68.5%) | 104 (68.0%) | ||
| Yes | 242 (31.6%) | 193 (31.5%) | 49 (32.0%) | ||
| Diabetes | 0.116 | 0.733 | |||
| No | 693 (90.6%) | 556 (90.8%) | 137 (89.5%) | ||
| Yes | 72 (9.4%) | 56 (9.2%) | 16 (10.5%) | ||
| Coronary heart disease | 0.031 | 0.861 | |||
| No | 710 (92.8%) | 567 (92.6%) | 143 (93.5%) | ||
| Yes | 55 (7.2%) | 45 (7.4%) | 10 (6.5%) | ||
| APAIS score | 0.000 | 0.986 | |||
| ＜8 | 372 (48.6%) | 297 (48.5%) | 75 (49.0%) | ||
| ≥8 | 393 (51.4%) | 315 (51.5%) | 78 (51.0%) | ||
| High anxiety related to anesthesia | 0.139 | 0.709 | |||
| No | 645 (84.3%) | 514 (84.0%) | 131 (85.6%) | ||
| Yes | 120 (15.7%) | 98 (16.0%) | 22 (14.4%) | ||
| High anxiety related to surgery | 0.562 | 0.454 | |||
| No | 614 (80.3%) | 495 (80.9%) | 119 (77.8%) | ||
| Yes | 151 (19.7%) | 117 (19.1%) | 34 (22.2%) | ||
| High information requirement | 1.013 | 0.314 | |||
| No | 680 (88.9%) | 540 (88.2%) | 140 (91.5%) | ||
| Yes | 85 (11.1%) | 72 (11.8%) | 13 (8.5%) | ||
| Preoperative fasting duration (h) | 3.471 | 0.062 | |||
| ＜16.3 | 396 (51.8%) | 306 (50.0%) | 90 (58.8%) | ||
| ≥16.3 | 369 (48.2%) | 306 (50.0%) | 63 (41.2%) | ||
| Preoperative pain intensity | 2.737 | 0.434 | |||
| 0 | 244 (31.9%) | 191 (31.2%) | 53 (34.6%) | ||
| Mild | 456 (59.6%) | 365 (59.6%) | 91 (59.5%) | ||
| Moderate | 62 (8.1%) | 54 (8.8%) | 8 (5.2%) | ||
| Severe | 3 (0.4%) | 2 (0.3%) | 1 (0.7%) | ||
| Opioid use within 24 hours before surgery | 0.454 | 0.500 | |||
| No | 741 (96.9%) | 591 (96.6%) | 150 (98.0%) | ||
| Yes | 24 (3.1%) | 21 (3.4%) | 3 (2.0%) | ||
| Intraoperative variables | |||||
| Type of surgery | 1.295 | 0.730 | |||
| Joint replacement | 357 (46.7%) | 280 (45.8%) | 77 (50.3%) | ||
| Arthroscope | 269 (35.2%) | 220 (35.9%) | 49 (32.0%) | ||
| Reduction of fracture | 109 (14.2%) | 87 (14.2%) | 22 (14.4%) | ||
| Other | 30 (3.9%) | 25 (4.1%) | 5 (3.3%) | ||
| Peak level of anesthesia ≥ T5 | 0.433 | 0.511 | |||
| No | 583 (76.2%) | 470 (76.8%) | 113 (73.9%) | ||
| Yes | 182 (23.8%) | 142 (23.2%) | 40 (26.1%) | ||
| Tourniquet pressure (kPa) | 0.040 | 0.842 | |||
| ＜35 | 382 (49.9%) | 304 (49.7%) | 78 (51.0%) | ||
| ≥35 | 383 (50.1%) | 308 (50.3%) | 75 (49.0%) | ||
| Tourniquet duration(min) | 0.016 | 0.899 | |||
| ＜73 | 374 (48.9%) | 298 (48.7%) | 76 (49.7%) | ||
| ≥73 | 391 (51.1%) | 314 (51.3%) | 77 (50.3%) | ||
| Baseline heart rate(beats/min) | 0.000 | 1.000 | |||
| ＜60 | 45 (5.9%) | 36 (5.9%) | 9 (5.9%) | ||
| ≥60 | 720 (94.1%) | 576 (94.1%) | 144 (94.1%) | ||
| Surgical duration (min) | 0.025 | 0.875 | |||
| ＜43.5 | 70 (9.2%) | 57 (9.3%) | 13 (8.5%) | ||
| ≥43.5 | 695 (90.8%) | 555 (90.7%) | 140 (91.5%) | ||
| Intraspinal opioid injection | 0.057 | 0.811 | |||
| No | 193 (25.2%) | 156 (25.5%) | 37 (24.2%) | ||
| Yes | 571 (74.6%) | 455 (74.3%) | 116 (75.8%) | ||
| Intraoperative sedation | 0.357 | 0.550 | |||
| No | 404 (52.8%) | 327 (53.4%) | 77 (50.3%) | ||
| Yes | 361 (47.2%) | 285 (46.6%) | 76 (49.7%) | ||
| Intraarticular bone cement | 0.801 | 0.371 | |||
