ORIGINAL ARTICLE
Does previous chemotherapy-induced nausea and vomiting predict postoperative nausea and vomiting? ~es, A. Slullitel and H. A. Ashmawi H. B. G. da Silva, A. M. Sousa, G. M. N. Guimara Department of Anaesthesia, Cancer Institute of the State of Sao Paulo, Pain Service, S~ao Paulo, Brazil
Correspondence A. M. Sousa, Chief of the Multidisciplinary Pain Center, of the Cancer Institute of the State of Sao Paulo, Anaesthesia Department, 8th floor. as de Carvalho Aguiar, 44 – Av. Dr. Ene Cerqueira Cesar, CEP 05403-900 S~ao Paulo, Brazil E-mail: [email protected] Conflicts of interest The authors declare no conflicts of interest. Funding No funds provided. Submitted 13 April 2015; accepted 15 April 2015; submission 7 February 2015. Citation da Silva HBG, Sousa AM, Guimar~aes GMN, Slullitel A, Ashmawi HA. Does previous chemotherapy-induced nausea and vomiting predict postoperative nausea and vomiting?. Acta Anaesthesiologica Scandinavica 2015 doi: 10.1111/aas.12552
Background: Postoperative nausea and vomiting (PONV) remains a problem in the postoperative period. Previous PONV in oncology patients has recently been associated with chemotherapy-induced nausea and vomiting (CINV). We assessed if CINV could improve Apfel’s heuristic for predicting PONV. Methods: We conducted a retrospective study of 1500 consecutive patients undergoing intermediate or major cancer surgery between April and July 2011. PONV was assessed in the first postoperative day during post-anaesthesia care. The assigned anaesthetist completed an electronic medical record with all of the studied variables. Multiple logistic regression analyses were performed to assess whether any of the variables could add predictive ability to Apfel’s tallying heuristic, and receiver operating characteristic (ROC) curves were modelled. The areas under the curve (AUC) were used to compare the model’s discriminating ability for predicting patients who vomited from those who did not vomit. Results: The overall incidence of PONV was 26%. Multiple logistic regressions identified two independent predictors for PONV (odds ratio; 95% CI), Apfel’s score (1.78; 1.23–2.63) and previous chemotherapy-induced vomiting (3.15; 1.71–5.9), Hosmer–Lemeshow’s P < 0.0001. Previous CINV was the most significant predictor to be added to Apfel’s heuristic in this population. Conclusions: A history of chemotherapy-induced nausea vomiting was a strong predictor for PONV and should be investigated as an added risk factor for PONV in the preoperative period of oncology surgery in prospective studies.
Editorial comment: what this article tells us
The occurrence of nausea or vomiting in the first 24 h after intermediate or major cancer surgery was significantly associated with a history of chemotherapy-induced vomiting.
Despite the introduction of new anaesthetic agents and anaesthesia techniques, the overall incidence of postoperative nausea and vomiting (PONV), which is defined as any nausea, retching and vomiting during the first 24–48 h after surgery, remains 20–40%.1–6 In addition to being an unpleasant symptom after surgery, PONV is an important cause, with
pain management and communication with the anaesthetist, of dissatisfaction during recovery from anaesthesia.7 PONV is a multifactorial phenomenon that could be triggered by the association of emetogenic conditions and susceptible patients. There is strong evidence that anaesthesia-related and patient-specific characteristics are the most
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1145
H. B. G. DA SILVA ET AL.
important independent risk factors for PONV.8 Risk-adapted strategies should consider these conditions for individualising therapeutics9. High-risk patients could benefit from prophylaxis, whereas low-risk patients do not benefit from prophylaxis, and absolute risk reduction is predominantly related to the patients’ baseline risk.10,11 Although the aetiology of PONV is complex, simple models for predicting PONV have been used.3,4 Risk scores to predict PONV are generally based on the results of logistic regression analyses;12 however, artificial neural networks (ANN) could be used for the prediction of PONV. These networks consider complex and non-linear relationships between predictive variables and the dependent item.13,14 Although improved predictive accuracy achieved by the ANN may be clinically relevant, the disadvantages of this systems prevail because of the need of computers for risk calculation. So, the use of a simplified risk score for clinical practice may still be preferred. Chemotherapy-induced nausea and vomiting (CINV) is estimated to occur in 30–90% of patients depending on the type of chemotherapeutic agent used.15 Additionally, CINV is strongly related to PONV, and previous PONV (a strong predictor of the symptom) is a predictor of CINV.16 The hypothesis that CINV could add predictive power to Apfel’s heuristic for PONV is based on these facts.
involve any modifications of the routine standard operations or any direct intervention with the patients. The inclusion criteria were adults (over 18 years old) undergoing elective intermediate or major cancer surgery. Patients unable to understand and communicate in Portuguese (patients with orotracheal tubes, mental illness, confusion, agitation or delirium) during the first 24 h of the postoperative period were not included. Incomplete data, death or unexpected intensive care unit (ICU) admission before the completion of the 24-h protocol were considered exclusion criteria. Anaesthesia The anaesthetic technique (general anaesthesia, neuraxial anaesthesia or a combination of the two types) and antiemetic prophylactic medication were at the discretion of the clinician in charge of each case, according to the routine standards of the institution. Although the majority of the patients received some type of intraoperative antiemetic (chiefly dexamethasone and/ or ondansetron), this information was not considered in the analysis. The surgical team followed their routine regimen and prescribed postoperative antiemetic medication for most of the patients. The respective antiemetic medication assigned for each patient was registered in the patient electronic medical record. The purpose was to obtain a representative sample of our surgical population.
