Preoperative Frailty and Chronic Pain After Cardiac Surgery: A Prospective Observational Study

A Prospective Observational Study

Britta C. Arends; Leon Timmerman; Lisette M. Vernooij; Lisa Verwijmeren; Douwe H. Biesma; Eric P. A. van Dongen; Peter G. Noordzij; Heleen J Blussé van Oud-Alblas


BMC Anesthesiol. 2022;22(201) 

In This Article


Study Design and Population

This sub-study of the Anesthesia Geriatric Evaluation and quality of life after cardiac surgery (AGE) study analyzed patients included at St. Antonius Hospital, The Netherlands.[16,17] The AGE study was a prospective observational cohort study in patients aged 70 years and older, that focused on the association between preoperative frailty with HRQL and disability after one year in elective cardiac surgery patients (i.e. coronary, valve, rhythm, aortic, or any combination of these procedures). The medical ethics committee approved the study protocol before patient recruitment (Medical Ethics Research Committees United (, number R15.039). The study was first registered at under NCT02535728 at 31/08/2015. This manuscript adheres to the applicable STROBE guidelines. All participants provided written informed consent. Details on design and analyses of the AGE study have been previously reported.[16]

Clinical Characteristics and Data Collection

After routine preoperative screening, an additional geriatric assessment was performed to assess physical, mental and social frailty in eleven domains. Physical frailty included the following domains: medication use, nutritional status using the Mini Nutritional Assessment[18] (MNA), mobility and gait speed using the Timed Get Up & Go test[19] (TGUG) and five-meter gait speed test[20] (5-MWT), daily physical functioning using Nagi's scale[20] and a handgrip strength test.[21] Screening for mental frailty included cognition using the Minimal Mental State Examination[22] (MMSE) and self-rated mental and physical health with the Short-Form 36 questionnaire (SF-36).[23,24] To assess social frailty, we evaluated the living situation and educational level. Based on the multidimensionality of the frailty syndrome, a patient was considered 'overall frail' if a positive test for physical, mental and social frailty was present. An elaborate description of frailty tests and chosen cut-off values is described in additional file Table A1. Demographics and medical history were derived from the electronic health record, including health status, comorbidities, previous surgical procedures and/or laboratory tests. Data from the SF-36 was used to identify presence of preoperative pain (see 'outcomes' section below). Information on preoperative use of analgesics was retrospectively collected from electronic patient files and included acetaminophen, non-steroid anti-inflammatory drugs (NSAIDs), opioids and antidepressants. Opioids included intravenous and subcutaneous administered morphine, oxycodone hydrochloride controlled-release (Oxycontin), oxycodone hydrochloride immediate-release (Oxynorm) and tramadol. Antidepressants were selective serotonin reuptake inhibitors (SSRIs), tricyclic antidepressants (TCAs), pregabalin and amitriptyline. Polypharmacy and excessive polypharmacy were defined as ≥ 5 and < 10 prescriptions and ≥ 10 prescriptions, respectively.

Perioperative Analgesia

Perioperative care was routinely performed according to local standard operating procedures. For intraoperative analgesia a continuous infusion of remifentanil was initiated directly after induction of anesthesia and intermittent fentanyl doses were used at predetermined times (i.e. prior to incision of the skin, sternotomy, aorta cannulation and opening of the pericardium). The dose of remifentanil and fentanyl was determined at the discretion of the attending anesthesiologist, depending on patient characteristics and intraoperative vital parameters. All patients received a loading dose of 10 mg intravenous morphine 30 min before the anticipated end of surgery. Postoperative pain management at the intensive care unit (ICU) consisted of intravenous paracetamol (1 g every six hours) and a continuous infusion of morphine (1–2 mg/h) according to protocol. After ICU discharge a standardized postoperative pain protocol was started including Oxycontin 10 mg twice daily, Oxynorm 5 mg as needed (maximum 6 times a day) and paracetamol 1 g four times a day during the first and second day at the ward. On the third day at the ward opioids were reduced and Oxynorm 5 mg as needed was prescribed, together with paracetamol 1 g four times a day. From the fourth day onwards patients received paracetamol 1 g four times a day. Insufficient pain control was managed by consultation of the hospital acute pain service that advised on an individualized pain treatment plan. Patients that suffered chronic pain preoperatively continued their pain therapy, with the exception of NSAID use. Preoperative opioid use was taken into account when defining postoperative opioid dose.