| No | 437 (57.1%) | 355 (58.0%) | 82 (53.6%) | ||
| Yes | 328 (42.9%) | 257 (42.0%) | 71 (46.4%) | ||
| Intravenous tranexamic acid | 2.604 | 0.107 | |||
| No | 576 (75.3%) | 469 (76.6%) | 107 (69.9%) | ||
| Yes | 189 (24.7%) | 143 (23.4%) | 46 (30.1%) | ||
| Intraarticular tranexamic acid | 0.922 | 0.337 | |||
| No | 409 (53.5%) | 333 (54.4%) | 76 (49.7%) | ||
| Yes | 356 (46.5%) | 279 (45.6%) | 77 (50.3%) | ||
| Hypotension | 0.000 | 1.000 | |||
| No | 624 (81.6%) | 499 (81.5%) | 125 (81.7%) | ||
| Yes | 141 (18.4%) | 113 (18.5%) | 28 (18.3%) | ||
| Bradycardia | 2.117 | 0.146 | |||
| No | 632 (82.6%) | 499 (81.5%) | 133 (86.9%) | ||
| Yes | 133 (17.4%) | 113 (18.5%) | 20 (13.1%) | ||
| Total fluid administration(ml) | 0.000 | 0.985 | |||
| ＜1075 | 437 (57.1%) | 349 (57.0%) | 88 (57.5%) | ||
| ≥1075 | 328 (42.9%) | 263 (43.0%) | 65 (42.5%) | ||
| Postoperative variables | |||||
| Early postoperative pain intensity | 6.079 | 0.193 | |||
| 0 | 15 (2.0%) | 13 (2.1%) | 2 (1.3%) | ||
| Mild | 229 (29.9%) | 185 (30.2%) | 44 (28.8%) | ||
| Moderate | 268 (35.0%) | 219 (35.8%) | 49 (32.0%) | ||
| Severe | 252 (32.9%) | 195 (31.9%) | 57 (37.3%) | ||
| Late postoperative pain intensity | 0.336 | 0.953 | |||
| 0 | 20 (2.6%) | 16 (2.6%) | 4 (2.6%) | ||
| Mild | 472 (61.7%) | 376 (61.4%) | 96 (62.7%) | ||
| Moderate | 213 (27.8%) | 173 (28.3%) | 40 (26.1%) | ||
| Severe | 60 (7.8%) | 47 (7.7%) | 13 (8.5%) | ||
| Analgesic pump | 0.384 | 0.825 | |||
| None | 270 (35.3%) | 219 (35.8%) | 51 (33.3%) | ||
| PCIA | 175 (22.9%) | 140 (22.9%) | 35 (22.9%) | ||
| PCNA | 320 (41.8%) | 253 (41.3%) | 67 (43.8%) | ||
| Postoperative opioid use | 0.043 | 0.837 | |||
| No | 133 (17.4%) | 105 (17.2%) | 28 (18.3%) | ||
| Yes | 631(82.5%) | 506 (82.7%) | 125 (81.7%) | ||
Normal distribution of data were expressed as mean ± standard deviation. Non-normal distribution of data were expressed by the median (interquartileinterval). Percentages are retained to one decimal place.
BMI Body Mass Index, PONV Postoperative nausea and vomiting, ASA American Society of Anesthesiologist, APAIS Assessment of Preoperative Anxiety and Information Scale, PCIA Patient-controlled Intravenous Analgesia, PCNA Patient-controlled Nerve Analgesia
P-values are retained to three decimal places
Statistical methods
All statistical analyses were performed using SPSS version26.0 and R software version4.4.2. A two-tailed P value < 0.05 was considered statistically significant. First, the Shapiro-Wilk test was conducted to assess the normality of continuous variables. As the continuous data in this study did not follow a normal distribution, they were described using the median and interquartile range [M (P25, P75)], and the Mann-Whitney U test was used for intergroup comparisons. Categorical variables were expressed as frequencies and percentages [n (%)] and compared using chi-square tests or Fisher’s exact tests. The entire dataset was randomly divided into training and validation sets in an 8:2 ratio using a computer-generated random number sequence.