Methods This work was a retrospective study of PONV in patients undergoing intermediate to major cancer surgeries. Patient selection After approval of the Institutional Ethics Committee (approval number 300/10, provided by the Ethical Committee of S~ ao Paulo’s Medical School, S~ ao Paulo, Brazil), we retrospectively reviewed 1500 consecutive elective surgical inpatients between 1 April 2011 and 31 July 2011 in a tertiary cancer referral hospital. Individual informed consent was not obtained and was not required by the Institutional Ethics Committee because the study was retrospective and did not
Data source The main data source was the institutional Multidisciplinary Pain Therapy Team (MPTT) standard form, which is part of the institutional electronic medical record. Complementary data were collected from other forms of the electronic medical records. An anaesthetist or a trained pain nurse interviewed each patient in the first postoperative day as part of the routine of the hospital. The interview occurred in the surgical ward or in the ICU, according to the case. The assigned anaesthetist completed a data sheet containing all the MPTT standard form variables. The occurrence of PONV was considered for analysis if there were complaints of nausea and/or vomiting occurred in the first Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1146
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
24 h postoperatively according to either of the following: (1) patient information during the interview; (2) nausea or vomiting that was properly documented in the patient’s medical record; (3) direct information from the day shift nurse regarding patient symptoms; or (4) administration of rescue antiemetic medication other than the routine antiemetic medications prescribed. Nausea and vomiting were not scored in terms of intensity of nausea or number of episodes of vomiting and were analysed independently for the registration of their occurrence on a dichotomous basis (yes or no). The information about previous CINV was collected based exclusively on the reports given by each patient; as some of them were submitted to ambulatory chemotherapy, records about these side effects may not be described in medical records. Just like Apfel’s criteria for previous PONV, CINV was considered present if the patient was exposed to previous chemotherapy and vomited, and was considered absent if the patient did not vomit during previous chemotherapy sessions or if the patient was not exposed to chemotherapy in the past. The odds ratio of vomiting for patients with previous CINV was calculated. The variables considered in the MPTT standard forms included the following: age, sex, surgical procedure, anaesthetic technique, Apfel’s tallying heuristic (all four risk factors), neuraxial opioid administration, chronic use of opioids, postoperative opioid administration route (epidural or systemic), postoperative analgesia regimen (patient-controlled [PCA] or intermittent analgesia), type and regimen (hourly or rescue medication) of antiemetic prescribed, and personal report of nausea and vomiting during previous chemotherapy sessions. Statistical analysis The statistical analysis was performed using R from the Comprehensive Archive Network (CRAN R) and pROC statistical package.17 PONV was set as the only dependent variable for all of the tests. Fisher’s exact test and logistic regression were used for testing the dichotomous variables, logistic regression and Mann–Whitney’s U-test were used for the ordinal variables, and simple logistic regression
was used for the categorical and numerical variables. It was not possible to verify that the missing data were completely missed at random; for this reason, case-wise deletion was not used. Logistic Regression Expectation Maximisation (from the BayesLogit package) was used to reduce the bias from the missing data.18 The score from Apfel and colleagues,3 which was originally based on a logistic regression analysis, was simplified to a tallying heuristic that ignores the weighing of cues; the goal of this technique was easier and faster decisionmaking.19 Apfel’s heuristic, as defined by Pierre,20 was then calculated. A multiple logistic regression analysis was performed to assess whether any of the variables could add risk to Apfel’s tallying heuristic. This theory-driven approach selected only CINV as a predictor to add discrimination power to Apfel’s heuristic. All of the possible variables except for CINV were excluded from the multiple logistic regression for predicting PONV when considering Apfel’s heuristic even by trying forward, backward or stepwise dataguided selection for predicting PONV, resulting in the identical model as the theory-driven model. Apfel’s heuristic was forced to enter as the best predictor by the data-driven models. A new tallying heuristic was built based on this model; the new heuristic includes Apfel’s heuristic and CINV. To estimate the discriminatory power of a selected model, a receiver operating characteristic curve (ROC) was plotted. The areas under the ROC curves (AUCs) were calculated as previously described21 and were used to discriminate the models’ goodness of fit for predicting the patients who vomited from the patients who did not vomit. The 95% confidence intervals of the AUC and the hypothesis test were performed by DeLong’s non-parametric methods.21 To avoid overfitting, the Akaike information criterion was used to select the models. Hosmer– Lemeshow’s goodness of fit test was used to assess the regression models. Results The final analysis included 1491 patients, 53% female (mean age 56.18 13.96), 47% male
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1147
H. B. G. DA SILVA ET AL.
(mean age 60.37 13); nine patients were excluded from the analysis because of missing data (Fig. 1). Main surgeries performed were prostatectomy (13.4%), breast cancer surgeries (10.7%), rectosigmoidectomy (4.6%), nephrectomy (6.4%), head and neck cancer surgeries (6%), hysterectomy (5.6%), thoracotomy (4.5%) and cystectomy (3.7%) (Table 1). Gynaecological surgeries (3.9%) and cholecystectomy (0.21%) incidences were low. Most other surgeries had a very low incidence (< 1%). The prevalence and distribution of the majority of the variables are shown in Table 2. The number of antiemetic medications prescribed according to the Apfel’s score are shown in Table 3. The most predictive risk factors for PONV were female sex, Apfel’s score, a history of motion sickness or previous PONV, non-smokers, postoperative intermittent systemic analgesia, previous CINV and movement-evoked pain. In this study, neither gynaecological surgery (P = 0.29), cholecystectomy (P = 1), laparoscopic surgery (P = 0.7) nor any other type of surgery were statistically significant PONV predictors, except for breast cancer surgeries. The incidence of PONV in patients submitted to mastectomy was 38.1%. The number of antiemetics prescribed, use of postoperative opioids, epidural analgesia, patient-controlled analgesia, type of anaesthesia or chronic opioid use did not predict PONV (Table 2). Patients who did not receive postoperative analgesia and patients with moderate or intense pain were more susceptible to PONV.
The observed incidence of PONV in the population studied is compared to the observed by Apfel, which is shown in Fig. 2. The incidences of PONV observed (95%CI) in the present population were 11.2% (9.6–12.8%), 16.5% (14.62– 18.38%), 20.2% (18.16–22.24%), 33.6% (31.2– 36%) and 57.6% (55.09–60.11%), which were associated with Apfel scores 0, 1, 2, 3 and 4, respectively, whereas the expected incidences for PONV from Apfel’s original study were 10%, 21%, 39%, 61% and 79% respectively. The 95% confidence interval included the incidence predicted by Apfel’s heuristic only for the patients scoring 0 risk factors. A multiple logistic regression for predicting PONV considered Apfel’s heuristic as a predictor, and other potential predictors were added and evaluated (Table 4). Apfel’s and the proposed heuristic’s definitions are compared in Table 5. A second logistic regression modelled Apfel’s heuristic for comparisons. Another multiple logistic regression for predicting PONV by either forward, backward, stepwise, greed or theory-driven selection using all variables selected in Table 2 with P < 0.05, excluding the Apfel Score, resulted the same model. All the models selected CINV and all four Apfel’s heuristic variables (female sex, history of PONV or motion sickness, non-smoker status and planning of postoperative opioid usage) thus making this multiple logistic regression useless when compared to the first logistic regression. The AUCs were 0.64 and 0.70 (Fig. 3), respectively, for Apfel’s heuristic (scores 0–4) and the proposed heuristic (scores 0–5), and the PONV incidences for each score are shown in Fig. 2. There were significant (P = 0.049) discrimination differences between the proposed heuristic and Apfel’s heuristic (assessed by the roc.test procedure based on DeLong’s non-parametric methods). Hosmer–Lemeshow’s goodness of fit test for both models resulted in P < 0.0001. Discussion
Fig. 1. Study flow chart.
This study, which was conducted to determine the incidence of PONV in an oncological population, demonstrated a high incidence of PONV in patients classified as high risk by the Apfel score, regardless of the antiemetic administered. Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1148
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
Table 1 Procedures and corresponding PONV probabilities. Procedure
Proportion %
P (PONV)
P (female)
Odds ratio (P)
P Bonferroni
Prostatectomy Mastectomy Nephrectomy Plastic Surgery Head and Neck Surgery Hysterectomy Retossigmoidectomy Mammary setorectomy Open Thoracoscopic tumour excision Cystectomy Hepatectomy Exploratory Laparotomy Colectomy Spine Arthrodesis Gastrectomy Oophorectomy Extensive lymphadenectomy Leg amputation Duodenopancreatectomy Esophagectomy Thyroidectomy Peritonectomy Citorreduction (debunking) Multiple bone fixation Biliodigestive surgery Other procedures
13.4 10.7 6.4 6.4 6.0 5.6 4.6 4.5 4.0 3.7 2.9 2.9 2.6 2.6 2.4 1.9 1.7 1.4 1.2 1.2 1.1 1.1 0.93 0.87 0.7 9.2
0.224 0.38 0.28 0.13 0.19 0.29 0.32 0.40 0.2 0.19 0.34 0.31 0.15 0.17 0.11 0.37 0.23 0.04 0.31 0.05 0.41 0.12 0.35 0.15 0.45 0.29
0 0.96 0.31 0.70 0.3 1 0.49 1 0.48 0.35 0.45 0.63 0.47 0.30 0.5 1 0.5 0.4 0.52 0.44 0.70 0.68 1 0.46 0.54 0.47
0.8 1.9 1.1 0.4 0.6 1.2 1.3 2 0.7 0.7 1.5 1.3 0.5 0.6 0.3 1.8 0.8 0.1 1.3 0.2 2 0.5 1.6 0.5 2.4 NA
1 0.02 1 0.2 1 1 1 0.55 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Additionally, the study showed that nausea and vomiting during previous chemotherapy added predictive ability for PONV, regardless of the chemotherapy protocol used to treat the cancer. The overall incidence of PONV was 26%. When interpreting these results, it must be remembered that most of the patients received one or two antiemetics and had not been classified according to the risk of PONV until the postoperative period. Our data suggest that prophylaxis of PONV was randomly prescribed, at the discretion at the clinician not based on a risk score. The analysis of our results considered the previously validated Apfel’s. Multiple logistic regressions selected only chemotherapy-induced nausea and vomiting to help predict PONV. This proposed heuristic could add to Apfel’s heuristic a hypothesis that CINV could add significant power to the Apfel score for predicting PONV.