One year after cardiac surgery, study patients were invited by letter to complete and return the SF-36 questionnaire. One phone-call was used to remind non-responders. The primary outcome was chronic pain following cardiac surgery after 12 months. Data from the SF-36 questionnaires prior to and one year after surgery were used to determine chronic pain by the following question: 'How much bodily pain did you have during the past 4 weeks?' Answers were graded 1 to 6 and represented; 'None' (1), 'Very Mild' (2), 'Mild' (3), 'Moderate' (4), 'Severe' (5) and 'Very Severe' (6).[22,23] For this study, chronic pain was divided in three groups: 'No pain' (grade 1), 'Mild pain' (grade 2–3) and 'Moderate to severe pain' (grade 4–6). Chronic pain was defined as a reclassification into a higher grade of pain or no improvement of preexistent moderate to severe pain one year after cardiac surgery. The source or location of pain symptoms were not registered. Our secondary outcome was HRQL according to the SF-36.[23,24] HRQL was measured before, and at three and twelve months after surgery. Change in HRQL was expressed by a delta score between the preoperative measurement and at one year after surgery, consisting of eight sub scores (i.e. physical functioning, role functioning, role emotional, social functioning, bodily pain, mental health, vitality and general health). Sub scores ranged from 0 to 100 and were summarized into a mental HRQL and physical HRQL score, with positive values representing improvement. Death was scored as 0 points.[16]

Statistical Analysis

Data are presented as frequencies and percentages (%) for categorical data and as median with interquartile range (IQR) or mean with standard deviation (SD) for continuous data, as appropriate. Normal distribution of the variables was assessed with visual inspection of the histograms and Q-Q plots. Differences between patients with and without chronic pain one year after surgery were compared using the Chi square test for dichotomous or categorical variables or the Mann–Whitney U test or Student's T-test for continuous variables as appropriate. To investigate the association between each frailty domain and chronic pain one year after cardiac surgery, multivariable log-binomial regression analysis was performed to present effect estimates as risk ratios (RR) with accompanying 99% confidence interval (99% CI). To take multiple testing into account, we tested against a p-value of 0.01 and used a CI of 99%. Bonferroni adjustment was deemed inappropriate and too conservative as the different frailty domains are highly dependent on each other.[17] As chronic pain one year after cardiac surgery was relatively common, the rare disease assumption would not hold. This means that an odds ratio, would not approach the corresponding risk ratio, hampering the interpretation of our results for clinical practice.[25] All associations were adjusted for EuroSCORE II to take age, sex, comorbidities and weight of the procedure into account. Additionally, the association was adjusted for intraoperative use of remifentanil, preexisting chronic pain and use of internal mammary artery.[1,2,26–30] These confounders were a priori selected based on literature.[1,2,26–30] Next, change of HRQL in all eight sub scores prior to and one year after surgery was compared between patients with and without chronic pain using a Wilcoxon signed-rank test, for this univariate analysis p-values ≤ 0.01 were considered statistical significant. To investigate the association of chronic pain with HRQL after one year, multivariable linear regression models were conducted, where physical and mental HRQL measured at 12 months were used as the outcome. All associations were adjusted for EuroSCORE II, preexisting chronic pain, overall frailty and physical or mental HRQL measured prior to surgery. Estimates are expressed as linear regression coefficients (β) with accompanying 99% CI. To assess the robustness of our findings, sensitivity analysis were performed using the same analytical approaches. The first post-hoc analysis excluded all patients who died within 12 months of follow up. In the second post-hoc analysis, only patients with new or worse pain one year after surgery (i.e. reclassified into a higher grade of pain) were scored as chronic pain and patients with preexistent moderate to severe pain were excluded from this definition. As SF-36 data was missing for 11% of cases and could lead to potential bias, multiple imputation was conducted using the 'mice' library in R.[31,32] Twenty data sets were created and the estimates and variances for each of the imputed datasets were pooled into an overall estimate using Rubin's rule. The imputed dataset was used for final analyses. In order to obtain chronic pain categories after imputation, the mean frequencies of the specific answers to the SF-36 questionnaire at baseline and one year after surgery across the 20 imputation datasets were rounded to the nearest integer. An a priori sample size was not performed, as the sample size was based on the available data of the AGE study.[16] Data analysis was performed using R statistics (version 3.6.3, 2020).