To identify potential predictive factors, univariate logistic regression analysis was conducted in the training set. The selected variables were then included in the multivariate logistic regression analysis. For variable inclusion, we combined statistical significance with clinical relevance to construct the model, thereby avoiding the omission of important clinical covariates. To ensure the clinical utility of the predictive model in preoperative decision-making and prevent information leakage, only preoperative variables were retained in the final model construction. Postoperative variables (such as the use of PCIA/PCNA) were excluded from the final multivariate logistic regression analysis. To enhance the clinical usability of the nomogram, continuous predictive variables (such as fasting duration and surgical duration) were converted into binary variables based on optimal cutoff values determined by ROC curve analysis (Youden’s index), with sensitivity analyses performed to confirm the robustness of the model. Stepwise selection (based on the Akaike Information Criterion, AIC) was employed for the multivariate logistic regression analysis to identify independent risk factors for PONV. The definition of independent risk factors was strictly revised to include only those variables with P values less than 0.05 in the multivariate analysis. Collinearity diagnostics were also conducted to ensure that there was no severe multicollinearity among the predictive factors.
The discriminative ability of the model was quantified using the area under the receiver operating characteristic curve (AUC), along with sensitivity and specificity. The calibration of the model (assessing the consistency between predicted probabilities and actual outcomes) was evaluated using the Hosmer-Lemeshow goodness-of-fit test (P > 0.05 indicates good fit) and visual calibration plots. Decision curve analysis (DCA) was employed to assess the clinical net benefit of the predictive model across different threshold probability ranges. Based on the independent predictive factors identified in the final multivariate analysis (P ≤ 0.05), the nomogram was constructed using the “rms” package in R to visualize the predictive model. Each patient’s total score was calculated by summing the scores of each risk factor, which corresponds to the predicted probability of PONV.
Results
A total of 800 patients who underwent elective orthopedic surgery with spinal anesthesia from March11,2024, to October15,2024, were collected for this study. After applying the exclusion criteria, 35 patients were removed, resulting in 765 patients included for statistical analysis, among whom 251 patients (32.8%) experienced PONV. The dataset was divided into training and validation sets in an 8:2 ratio, comprising 612 patients in the training set and 153 patients in the validation set (Fig. 1). Statistical analysis of baseline data and clinical characteristics of patients in both the training and validation sets revealed no statistically significant differences between the two groups (P > 0.05) (Table 1).
Fig. 1.
Study flow chart
After conducting univariate logistic regression screening for all variables and preoperative variables separately, multivariate analysis was performed (Table 2). To avoid information leakage, we ultimately selected the preoperative variables (Model1) as the final model. Based on the identified risk factors from the multivariate analysis, the logistic regression equation affecting the occurrence of PONV in patients undergoing orthopedic surgery with spinal anesthesia is as follows: Risk Score = Intercept + β1 × Gender + History of Smoking + History of Motion Sickness + History of PONV + High preoperative anxiety (APAIS score ≥ 8)+ Fasting Duration + Preoperative Pain Score. The predictive performance of the model was evaluated using ROC analysis, with Model1 demonstrating a sensitivity of 72.0% and an area under the curve (AUC) of 0.649 in the validation set. Although the overall accuracy (62.7%) and specificity (58.3%) were moderate, the high sensitivity is advantageous for preoperative screening tools (Fig. 2). In the clinical context of PONV prevention, the primary goal is to reduce false negatives (i.e., avoid missing high-risk patients) to ensure that this group receives timely preventive measures. Furthermore, the performance of our model is comparable to existing classic scoring systems (such as the Apfel score) [11], which typically show an AUC between 0.60 and 0.70 in external validation, indicating that our model provides acceptable discrimination for guiding clinical decision-making in spinal anesthesia orthopedic patients.
Table 2.