The hypothesis that some surgical procedures are independent risk factors is controversial. This study is underpowered to discriminate the odds ratio for PONV for cholecystectomy and laparoscopy but it suggests that it might not be high. The only surgical procedure with a significant odds ratio for PONV after Bonferroni correction was mastectomy but after controlling for Apfel’s heuristic model’s risk factors, we could not discriminate the odds ratio for PONV for mastectomy (P = 0.8). In this study, Apfel’s original heuristic model prediction by score was compared with the observed incidence of PONV in cancer patients. In the incidences of PONV in patients with scores greater than 0, which occurred in a smaller incidence than predicted by Apfel,3 there were greater differences when the score was between 2 and 3. These results are most likely related to the perioperative use of antiemetic because prophylaxis was administered in most
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1149
H. B. G. DA SILVA ET AL.
Table 2 Incidence of postoperative according to the predictors.
Predictor
Prevalence
nausea
PONV incidence
and
vomiting
Table 2 (Continued)
Predictor
P
PONV incidence
Prevalence
Anaesthesia Sex Female Male Apfel’s score
0 1 2 3 4 History of motion sickness or PONV Yes No Smoker Yes No Systemic postoperative analgesia route Yes No CINV Yes No Postoperative pain No or mild pain Mild or intense pain Postoperative analgesia Did not receive Received Number of antiemetic drugs 0 1 2 3 Use of postoperative opioids Yes No Epidural postoperative analgesia Yes No Patient-controlled analgesia Yes No
0.61 0.39
0.01 0.11 0.42 0.36 0.10
0.33 0.20
0.11 0.16 0.20 0.33 0.57
Fisher’s exact 0.02 Mann– Whitney’s U-test < 0.0001
General General and epidural General and spinal Other Chronic opioid user Yes No
0.42 0.43 0.07 0.08
0.27 0.28 0.27 0.28
0.17 0.83
0.22 0.30
P Logistic regression 0.998 0.998 0.998 Fisher’s exact 0.36
Fisher’s exact 0.19 0.81 0.15 0.85
0.49 0.22 0.12 0.32
0.61 0.39
0.23 0.29
0.42 0.58
0.43 0.18
0.77 0.23
0.23 0.33
Table 3 Mean number of prescribed antiemetics according to Apfel’s score.
0.0009
Fisher’s exact 0.02
Mean number of antiemetics
Fisher’s exact < 0.0001 Fisher’s exact 0.0007 Fisher’s exact
0.07 0.93
0.18 0.75 0.06 0.01
0.46 0.24
0.31 0.28 0.29 0
< 0.001 Mann– Whitney’s U-test 0.67
Fisher’s exact 0.66 0.34
0.29 0.28
0.88 Fisher’s exact
0.15 0.85
0.30 0.25
0.088 Fisher’s exact
0.27 0.73
0.23 0.26
Apfel’s score
Fisher’s exact 0.013
0.25
0
1
2
3
4
1
0.88
0.92
0.88
0.83
of the patients, without categorisation by risk. Ongoing prophylaxis with ondansetron and/or metoclopramide reduced the incidence of PONV in patients with one or two risk factors, as expected. However, patients at both extremes of the curve were either overmedicated or received insufficient prophylaxis. It is important to notice that the last PONV guidelines were not published yet by the time of the prophylaxis. This is the reason why metoclopramide was used for PONV prophylaxis. Although previous CINV and previous PONV have only a correlative relationship with PONV, they could be useful for predicting PONV. This finding is important because previous PONV and previous CINV must be collinear to the causal risk factors and might be excluded from multiple logistic regressions for this reason. PONV is a multifactorial phenomenon that could be triggered by multiple receptor pathways at peripheral, central or both sites,22 whereas CINV has several factors involved in its aetiology.15 The likelihood that PONV would develop in patients with a previous history of CINV is not yet established. Of all of the predictive factors Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1150
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
Fig. 2. PONV incidence for Apfel’s and the new proposed model, when CINV was added as a potential risk factor. A second logistic regression modelled Apfel’s heuristic for comparisons.
Table 4 Regression coefficients (standard errors) and odds ratios (95% confidence intervals) derived from multiple logistic regression analyses.
Intercept Apfel’s score CINV (Yes)
Coefficient (SE)
Odds Ratio (95% CI)
Pr (> Z)
2.9 (0.52) 0.58 (0.19) 1.15 (0.31)
0.05 (0.01–0.13) 1.78 (1.23–2.63) 3.15 (1.71–5.9)
1.18e-08 0.0023 0.0002
CI, confidence interval; CINV, Chemotherapy-induced nausea and vomiting.
Table 5 Definition of Apfel’s tallying heuristic and the new proposed tallying heuristic.
Risk factors Female sex Non-smoker History of PONV or motion sickness Plan for postoperative opioids Previous CINV Number of risk scores
Points (new heuristic)
Points (Apfel’s heuristic)
1 1 1
1 1 1
1
1
1 5
0 4
for CINV, the emetogenicity of a given chemotherapeutic agent is the predominant factor for predicting CINV.15 In this study, a history of previous CINV was found to be a predictive factor for PONV, regardless of the chemotherapy
Fig. 3. Receiver operating curves for Apfel’s and the new proposed model.
agent or antiemetogenic regimen used. These results link previous negative experiences of vomiting after chemotherapy to future emetogenic conditions, such as the postoperative period, irrespective of the chemotherapeutic protocol prescribed. To explain this phenomenon, the brain mechanisms involved in the physiology of nausea should be studied. The brainstem, vestibulocerebellum nuclei and amygdala are implicated in the development of emesis22; however, the mechanisms involved in this circuit are not completely understood. Visual scenes of vomiting induce emotional disgust and nausea, which are linked to anterior insula and mid-cingulate cortex activation.23 This condition best fits a Pavlovian conditioning reflex, which is similar to that which occurs after anticipatory nausea and vomiting (ANV) related to CINV.24 Nausea is associated with interoception, which is classically referred to as conscious awareness of visceral afference.25 It is likely that nausea and vomiting (NV) is not only a symptom, but a negative and emotional experience that sensitises the central nervous system for a longer duration than the exposure to the triggering factor.26 The main limitations of this study are a lack of information from the intraoperative period, such as the duration of the surgical procedure, type and dose of anaesthetics used, type and
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1151
H. B. G. DA SILVA ET AL.
doses of opioids prescribed in the postoperative period and dose of neostigmine administered. This study shows that individual patient characteristics play an important role in the prevalence of nausea and vomiting regardless of the emetogenic potential of the anaesthetics used during the anaesthesia procedure. We determined that the presence of CINV prior to surgery increased the probability of developing PONV, which is one of the most troublesome side effects after anaesthesia. As a retrospective study, sometimes it may be difficult to retrieve accurate data from medical records. As with all retrospective studies one cannot get final answers, just generate interesting hypotheses that require further prospective studies. We suggest that new prospective studies consider previous CINV as an additional risk factor for PONV in oncological patients to test the feasibility and statistical power of this new variable, including multicentre approach.
4.
5.
6.
7.
8.
Acknowledgements The authors thank the Pain team from the Cancer Institute of the State of Sao Paulo for helping to obtain these data.
9.
Author contributions H. B. G. S.: Collected data and helped to write the manuscript. A. M. S. and A. S.: Designed the study, collected data and wrote the manuscript. G. M. N. G.: Analysed and interpreted the data, helped to write the manuscript. H. A. A.: Helped to write the manuscript and supervised the project.
10.
References 1. Morino R, Ozaki M, Nagata O, Yokota M. Incidence of and risk factors for postoperative nausea and vomiting at a Japanese Cancer Center: first largescale study in Japan. J Anesth 2013; 27: 18–24. 2. Sinclair DR, Chung F, Mezei G. Can postoperative nausea and vomiting be predicted? Anesthesiology 1999; 91: 109–18. 3. Apfel CC, L€ a€ ar€ a E, Koivuranta M, Greim CA, Roewer N. A simplified risk score for predicting postoperative nausea and vomiting: conclusions
11.
12.
from cross-validations between two centers. Anesthesiology 1999; 91: 693–700. Koivuranta M, L€ a€ ar€ a E, Sn are L, Alahuhta S. A survey of postoperative nausea and vomiting. Anaesthesia 1997; 52: 443–9. Apfel CC, Kranke P, Katz MH, Goepfert C, Papenfuss T, Rauch S, Heineck R, Greim CA, Roewer N. Volatile anaesthetics may be the main cause of early but not delayed postoperative vomiting: a randomized controlled trial of factorial design. Br J Anaesth 2002; 88: 659–68. Apfel CC, Philip BK, Cakmakkaya OS, Shilling A, Shi YY, Leslie JB, Allard M, Turan A, Windle P, Odom-Forren J, Hooper VD, Radke OC, Ruiz J, Kovac A. Who is at risk for postdischarge nausea and vomiting after ambulatory surgery? Anesthesiology 2012; 117: 475–86. Hocking G, Weightman WM, Smith C, Gibbs NM, Sherrard K. Measuring the quality of anaesthesia from a patient’s perspective: development, validation, and implementation of a short questionnaire. Br J Anaesth 2013; 111: 979–89. Apfel CC, Heidrich FMFM, Jukar-Rao S, Jalota L, Hornuss C, Whelan RP, Zhang K, Cakmakkaya OS. Evidence-based analysis of risk factors for postoperative nausea and vomiting. Br J Anaesth 2012; 109: 742–53. Gan TJ, Diemunsch P, Habib AS, Kovac A, Kranke P, Meyer TA, Watcha M, Chung F, Angus S, Apfel CC, Bergese SD, Candiotti KA, Chan MT, Davis PJ, Hooper VD, Lagoo-Deenadayalan S, Myles P, Nezat G, Philip BK, Tram er MR, Society for Ambulatory Anesthesia. Consensus guidelines for the management of postoperative nausea and vomiting. Anesth Analg 2014; 118: 85–113. Apfel CC, Korttila K, Abdalla M, Kerger H, Turan A, Vedder I, Zernak C, Danner K, Jokela R, Pocock SJ, Trenkler S, Kredel M, Biedler A, Sessler DI, Roewer N; IMPACT Investigators. A factorial trial of six interventions for the prevention of postoperative nausea and vomiting. N Engl J Med 2004; 350: 575–85. White PF, O’Hara JF, Roberson CR, Wender RH, Candiotti KA, Wender RH, Candiotti KA, POST-OP Study Group. The impact of current antiemetic practices on patient outcomes: a prospective study on high-risk patients. Anesth Analg 2008; 107: 452–8. Eberhart LH, H€ ogel J, Seeling W, Staack AM, Geldner G, Georgieff M. Evaluation of three risk scores to predict postoperative nausea and vomiting. Acta Anaesthesiol Scand 2000; 44: 480–8. Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1152
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
13. Traeger M, Eberhart A, Geldner G, Morin AM, Putzke C, Wulf H, Eberhart LH. Prediction of postoperative nausea and vomiting using an artificial neural network. Anaesthesist 2003; 52: 1132–38. 14. Peng SY, Wu KC, Wang JJ, Chuang JH, Peng SK, Lai YH. Predicting postoperative nausea and vomiting with the application of an artificial neural network. Br J Anaesth 2007; 98: 60–65. 15. Hesketh PJ. Chemotherapy-induced nausea and vomiting. N Engl J Med 2008; 358: 2482–94. 16. Oddby-Muhrbeck E, Obrink E, Eksborg S, Rotstein S, L€ onnqvist PA. Is there an association between PONV and chemotherapy-induced nausea and vomiting? Acta Anaesthesiol Scand 2013; 57: 749– 53. 17. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, M€ uller M. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2001; 12: 77. 18. Dunning T, Freedman DA. Modeling selection effects. In: Outhwaite W, Turner S, eds. The Sage Handbook of Social Science Methodology. London: Sage Publications, 2007; 225–31. 19. Gigerenzer G, Gaissmaier W. Heuristic decision making. Annu Rev Psychol 2011; 62: 451–82.
20. Pierre S, Benais H, Pouymayou J. Apfel’s simplified score may favourably predict the risk of postoperative nausea and vomiting. Can J Anaesth 2002; 49: 237–42. 21. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 837–45. 22. Horn CC, Wallisch WJ, Homanics GE, Williams JP. Pathophysiological and neurochemical mechanisms of postoperative nausea and vomiting. Eur J Pharmacol 2014; 722: 55–66. 23. Harrison NA, Gray MA, Gianaros PJ, Critchley HD. The embodiment of emotional feelings in the brain. J Neurosci 2010; 30: 12878–84. 24. Roscoe JA, Morrow GR, Aapro MS, Molassiotis A, Olver I. Anticipatory nausea and vomiting. Support Care Cancer 2011; 19: 1533–8. 25. Herbert BM, Muth ER, Pollatos O, Herbert C. Interoception across modalities: on the relationship between cardiac awareness and the sensitivity for gastric functions. PLoS ONE 2012; 7: e36646. doi:10.1371/journal.pone.0036646. 26. Lichtor JL. Nausea and vomiting after surgery: it is not just postoperative. Curr Opin Anaesthesiol 2012; 25: 673–9.