Results of multivariable regression analysis for two models
| Variable | Uni_OR (95% CI) | Uni_P | M1_OR (95% CI) | M1_P | M2_OR (95% CI) | M2_P |
|---|---|---|---|---|---|---|
| Gender | 3.45 (2.36–5.06) | 0.000 | 2.51 (1.55–4.06) | 0.000 | 2.40 (1.47–3.93) | 0.000 |
| Age ≥ 65 years | 1.16 (0.81–1.65) | 0.429 | - | - | - | - |
| BMI ≥ 28 kg/m² | 1.25 (0.84–1.86) | 0.270 | - | - | - | - |
| History of Smoking | 0.42 (0.26–0.68) | 0.000 | 0.97 (0.54–1.73) | 0.916 | 0.91 (0.51–1.64) | 0.760 |
| History of Alcohol | 0.34 (0.17–0.68) | 0.002 | - | - | - | - |
| History of MotionSickness | 1.91 (1.32–2.76) | 0.001 | 1.28 (0.85–1.93) | 0.229 | 1.37 (0.90–2.07) | 0.139 |
| History of PONV | 4.71 (2.66–8.32) | 0.000 | 3.13 (1.73–5.67) | 0.000 | 3.09 (1.69–5.63) | 0.000 |
| Hypertension | 0.89 (0.62–1.28) | 0.530 | - | - | - | - |
| Diabetes | 1.25 (0.71–2.21) | 0.437 | - | - | - | - |
| Coronary heart disease | 0.92 (0.48–1.77) | 0.797 | - | - | - | - |
| High Anxiety(APAIS) | 1.54 (1.10–2.17) | 0.012 | 1.19 (0.82–1.72) | 0.362 | 1.17 (0.81–1.70) | 0.403 |
| Fasting | 1.14 (0.82–1.60) | 0.439 | 1.03 (0.71–1.48) | 0.887 | 1.05 (0.72–1.52) | 0.813 |
| Preoperative pain Score | 1.59 (1.20–2.11) | 0.001 | 1.46 (1.08–1.98) | 0.013 | 1.36 (0.99–1.85) | 0.058 |
| Surgery duration | 1.41 (0.76–2.61) | 0.272 | - | - | 1.03 (0.52–2.05) | 0.926 |
| Intraoperative Hypotension | 1.67 (1.10–2.54) | 0.016 | - | - | 1.48 (0.94–2.33) | 0.094 |
| Bradycardia | 1.26 (0.83–1.94) | 0.279 | - | - | - | - |
| Bone Cement | 1.42 (1.01–2.00.01.00) | 0.044 | - | - | - | - |
| Sedation | 1.25 (0.89–1.75) | 0.202 | - | - | - | - |
| Intraspinal Opioid injection | 1.58 (1.05–2.38) | 0.027 | - | - | 1.56 (0.99–2.44) | 0.053 |
| PCIA | 1.00 (0.67–1.50) | 0.997 | - | - | 1.11 (0.66–1.87) | 0.695 |
| PCNA | 1.48 (1.05–2.08) | 0.024 | - | - | 1.18 (0.76–1.85) | 0.456 |
PONV Postoperative nausea and vomiting, BMI Body Mass Index, APAIS Assessment of Preoperative Anxiety and Information Scale, PCIA Patient-controlled Intravenous Analgesia, PCNA Patient-controlled Nerve Analgesia, OR odds ratio, CI confidence interval
“_” indicates that the variable was not included in the final model of the multivariable analysis
P-values are retained to three decimal places
Fig. 2.
Comparison of ROC Curves for Model1 (Preoperative Variables) and Model2(Full Variables) in (a) the training dataset and (b) the validation dataset. AUC, Area Under the Curve
The Hosmer-Lemeshow test indicated that the model exhibited good calibration in both the training set (χ²=2.926, P = 0.939) and the validation set (χ²=10.099, P = 0.258) (P > 0.05). Additionally, the calibration ability was quantitatively assessed using the Brier Score (BS), which showed good accuracy in both the training set (BS = 0.195) and the validation set (BS = 0.212) (Fig. 3). As illustrated in Fig. 3, the model demonstrated considerable clinical value in the training set. In the validation set, the net benefit of the model consistently exceeded that of the “All” and “None” strategies when the threshold probability was within the range of 0.1 to 0.45 (or 0.1 to 0.4). Although the curve exhibited fluctuations at higher thresholds due to the limited sample size of the validation cohort, this range (10% to 45%) still encompassed most of the critical threshold values for clinical decision-making related to the prevention of PONV(Fig. 4). Finally, we presented the predictive model using a nomogram created with R software (Fig. 5), along with example illustrations for clinical use (Fig. 6)
Fig. 3.
Comparison of Calibration Curves for Model1 (Preoperative Variables) and Model2 (Full Variables) in (a) the training dataset and (b) the validation dataset. BS, Brier Score
Fig. 4.
Comparison of DCA for Model1 (Preoperative Variables) and Model2 (Full Variables) in (a) the training dataset and (b) the validation dataset
Fig. 5.
Nomogram for the Final Model— Model1 (Preoperative Variables). PONV, Postoperative nausea and vomiting; APAIS, Assessment of Preoperative Anxiety and Information Scale;“0” indicates “No,” “1” indicates “Yes.”
Fig. 6.
Example Nomogram for Clinical Application of Model1 (Preoperative Variables). *P < 0.05,***P < 0.001
Discussion
In recent years, there has been increasing attention on comfort-oriented medical care, driven by the growing pursuit of comfortable living. This has prompted physicians to consider how to enable patients to face postoperative recovery more comfortably while alleviating their medical concerns. PONV is a common complication following anesthesia, which can lead to significant issues such as wound dehiscence, electrolyte imbalances, delayed postoperative recovery, and increased healthcare costs [12]. Currently, due to the widespread application of general anesthesia, many studies have focused on the prevention and treatment of PONV associated with general anesthesia, often overlooking the importance of spinal anesthesia for patients at high risk for general anesthesia. This study employed logistic regression statistical methods to identify relevant factors associated with PONV, ultimately determining that female gender, a history of PONV, and preoperative pain scores are independent risk factors for PONV in patients undergoing orthopedic surgery with spinal anesthesia. The study successfully constructed a predictive model, revealing that the incidence of PONV in this patient population is 32.8%, thereby serving as a reminder for the prevention of PONV in clinical practice.
Regarding the incidence of PONV in patients undergoing orthopedic surgery with spinal anesthesia, previous studies have rarely provided definitive answers; however, several studies have described the overall incidence in patients. Reports indicate that the overall incidence of PONV without any preventive measures is approximately 30–40%, with rates rising to as high as 80% in high-risk populations [3]. Other studies have shown that even with the use of antiemetic medications, the incidence of PONV can reach 18–38% [13, 14]. The incidence found in this study was 32.8%, which is largely consistent with previous research. Compared to earlier studies, we have more precisely defined the population of patients undergoing spinal anesthesia for orthopedic surgery. Statistics show that the total number of orthopedic surgeries in China increased by over 9% in the first half of 2024, outpacing growth in other specialties. Spinal anesthesia offers advantages over general anesthesia, including fewer complications, lower costs, and reduced postoperative pain levels [15]. Therefore, it is essential to conduct research on comfort-oriented medical care for this patient population.
In this study, female gender and a history of PONV were identified as independent risk factors for PONV in patients undergoing orthopedic surgery with spinal anesthesia, which is consistent with the guidelines for PONV management and prevention [3]. However, the high-risk factor of non-smoking mentioned in the guidelines did not show a statistically significant association in this study (P = 0.916). This may be attributed to the high proportion of Hui patients in this study, who generally do not smoke due to religious beliefs, resulting in a lack of a significant comparison with the smoking population.Additionally, this study assessed patients’ preoperative anxiety levels, revealing that when the APAIS score exceeded 8, patients exhibited significant preoperative anxiety and were more likely to experience PONV. This finding aligns with the research by Tatsuro et al., which identified anxiety as a risk factor for PONV [16]. Reports by Rita et al. also indicated that PONV is more likely to occur in individuals with pronounced anxiety symptoms [17].Moreover, whether prolonged fasting increases the incidence of PONV remains a contentious issue [3, 18]. Apfel et al. suggested in their evidence-based analysis of PONV risk factors that there is insufficient evidence to link preoperative fasting with PONV [19]. Similarly, Jonathon et al. reported that administering carbohydrates preoperatively according to Enhanced Recovery After Surgery (ERAS) principles does not alter the incidence of PONV in orthopedic surgery patients [20]. However, some studies have posited that fasting after midnight may increase the risk of PONV [21]. This is generally consistent with our findings. Multiple studies have indicated that patients receiving carbohydrate supplementation before surgery have a lower incidence of PONV compared to those fasting for extended periods [22, 23]. In our study, the continuous variable of preoperative fasting time was categorized using a cutoff value, indicating that a fasting duration greater than 16.3 h is a potential predictor of PONV.
The study also revealed a strong association between preoperative pain stimulation and the occurrence of PONV. This relationship has been less explored in previous research on PONV. Preoperative pain may lead to preoperative anxiety in patients and can impact the severity of postoperative pain [24], which was also evident in the data collection process of this study. Patients experiencing varying degrees of preoperative pain exhibited significant anxiety before surgery, which even persisted postoperatively. Whether there is a statistically significant correlation between preoperative pain and anxiety requires further exploration. Unfortunately, statistical analysis was not conducted on preoperative pain and anxiety in this study. However, Michael et al. found in their study on Post-Anesthesia Care Unit (PACU) patients, especially in orthopedic surgery patients, that postoperative pain significantly increases the incidence of PONV [25]. Therefore, it is reasonable to believe that sustained preoperative pain may trigger anxiety and hyperalgesia in patients, leading to increased postoperative pain levels and consequently triggering PONV [26].