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1153
Does previous chemotherapy-induced nausea and vomiting predict postoperative nausea and vomiting? ~es, A. Slullitel and H. A. Ashmawi H. B. G. da Silva, A. M. Sousa, G. M. N. Guimara Department of Anaesthesia, Cancer Institute of the State of Sao Paulo, Pain Service, S~ao Paulo, Brazil
Correspondence A. M. Sousa, Chief of the Multidisciplinary Pain Center, of the Cancer Institute of the State of Sao Paulo, Anaesthesia Department, 8th floor. as de Carvalho Aguiar, 44 – Av. Dr. Ene Cerqueira Cesar, CEP 05403-900 S~ao Paulo, Brazil E-mail: [email protected] Conflicts of interest The authors declare no conflicts of interest. Funding No funds provided. Submitted 13 April 2015; accepted 15 April 2015; submission 7 February 2015. Citation da Silva HBG, Sousa AM, Guimar~aes GMN, Slullitel A, Ashmawi HA. Does previous chemotherapy-induced nausea and vomiting predict postoperative nausea and vomiting?. Acta Anaesthesiologica Scandinavica 2015 doi: 10.1111/aas.12552
Background: Postoperative nausea and vomiting (PONV) remains a problem in the postoperative period. Previous PONV in oncology patients has recently been associated with chemotherapy-induced nausea and vomiting (CINV). We assessed if CINV could improve Apfel’s heuristic for predicting PONV. Methods: We conducted a retrospective study of 1500 consecutive patients undergoing intermediate or major cancer surgery between April and July 2011. PONV was assessed in the first postoperative day during post-anaesthesia care. The assigned anaesthetist completed an electronic medical record with all of the studied variables. Multiple logistic regression analyses were performed to assess whether any of the variables could add predictive ability to Apfel’s tallying heuristic, and receiver operating characteristic (ROC) curves were modelled. The areas under the curve (AUC) were used to compare the model’s discriminating ability for predicting patients who vomited from those who did not vomit. Results: The overall incidence of PONV was 26%. Multiple logistic regressions identified two independent predictors for PONV (odds ratio; 95% CI), Apfel’s score (1.78; 1.23–2.63) and previous chemotherapy-induced vomiting (3.15; 1.71–5.9), Hosmer–Lemeshow’s P < 0.0001. Previous CINV was the most significant predictor to be added to Apfel’s heuristic in this population. Conclusions: A history of chemotherapy-induced nausea vomiting was a strong predictor for PONV and should be investigated as an added risk factor for PONV in the preoperative period of oncology surgery in prospective studies.
Editorial comment: what this article tells us
The occurrence of nausea or vomiting in the first 24 h after intermediate or major cancer surgery was significantly associated with a history of chemotherapy-induced vomiting.
Despite the introduction of new anaesthetic agents and anaesthesia techniques, the overall incidence of postoperative nausea and vomiting (PONV), which is defined as any nausea, retching and vomiting during the first 24–48 h after surgery, remains 20–40%.1–6 In addition to being an unpleasant symptom after surgery, PONV is an important cause, with
pain management and communication with the anaesthetist, of dissatisfaction during recovery from anaesthesia.7 PONV is a multifactorial phenomenon that could be triggered by the association of emetogenic conditions and susceptible patients. There is strong evidence that anaesthesia-related and patient-specific characteristics are the most
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1145
H. B. G. DA SILVA ET AL.
important independent risk factors for PONV.8 Risk-adapted strategies should consider these conditions for individualising therapeutics9. High-risk patients could benefit from prophylaxis, whereas low-risk patients do not benefit from prophylaxis, and absolute risk reduction is predominantly related to the patients’ baseline risk.10,11 Although the aetiology of PONV is complex, simple models for predicting PONV have been used.3,4 Risk scores to predict PONV are generally based on the results of logistic regression analyses;12 however, artificial neural networks (ANN) could be used for the prediction of PONV. These networks consider complex and non-linear relationships between predictive variables and the dependent item.13,14 Although improved predictive accuracy achieved by the ANN may be clinically relevant, the disadvantages of this systems prevail because of the need of computers for risk calculation. So, the use of a simplified risk score for clinical practice may still be preferred. Chemotherapy-induced nausea and vomiting (CINV) is estimated to occur in 30–90% of patients depending on the type of chemotherapeutic agent used.15 Additionally, CINV is strongly related to PONV, and previous PONV (a strong predictor of the symptom) is a predictor of CINV.16 The hypothesis that CINV could add predictive power to Apfel’s heuristic for PONV is based on these facts.
involve any modifications of the routine standard operations or any direct intervention with the patients. The inclusion criteria were adults (over 18 years old) undergoing elective intermediate or major cancer surgery. Patients unable to understand and communicate in Portuguese (patients with orotracheal tubes, mental illness, confusion, agitation or delirium) during the first 24 h of the postoperative period were not included. Incomplete data, death or unexpected intensive care unit (ICU) admission before the completion of the 24-h protocol were considered exclusion criteria. Anaesthesia The anaesthetic technique (general anaesthesia, neuraxial anaesthesia or a combination of the two types) and antiemetic prophylactic medication were at the discretion of the clinician in charge of each case, according to the routine standards of the institution. Although the majority of the patients received some type of intraoperative antiemetic (chiefly dexamethasone and/ or ondansetron), this information was not considered in the analysis. The surgical team followed their routine regimen and prescribed postoperative antiemetic medication for most of the patients. The respective antiemetic medication assigned for each patient was registered in the patient electronic medical record. The purpose was to obtain a representative sample of our surgical population.