To minimize the risk of overlooking factors related to PONV in orthopedic surgery patients under spinal anesthesia, we systematically collected all perioperative data that could be associated with PONV. Considering the potential for postoperative data leakage leading to an underestimation of the incidence of PONV, we constructed two independent models for a head-to-head performance comparison: Model1 (pure preoperative model) included only preoperative variables, intended for preoperative risk stratification; Model2 (full variable model) additionally incorporated intraoperative and postoperative variables. Through comprehensive performance evaluation of the two models (ROC, calibration curve, and DCA), we found that the inclusion of postoperative variables did not result in a clinically significant performance improvement. As shown in Fig. 1, the ROC curves of Model1 and Model2 highly overlapped in the training set (AUC:0.694 vs. 0.707) and validation set (AUC:0.649 vs. 0.678), with minimal differences in AUC. Additionally, calibration curve analysis (Fig. 2) revealed nearly identical Brier Scores for both models (validation set:0.212 vs. 0.207), indicating that the predictive accuracy of the pure preoperative model was not significantly affected by the absence of postoperative variables. Furthermore, decision curve analysis (DCA, Fig. 3) further confirmed that within the vast majority of reasonable threshold probability ranges, intervening decisions based solely on preoperative variables (Model1) provided almost the same net benefit to patients as the full variable model. Based on the robust data evidence and Occam’s Razor principle, we decided to ultimately adopt Model1 (pure preoperative model) as the recommended model for this study. This choice not only completely eliminates the risk of “information leakage,” ensuring that all predictive indicators can be obtained before the start of surgery, but also greatly enhances the clinical utility of the model - allowing anesthesiologists to assess risk directly using this model during preoperative visits, thereby enabling early preventive strategies to be established.
In comparison to previous retrospective studies on PONV in orthopedic patients, a notable strength of this study is its prospective cohort design. Firstly, to ensure data integrity and methodological rigor, we systematically collected perioperative data from765 orthopedic surgery patients under spinal anesthesia. Predictive variables (especially preoperative pain, preoperative anxiety, and other subjective indicators) were gathered preoperatively through standardized questionnaires or clearly defined criteria, thereby avoiding incomplete medical records or retrospective biases. Secondly, to effectively control confounding factors and assess the actual impact of clinically important variables, our model not only included significant variables identified in univariate analysis but also, based on existing literature and clinical judgment, forcibly incorporated variables such as preoperative fasting time, surgical duration, and the use of patient-controlled intravenous analgesia (PCIA). Thirdly, all predictive factors included in our study were strictly assessed before the occurrence of the PONV outcome, aligning perfectly with the logic of “cause before effect” in predictive modeling, thereby enhancing the causal inference strength of the conclusions.
However, this study also has certain limitations: Firstly, although this prospective study could address the data biases present in retrospective studies, the large and challenging workload of data collection hindered the acquisition of a substantial sample size. Secondly, the patient data collected in this study mainly consisted of elderly patients with lower levels of education, potentially leading to an underestimation of PONV occurrence due to their inability to accurately describe or endure discomfort. Thirdly, being a single-center study utilizing internal validation for model testing, this study may not yield a more generalizable predictive model, necessitating validation with larger sample sizes and external validation. Fourthly, to enhance the clinical simplicity and interpretability of the model, some continuous variables (such as fasting time and preoperative pain scores) were transformed into categorical variables for processing. While this approach was based on clinical or statistical distributions and simplified bedside risk assessment, it inevitably resulted in some information loss. For instance, using16.3 h as the cutoff for fasting time may not distinguish potential gradient differences in PONV risk between 4 h and 12 h. Fifthly, as a predictive model aimed at providing preoperative decision support, the deliberate exclusion of postoperative variables in the final model ensured the practical value of the model at the preoperative stage. However, this choice also meant that the model could not capture the impact of postoperative dynamic processes on PONV. Since PONV is a dynamic event, postoperative variables may also contribute to its occurrence. Developing dynamic predictive models incorporating early postoperative data will be the next step towards achieving more precise and real-time risk alerts.
Conclusions
Through a prospective cohort study, we found that the incidence of PONV in orthopedic surgery patients under spinal anesthesia was 32.8%. Female gender, history of prior PONV, and preoperative pain scores were identified as independent risk factors for PONV occurrence. Based on these predictive factors, we successfully developed a risk prediction model with good predictive performance and clinical utility, aiding in the identification of high-risk patients and the implementation of preventive measures.