Methods This work was a retrospective study of PONV in patients undergoing intermediate to major cancer surgeries. Patient selection After approval of the Institutional Ethics Committee (approval number 300/10, provided by the Ethical Committee of S~ ao Paulo’s Medical School, S~ ao Paulo, Brazil), we retrospectively reviewed 1500 consecutive elective surgical inpatients between 1 April 2011 and 31 July 2011 in a tertiary cancer referral hospital. Individual informed consent was not obtained and was not required by the Institutional Ethics Committee because the study was retrospective and did not
Data source The main data source was the institutional Multidisciplinary Pain Therapy Team (MPTT) standard form, which is part of the institutional electronic medical record. Complementary data were collected from other forms of the electronic medical records. An anaesthetist or a trained pain nurse interviewed each patient in the first postoperative day as part of the routine of the hospital. The interview occurred in the surgical ward or in the ICU, according to the case. The assigned anaesthetist completed a data sheet containing all the MPTT standard form variables. The occurrence of PONV was considered for analysis if there were complaints of nausea and/or vomiting occurred in the first Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1146
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
24 h postoperatively according to either of the following: (1) patient information during the interview; (2) nausea or vomiting that was properly documented in the patient’s medical record; (3) direct information from the day shift nurse regarding patient symptoms; or (4) administration of rescue antiemetic medication other than the routine antiemetic medications prescribed. Nausea and vomiting were not scored in terms of intensity of nausea or number of episodes of vomiting and were analysed independently for the registration of their occurrence on a dichotomous basis (yes or no). The information about previous CINV was collected based exclusively on the reports given by each patient; as some of them were submitted to ambulatory chemotherapy, records about these side effects may not be described in medical records. Just like Apfel’s criteria for previous PONV, CINV was considered present if the patient was exposed to previous chemotherapy and vomited, and was considered absent if the patient did not vomit during previous chemotherapy sessions or if the patient was not exposed to chemotherapy in the past. The odds ratio of vomiting for patients with previous CINV was calculated. The variables considered in the MPTT standard forms included the following: age, sex, surgical procedure, anaesthetic technique, Apfel’s tallying heuristic (all four risk factors), neuraxial opioid administration, chronic use of opioids, postoperative opioid administration route (epidural or systemic), postoperative analgesia regimen (patient-controlled [PCA] or intermittent analgesia), type and regimen (hourly or rescue medication) of antiemetic prescribed, and personal report of nausea and vomiting during previous chemotherapy sessions. Statistical analysis The statistical analysis was performed using R from the Comprehensive Archive Network (CRAN R) and pROC statistical package.17 PONV was set as the only dependent variable for all of the tests. Fisher’s exact test and logistic regression were used for testing the dichotomous variables, logistic regression and Mann–Whitney’s U-test were used for the ordinal variables, and simple logistic regression
was used for the categorical and numerical variables. It was not possible to verify that the missing data were completely missed at random; for this reason, case-wise deletion was not used. Logistic Regression Expectation Maximisation (from the BayesLogit package) was used to reduce the bias from the missing data.18 The score from Apfel and colleagues,3 which was originally based on a logistic regression analysis, was simplified to a tallying heuristic that ignores the weighing of cues; the goal of this technique was easier and faster decisionmaking.19 Apfel’s heuristic, as defined by Pierre,20 was then calculated. A multiple logistic regression analysis was performed to assess whether any of the variables could add risk to Apfel’s tallying heuristic. This theory-driven approach selected only CINV as a predictor to add discrimination power to Apfel’s heuristic. All of the possible variables except for CINV were excluded from the multiple logistic regression for predicting PONV when considering Apfel’s heuristic even by trying forward, backward or stepwise dataguided selection for predicting PONV, resulting in the identical model as the theory-driven model. Apfel’s heuristic was forced to enter as the best predictor by the data-driven models. A new tallying heuristic was built based on this model; the new heuristic includes Apfel’s heuristic and CINV. To estimate the discriminatory power of a selected model, a receiver operating characteristic curve (ROC) was plotted. The areas under the ROC curves (AUCs) were calculated as previously described21 and were used to discriminate the models’ goodness of fit for predicting the patients who vomited from the patients who did not vomit. The 95% confidence intervals of the AUC and the hypothesis test were performed by DeLong’s non-parametric methods.21 To avoid overfitting, the Akaike information criterion was used to select the models. Hosmer– Lemeshow’s goodness of fit test was used to assess the regression models. Results The final analysis included 1491 patients, 53% female (mean age 56.18 13.96), 47% male
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1147
H. B. G. DA SILVA ET AL.
(mean age 60.37 13); nine patients were excluded from the analysis because of missing data (Fig. 1). Main surgeries performed were prostatectomy (13.4%), breast cancer surgeries (10.7%), rectosigmoidectomy (4.6%), nephrectomy (6.4%), head and neck cancer surgeries (6%), hysterectomy (5.6%), thoracotomy (4.5%) and cystectomy (3.7%) (Table 1). Gynaecological surgeries (3.9%) and cholecystectomy (0.21%) incidences were low. Most other surgeries had a very low incidence (< 1%). The prevalence and distribution of the majority of the variables are shown in Table 2. The number of antiemetic medications prescribed according to the Apfel’s score are shown in Table 3. The most predictive risk factors for PONV were female sex, Apfel’s score, a history of motion sickness or previous PONV, non-smokers, postoperative intermittent systemic analgesia, previous CINV and movement-evoked pain. In this study, neither gynaecological surgery (P = 0.29), cholecystectomy (P = 1), laparoscopic surgery (P = 0.7) nor any other type of surgery were statistically significant PONV predictors, except for breast cancer surgeries. The incidence of PONV in patients submitted to mastectomy was 38.1%. The number of antiemetics prescribed, use of postoperative opioids, epidural analgesia, patient-controlled analgesia, type of anaesthesia or chronic opioid use did not predict PONV (Table 2). Patients who did not receive postoperative analgesia and patients with moderate or intense pain were more susceptible to PONV.
The observed incidence of PONV in the population studied is compared to the observed by Apfel, which is shown in Fig. 2. The incidences of PONV observed (95%CI) in the present population were 11.2% (9.6–12.8%), 16.5% (14.62– 18.38%), 20.2% (18.16–22.24%), 33.6% (31.2– 36%) and 57.6% (55.09–60.11%), which were associated with Apfel scores 0, 1, 2, 3 and 4, respectively, whereas the expected incidences for PONV from Apfel’s original study were 10%, 21%, 39%, 61% and 79% respectively. The 95% confidence interval included the incidence predicted by Apfel’s heuristic only for the patients scoring 0 risk factors. A multiple logistic regression for predicting PONV considered Apfel’s heuristic as a predictor, and other potential predictors were added and evaluated (Table 4). Apfel’s and the proposed heuristic’s definitions are compared in Table 5. A second logistic regression modelled Apfel’s heuristic for comparisons. Another multiple logistic regression for predicting PONV by either forward, backward, stepwise, greed or theory-driven selection using all variables selected in Table 2 with P < 0.05, excluding the Apfel Score, resulted the same model. All the models selected CINV and all four Apfel’s heuristic variables (female sex, history of PONV or motion sickness, non-smoker status and planning of postoperative opioid usage) thus making this multiple logistic regression useless when compared to the first logistic regression. The AUCs were 0.64 and 0.70 (Fig. 3), respectively, for Apfel’s heuristic (scores 0–4) and the proposed heuristic (scores 0–5), and the PONV incidences for each score are shown in Fig. 2. There were significant (P = 0.049) discrimination differences between the proposed heuristic and Apfel’s heuristic (assessed by the roc.test procedure based on DeLong’s non-parametric methods). Hosmer–Lemeshow’s goodness of fit test for both models resulted in P < 0.0001. Discussion
Fig. 1. Study flow chart.
This study, which was conducted to determine the incidence of PONV in an oncological population, demonstrated a high incidence of PONV in patients classified as high risk by the Apfel score, regardless of the antiemetic administered. Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1148
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
Table 1 Procedures and corresponding PONV probabilities. Procedure
Proportion %
P (PONV)
P (female)
Odds ratio (P)
P Bonferroni
Prostatectomy Mastectomy Nephrectomy Plastic Surgery Head and Neck Surgery Hysterectomy Retossigmoidectomy Mammary setorectomy Open Thoracoscopic tumour excision Cystectomy Hepatectomy Exploratory Laparotomy Colectomy Spine Arthrodesis Gastrectomy Oophorectomy Extensive lymphadenectomy Leg amputation Duodenopancreatectomy Esophagectomy Thyroidectomy Peritonectomy Citorreduction (debunking) Multiple bone fixation Biliodigestive surgery Other procedures
13.4 10.7 6.4 6.4 6.0 5.6 4.6 4.5 4.0 3.7 2.9 2.9 2.6 2.6 2.4 1.9 1.7 1.4 1.2 1.2 1.1 1.1 0.93 0.87 0.7 9.2
0.224 0.38 0.28 0.13 0.19 0.29 0.32 0.40 0.2 0.19 0.34 0.31 0.15 0.17 0.11 0.37 0.23 0.04 0.31 0.05 0.41 0.12 0.35 0.15 0.45 0.29
0 0.96 0.31 0.70 0.3 1 0.49 1 0.48 0.35 0.45 0.63 0.47 0.30 0.5 1 0.5 0.4 0.52 0.44 0.70 0.68 1 0.46 0.54 0.47
0.8 1.9 1.1 0.4 0.6 1.2 1.3 2 0.7 0.7 1.5 1.3 0.5 0.6 0.3 1.8 0.8 0.1 1.3 0.2 2 0.5 1.6 0.5 2.4 NA
1 0.02 1 0.2 1 1 1 0.55 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
Additionally, the study showed that nausea and vomiting during previous chemotherapy added predictive ability for PONV, regardless of the chemotherapy protocol used to treat the cancer. The overall incidence of PONV was 26%. When interpreting these results, it must be remembered that most of the patients received one or two antiemetics and had not been classified according to the risk of PONV until the postoperative period. Our data suggest that prophylaxis of PONV was randomly prescribed, at the discretion at the clinician not based on a risk score. The analysis of our results considered the previously validated Apfel’s. Multiple logistic regressions selected only chemotherapy-induced nausea and vomiting to help predict PONV. This proposed heuristic could add to Apfel’s heuristic a hypothesis that CINV could add significant power to the Apfel score for predicting PONV.