Supplementary Information
Acknowledgements
The authors extend their gratitude to all colleagues and nursing staff in the Department of Anesthesiology at the People's Hospital of Ningxia Hui Autonomous Region for their support and assistance during the data collection process for this study.
Abbreviations
- PONV
Postoperative nausea and vomiting
- ROC
Receiver Operating Characteristic
- DCA
Decision Curve Analysis
- AUC
Area Under Curve
- TRIPOD
Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis
- ASA
American Society of Anesthesiologists
- ICU
Intensive Care Unit
- VAS
Visual Analog Scale
- BMI
Body Mass Index
- APAIS
Assessment of Preoperative Anxiety and Information Scale
- AIC
Akaike Information Criterion
- BS
Brier Score
- PCIA
Patient-controlled Intravenous Analgesia
- PCNA
Patient-controlled Nerve Analgesia
- OR
Odds Ratio
- CI
Confidence Interval
- ERAS
Enhanced Recovery After Surgery
- PACU
Post-anesthesia Care Unit
Authors’ contributions
YHZ was responsible for the study conception, data collection, review, statistical analysis, and manuscript writing. NKL was involved in the study conception, data collection, and review. CLL and XT contributed to data collection, review, and data entry. YNW conducted data statistical analysis. XXS participated in data collection and review.QSY was responsible for defining and revising the research protocol, as well as offering guidance and making modifications to the manuscript.
Funding
None.
Data availability
All data generated or analyzed during this study are included in this publishedarticle.
Declarations
Ethics approval and consent to participate
This study received approval from the Ethics Committee of the People’s Hospital of Ningxia Hui Autonomous Region (Approval No.: Ethics [2024]--LL-045). Written informed consent was obtained from all subjects participating in the trial. For participants aged12–18 years, written informed consent was also obtained from their parents or legal guardians. All methods were conducted in accordance with the relevant guidelines and regulations. This manuscript adheres to the applicable TRIPOD statement.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Yuhua Zou and Ningkang Li contributed equally to this work.
References
- 1.Pierre S, Whelan R. Nausea and vomiting after surgery. Contin Educ Anaesth Crit Care Pain. 2013;13(1):28–32. [Google Scholar]
- 2.Singh PM, Borle A, Rewari V, Makkar JK, Trikha A, Sinha AC, et al. Aprepitant for postoperative nausea and vomiting: a systematic review and meta-analysis. Postgrad Med J. 2016;92(1084):87–98. [DOI] [PubMed] [Google Scholar]
- 3.Gan TJ, Belani KG, Bergese S, Chung F, Diemunsch P, Habib AS, et al. Fourth consensus guidelines for the management of postoperative nausea and vomiting. Anesth Analg. 2020;131(2):411–48. [DOI] [PubMed] [Google Scholar]
- 4.Johnson JP. Preoperative assessment of high-risk orthopedic surgery patients. Nurse Pract. 2011;36(7):40–7. [DOI] [PubMed] [Google Scholar]
- 5.Mori H, Yamasaki K, Itoh T, Saishoji Y, Torisu Y, Mori T, et al. Predictors of prolonged mechanical ventilation identified at an emergency visit for elderly people: a retrospective cohort study. Medicine (Baltimore). 2020;99(49):e23472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Anwer HM, Swelem SE, El-Sheshai A, Moustafa AA. Postoperative cognitive dysfunction in adult and elderly patients. Middle East J Anesth. 2006;18(6):1123–38. [PubMed] [Google Scholar]
- 7.Neuman MD, Feng R, Carson JL, Gaskins LJ, Dillane D, Sessler DI, et al. Spinal anesthesia or general anesthesia for hip surgery in older adults. N Engl J Med. 2021;385(22):2025–35. [DOI] [PubMed] [Google Scholar]
- 8.Borgeat A, Ekatodramis G, Schenker CA. Postoperative nausea and vomiting in regional anesthesia: a review. Anesthesiology. 2003;98(2):530–47. [DOI] [PubMed] [Google Scholar]
- 9.Campos GO, de Jesus Martins M, Jesus GN, de Oliveira PRR, Lessa CN, de Oliveira Junior JCMF, et al. Palonosetron versus ondansetron for prevention of nausea and vomiting after total abdominal hysterectomy under spinal anesthesia with intrathecal morphine: a double-blind, randomized controlled trial. BMC Anesthesiol. 2019;19(1):159. [DOI] [PMC free article] [PubMed]
- 10.Srihari SS, Henjarappa KS, Arathi BH, Rashmi NR, Gowda VB, Ranganath N, et al. A randomised blinded comparison of epidural infusion of ropivacaine, ropivacaine with fentanyl for ponv and sedation in elective lower abdominal oncosurgeries. J Clin Diagn Res. 2021;15(5). 10.7860/jcdr/2021/49337.14868.