The hypothesis that some surgical procedures are independent risk factors is controversial. This study is underpowered to discriminate the odds ratio for PONV for cholecystectomy and laparoscopy but it suggests that it might not be high. The only surgical procedure with a significant odds ratio for PONV after Bonferroni correction was mastectomy but after controlling for Apfel’s heuristic model’s risk factors, we could not discriminate the odds ratio for PONV for mastectomy (P = 0.8). In this study, Apfel’s original heuristic model prediction by score was compared with the observed incidence of PONV in cancer patients. In the incidences of PONV in patients with scores greater than 0, which occurred in a smaller incidence than predicted by Apfel,3 there were greater differences when the score was between 2 and 3. These results are most likely related to the perioperative use of antiemetic because prophylaxis was administered in most
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1149
H. B. G. DA SILVA ET AL.
Table 2 Incidence of postoperative according to the predictors.
Predictor
Prevalence
nausea
PONV incidence
and
vomiting
Table 2 (Continued)
Predictor
P
PONV incidence
Prevalence
Anaesthesia Sex Female Male Apfel’s score
0 1 2 3 4 History of motion sickness or PONV Yes No Smoker Yes No Systemic postoperative analgesia route Yes No CINV Yes No Postoperative pain No or mild pain Mild or intense pain Postoperative analgesia Did not receive Received Number of antiemetic drugs 0 1 2 3 Use of postoperative opioids Yes No Epidural postoperative analgesia Yes No Patient-controlled analgesia Yes No
0.61 0.39
0.01 0.11 0.42 0.36 0.10
0.33 0.20
0.11 0.16 0.20 0.33 0.57
Fisher’s exact 0.02 Mann– Whitney’s U-test < 0.0001
General General and epidural General and spinal Other Chronic opioid user Yes No
0.42 0.43 0.07 0.08
0.27 0.28 0.27 0.28
0.17 0.83
0.22 0.30
P Logistic regression 0.998 0.998 0.998 Fisher’s exact 0.36
Fisher’s exact 0.19 0.81 0.15 0.85
0.49 0.22 0.12 0.32
0.61 0.39
0.23 0.29
0.42 0.58
0.43 0.18
0.77 0.23
0.23 0.33
Table 3 Mean number of prescribed antiemetics according to Apfel’s score.
0.0009
Fisher’s exact 0.02
Mean number of antiemetics
Fisher’s exact < 0.0001 Fisher’s exact 0.0007 Fisher’s exact
0.07 0.93
0.18 0.75 0.06 0.01
0.46 0.24
0.31 0.28 0.29 0
< 0.001 Mann– Whitney’s U-test 0.67
Fisher’s exact 0.66 0.34
0.29 0.28
0.88 Fisher’s exact
0.15 0.85
0.30 0.25
0.088 Fisher’s exact
0.27 0.73
0.23 0.26
Apfel’s score
Fisher’s exact 0.013
0.25
0
1
2
3
4
1
0.88
0.92
0.88
0.83
of the patients, without categorisation by risk. Ongoing prophylaxis with ondansetron and/or metoclopramide reduced the incidence of PONV in patients with one or two risk factors, as expected. However, patients at both extremes of the curve were either overmedicated or received insufficient prophylaxis. It is important to notice that the last PONV guidelines were not published yet by the time of the prophylaxis. This is the reason why metoclopramide was used for PONV prophylaxis. Although previous CINV and previous PONV have only a correlative relationship with PONV, they could be useful for predicting PONV. This finding is important because previous PONV and previous CINV must be collinear to the causal risk factors and might be excluded from multiple logistic regressions for this reason. PONV is a multifactorial phenomenon that could be triggered by multiple receptor pathways at peripheral, central or both sites,22 whereas CINV has several factors involved in its aetiology.15 The likelihood that PONV would develop in patients with a previous history of CINV is not yet established. Of all of the predictive factors Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1150
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
Fig. 2. PONV incidence for Apfel’s and the new proposed model, when CINV was added as a potential risk factor. A second logistic regression modelled Apfel’s heuristic for comparisons.
Table 4 Regression coefficients (standard errors) and odds ratios (95% confidence intervals) derived from multiple logistic regression analyses.
Intercept Apfel’s score CINV (Yes)
Coefficient (SE)
Odds Ratio (95% CI)
Pr (> Z)
2.9 (0.52) 0.58 (0.19) 1.15 (0.31)
0.05 (0.01–0.13) 1.78 (1.23–2.63) 3.15 (1.71–5.9)
1.18e-08 0.0023 0.0002
CI, confidence interval; CINV, Chemotherapy-induced nausea and vomiting.
Table 5 Definition of Apfel’s tallying heuristic and the new proposed tallying heuristic.
Risk factors Female sex Non-smoker History of PONV or motion sickness Plan for postoperative opioids Previous CINV Number of risk scores
Points (new heuristic)
Points (Apfel’s heuristic)
1 1 1
1 1 1
1
1
1 5
0 4
for CINV, the emetogenicity of a given chemotherapeutic agent is the predominant factor for predicting CINV.15 In this study, a history of previous CINV was found to be a predictive factor for PONV, regardless of the chemotherapy
Fig. 3. Receiver operating curves for Apfel’s and the new proposed model.
agent or antiemetogenic regimen used. These results link previous negative experiences of vomiting after chemotherapy to future emetogenic conditions, such as the postoperative period, irrespective of the chemotherapeutic protocol prescribed. To explain this phenomenon, the brain mechanisms involved in the physiology of nausea should be studied. The brainstem, vestibulocerebellum nuclei and amygdala are implicated in the development of emesis22; however, the mechanisms involved in this circuit are not completely understood. Visual scenes of vomiting induce emotional disgust and nausea, which are linked to anterior insula and mid-cingulate cortex activation.23 This condition best fits a Pavlovian conditioning reflex, which is similar to that which occurs after anticipatory nausea and vomiting (ANV) related to CINV.24 Nausea is associated with interoception, which is classically referred to as conscious awareness of visceral afference.25 It is likely that nausea and vomiting (NV) is not only a symptom, but a negative and emotional experience that sensitises the central nervous system for a longer duration than the exposure to the triggering factor.26 The main limitations of this study are a lack of information from the intraoperative period, such as the duration of the surgical procedure, type and dose of anaesthetics used, type and
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1151
H. B. G. DA SILVA ET AL.
doses of opioids prescribed in the postoperative period and dose of neostigmine administered. This study shows that individual patient characteristics play an important role in the prevalence of nausea and vomiting regardless of the emetogenic potential of the anaesthetics used during the anaesthesia procedure. We determined that the presence of CINV prior to surgery increased the probability of developing PONV, which is one of the most troublesome side effects after anaesthesia. As a retrospective study, sometimes it may be difficult to retrieve accurate data from medical records. As with all retrospective studies one cannot get final answers, just generate interesting hypotheses that require further prospective studies. We suggest that new prospective studies consider previous CINV as an additional risk factor for PONV in oncological patients to test the feasibility and statistical power of this new variable, including multicentre approach.
4.
5.
6.
7.
8.
Acknowledgements The authors thank the Pain team from the Cancer Institute of the State of Sao Paulo for helping to obtain these data.
9.
Author contributions H. B. G. S.: Collected data and helped to write the manuscript. A. M. S. and A. S.: Designed the study, collected data and wrote the manuscript. G. M. N. G.: Analysed and interpreted the data, helped to write the manuscript. H. A. A.: Helped to write the manuscript and supervised the project.
10.
References 1. Morino R, Ozaki M, Nagata O, Yokota M. Incidence of and risk factors for postoperative nausea and vomiting at a Japanese Cancer Center: first largescale study in Japan. J Anesth 2013; 27: 18–24. 2. Sinclair DR, Chung F, Mezei G. Can postoperative nausea and vomiting be predicted? Anesthesiology 1999; 91: 109–18. 3. Apfel CC, L€ a€ ar€ a E, Koivuranta M, Greim CA, Roewer N. A simplified risk score for predicting postoperative nausea and vomiting: conclusions
11.
12.
from cross-validations between two centers. Anesthesiology 1999; 91: 693–700. Koivuranta M, L€ a€ ar€ a E, Sn are L, Alahuhta S. A survey of postoperative nausea and vomiting. Anaesthesia 1997; 52: 443–9. Apfel CC, Kranke P, Katz MH, Goepfert C, Papenfuss T, Rauch S, Heineck R, Greim CA, Roewer N. Volatile anaesthetics may be the main cause of early but not delayed postoperative vomiting: a randomized controlled trial of factorial design. Br J Anaesth 2002; 88: 659–68. Apfel CC, Philip BK, Cakmakkaya OS, Shilling A, Shi YY, Leslie JB, Allard M, Turan A, Windle P, Odom-Forren J, Hooper VD, Radke OC, Ruiz J, Kovac A. Who is at risk for postdischarge nausea and vomiting after ambulatory surgery? Anesthesiology 2012; 117: 475–86. Hocking G, Weightman WM, Smith C, Gibbs NM, Sherrard K. Measuring the quality of anaesthesia from a patient’s perspective: development, validation, and implementation of a short questionnaire. Br J Anaesth 2013; 111: 979–89. Apfel CC, Heidrich FMFM, Jukar-Rao S, Jalota L, Hornuss C, Whelan RP, Zhang K, Cakmakkaya OS. Evidence-based analysis of risk factors for postoperative nausea and vomiting. Br J Anaesth 2012; 109: 742–53. Gan TJ, Diemunsch P, Habib AS, Kovac A, Kranke P, Meyer TA, Watcha M, Chung F, Angus S, Apfel CC, Bergese SD, Candiotti KA, Chan MT, Davis PJ, Hooper VD, Lagoo-Deenadayalan S, Myles P, Nezat G, Philip BK, Tram er MR, Society for Ambulatory Anesthesia. Consensus guidelines for the management of postoperative nausea and vomiting. Anesth Analg 2014; 118: 85–113. Apfel CC, Korttila K, Abdalla M, Kerger H, Turan A, Vedder I, Zernak C, Danner K, Jokela R, Pocock SJ, Trenkler S, Kredel M, Biedler A, Sessler DI, Roewer N; IMPACT Investigators. A factorial trial of six interventions for the prevention of postoperative nausea and vomiting. N Engl J Med 2004; 350: 575–85. White PF, O’Hara JF, Roberson CR, Wender RH, Candiotti KA, Wender RH, Candiotti KA, POST-OP Study Group. The impact of current antiemetic practices on patient outcomes: a prospective study on high-risk patients. Anesth Analg 2008; 107: 452–8. Eberhart LH, H€ ogel J, Seeling W, Staack AM, Geldner G, Georgieff M. Evaluation of three risk scores to predict postoperative nausea and vomiting. Acta Anaesthesiol Scand 2000; 44: 480–8. Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153
1152
ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
POSTOPERATIVE NAUSEA AND VOMITING
13. Traeger M, Eberhart A, Geldner G, Morin AM, Putzke C, Wulf H, Eberhart LH. Prediction of postoperative nausea and vomiting using an artificial neural network. Anaesthesist 2003; 52: 1132–38. 14. Peng SY, Wu KC, Wang JJ, Chuang JH, Peng SK, Lai YH. Predicting postoperative nausea and vomiting with the application of an artificial neural network. Br J Anaesth 2007; 98: 60–65. 15. Hesketh PJ. Chemotherapy-induced nausea and vomiting. N Engl J Med 2008; 358: 2482–94. 16. Oddby-Muhrbeck E, Obrink E, Eksborg S, Rotstein S, L€ onnqvist PA. Is there an association between PONV and chemotherapy-induced nausea and vomiting? Acta Anaesthesiol Scand 2013; 57: 749– 53. 17. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, M€ uller M. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2001; 12: 77. 18. Dunning T, Freedman DA. Modeling selection effects. In: Outhwaite W, Turner S, eds. The Sage Handbook of Social Science Methodology. London: Sage Publications, 2007; 225–31. 19. Gigerenzer G, Gaissmaier W. Heuristic decision making. Annu Rev Psychol 2011; 62: 451–82.
20. Pierre S, Benais H, Pouymayou J. Apfel’s simplified score may favourably predict the risk of postoperative nausea and vomiting. Can J Anaesth 2002; 49: 237–42. 21. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 837–45. 22. Horn CC, Wallisch WJ, Homanics GE, Williams JP. Pathophysiological and neurochemical mechanisms of postoperative nausea and vomiting. Eur J Pharmacol 2014; 722: 55–66. 23. Harrison NA, Gray MA, Gianaros PJ, Critchley HD. The embodiment of emotional feelings in the brain. J Neurosci 2010; 30: 12878–84. 24. Roscoe JA, Morrow GR, Aapro MS, Molassiotis A, Olver I. Anticipatory nausea and vomiting. Support Care Cancer 2011; 19: 1533–8. 25. Herbert BM, Muth ER, Pollatos O, Herbert C. Interoception across modalities: on the relationship between cardiac awareness and the sensitivity for gastric functions. PLoS ONE 2012; 7: e36646. doi:10.1371/journal.pone.0036646. 26. Lichtor JL. Nausea and vomiting after surgery: it is not just postoperative. Curr Opin Anaesthesiol 2012; 25: 673–9.
Acta Anaesthesiologica Scandinavica 59 (2015) 1145–1153 ª 2015 The Acta Anaesthesiologica Scandinavica Foundation. Published by John Wiley & Sons Ltd
1153