- 11.Apfel CC. Comparison of predictive models for postoperative nausea and vomiting. Br J Anaesth. 2002;88(2):234–40. [DOI] [PubMed] [Google Scholar]
- 12.Hambridge K. Assessing the risk of post-operative nausea and vomiting. Nurs Stand. 2013;27(18):35. [DOI] [PubMed] [Google Scholar]
- 13.Hyman JB, Park C, Lin HM, Cole B, Rosen L, Fenske SS, et al. Olanzapine for the prevention of postdischarge nausea and vomiting after ambulatory surgery: a randomized controlled trial. (pg 1419). Anesthesiology. 2020;132(3):133. [DOI] [PubMed]
- 14.Stjernberg M, Schlichting E, Rustoen T, Valeberg BT, Småstuen MC, Raeder JC. Postdischarge pain, nausea and patient satisfaction after diagnostic and breast-conserving ambulatory surgery for breast cancer: a cross‐sectional study. Acta Anaesthesiol Scand. 2022;66(3):317–25. [DOI] [PubMed] [Google Scholar]
- 15.Wilson JM, Farley KX, Erens GA, Guild IIIGN. General vs spinal anesthesia for revision total knee arthroplasty: do complication rates differ? J Arthroplasty. 2019;34(7):1417–22. [DOI] [PubMed] [Google Scholar]
- 16.Yokoyama T, Tamura T, Nishida K, Ito R, Nishiwaki K. Anxiety evaluated by the hospital anxiety and depression scale as a predictor of postoperative nausea and vomiting: a pilot study. Nagoya J Med Sci. 2024;86(1):72–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Laufenberg-Feldmann R, Müller M, Ferner M, Engelhard K, Kappis B. Is ‘anxiety sensitivity’ predictive of postoperative nausea and vomiting?: a prospective observational study. Eur J Anaesthesiol. 2019;36(5):369–74. [DOI] [PubMed]
- 18.Schlesinger T, Meybohm P, Kranke P. Postoperative nausea and vomiting: risk factors, prediction tools, and algorithms. Curr Opin Anaesthesiol. 2023;36(1):117–23. [DOI] [PubMed] [Google Scholar]
- 19.Apfel CC, Heidrich FM, Jukar-Rao S, Jalota L, Hornuss C, Whelan RP, et al. Evidence-based analysis of risk factors for postoperative nausea and vomiting. Br J Anaesth. 2012;109(5):742–53. [DOI] [PubMed] [Google Scholar]
- 20.Clinical JSM, Mbachair TJGM. Management of postoperative nausea and vomiting in the context of an enhanced recovery after surgery program. Best Pract Res Clin Anaesthesiol. 2020;34(4):687–700. [DOI] [PubMed] [Google Scholar]
- 21.TIAN X, YANG Y. Concept renewal leads to behavior progress: interpretation on the surgical part of consensus on eras and guidelines for pathway management in China (2018). Med J Peking Union Med Coll Hosp. 2018;9(6):485–9. [Google Scholar]
- 22.Zhao L, Liu X, Zhao Z, Gu R, Ni X, Deng R, et al. Effectiveness of enhanced recovery after surgery in the perioperative management of patients with bone surgery in China. World J Clin Cases. 2021;9(33):10151–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Marsman M, Kappen TH, Vernooij LM, Van Der Hout EC, Van Waes JA, Van Klei WA. Association of a liberal fasting policy of clear fluids before surgery with fasting duration and patient well-being and safety. JAMA Surg. 2023;158(3):254–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bradshaw P, Hariharan S, Chen D. Does preoperative psychological status of patients affect postoperative pain? A prospective study from the Caribbean. Br J Pain. 2016;10(2):108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ganter MT, Blumenthal S, Dübendorfer S, Brunnschweiler S, Hofer T, Klaghofer R, et al. The length of stay in the post-anaesthesia care unit correlates with pain intensity, nausea and vomiting on arrival. Perioper Med (Lond). 2014;3(1):10. [DOI] [PMC free article] [PubMed]
- 26.Wu H, Huang Y, Tian X, Zhang Z, Zhang Y, Mao Y, et al. Preoperative anxiety-induced glucocorticoid signaling reduces Gabaergic markers in spinal cord and promotes postoperative hyperalgesia by affecting neuronal Pas domain protein 4. Mol Pain. 2019;15:2069313137. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analyzed during this study are included in this publishedarticle.