Prevalence and Risk Factors for Chronic Postsurgical Pain after Cardiac Surgery: A Single-center Prospective Cohort Study

We undertook a substudy of a completed prospective cohort study that recruited consecutive patients undergoing cardiothoracic surgery at Toronto General Hospital (Toronto, Canada) between August 2012 and April 2020. The study was approved by the University Health Network Research Ethics Board (Toronto, Canada; REB #12-0078.21) and York University (Toronto, Canada; REB #12-0078-AE). The protocol was retrospectively registered on clinicaltrials.gov (NCT04798573, March 15, 2021, by Dr. Clarke) after patient enrollment but before data analysis. We followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guideline.¹⁰ 

Patients included were 18 yr or older undergoing nonemergent open cardiac surgery via median sternotomy (coronary artery bypass grafting [CABG] surgery, valve replacement or repair, aneurysm repair, heart transplant, or insertion of a ventricular assist device). Exclusion criteria were a history of psychotic disorders (e.g., schizophrenia) or major cognitive disorder precluding participation, lack of capacity to provide informed consent, and nonfluency in English. All participants provided written informed consent before enrollment.

Before surgery, participants provided data on demographic factors (age, sex, body mass index) and baseline pain intensity (including angina pectoris) assessed using the numeric rating scale from 0 (no pain) to 10 (worst pain imaginable).¹¹  In addition, two standardized psychologic questionnaires were administered to assess emotional and pain-related distress: the Hospital Anxiety and Depression Scale (HADS)¹²  and the Pain Catastrophizing Scale.¹³  The HADS comprises seven anxiety and seven depression-related items, with higher scores on each subscale (HADS anxiety [HADS-A], HADS depression [HADS-D]) indicating greater levels of anxiety and depression. The Pain Catastrophizing Scale consists of 13 self-reported items to assess the extent of catastrophic thinking about pain.¹³ 

Postoperatively, the following medical-surgical variables were extracted via chart review: past medical history (hypertension, myocardial infarction, congestive heart failure, coronary artery disease, angina, asthma, diabetes, osteoarthritis, smoking history); surgical variables (type of procedure, total surgical time [in minutes], cardiopulmonary bypass time [in minutes], anesthetic type); and postoperative variables (days on acute pain service, hospital length of stay, cumulative equianalgesic dose during the first 5 days after surgery in morphine milligram equivalent).¹⁴  All patients were assigned to routine institutional analgesic protocols, administered, and supervised by the acute pain service. Acute postoperative pain was also measured on each day for the first 5 postoperative days, during which participants were asked to rate the intensity of their worst and average postoperative pain during the preceding 24 h (0 to 10 rating scale). The frequency of missingness for each variable ranged from 0.5% for age to 2.3% for the HADS score.

The primary outcome is the presence of nonzero chronic chest pain beyond the expected healing time of 3 months, as per the International Association for the Study of Pain (Washington, D.C.) definition.^2,15  Patients were mailed questionnaires at 3, 6, and 12 months postoperatively to assess for chronic pain around their surgical scar sites (i.e., the chest area and leg area if the saphenous vein was harvested). Patients who reported chronic chest or leg pain had their medical records screened to rule out other potential causes of surgical-site pain (e.g., infection, cancer). To assess the severity of pain, we modified the Brief Pain Inventory questionnaire and collected information on pain intensity in the surgical area (0 to 10 rating scale); pain unpleasantness (0 [not unpleasant] to 10 [most unpleasant sensation]); and nature of pain (presence of numbness, location of pain [skin, scar, muscle or joint], frequency of pain episodes, and duration of pain). A version of the questionnaire is available in the Supplemental Digital Content (https://links.lww.com/ALN/D158).

The Leeds Assessment of Neuropathic Signs and Symptoms Pain Questionnaire (S-LANSS) was used to investigate whether chronic pain was compatible with a neuropathic pain phenotype, using a cutoff of 12 or greater.¹⁶  Finally, the Pain Self-Efficacy Questionnaire was used to evaluate the impact of chronic pain on the respondent’s life. This is a 10-item questionnaire that addresses the belief and confidence that patients with chronic pain have in performing daily activities, including household chores, socializing, and work—a cutoff score of 40 or less indicates low self-efficacy.¹⁷  Patients who failed to respond within 1 month were contacted by telephone to complete the questionnaires.

Only patients who provided complete outcome data at one or more follow-up visits were included in the complete case analysis. Descriptive statistics were used to summarize the cohort’s demographic, clinical, and psychologic characteristics. The maximum postoperative pain intensity was determined as the highest pain score at any timepoint within the first 5 postoperative days. Pain scores at follow-up were categorized into none (0), mild (1 to 3), moderate (4 to 6), and severe (7 to 10). We also sought to determine the proportion of patients whose pain was likely neuropathic (S-LANSS 12 or greater) and whether the presence of neuropathic pain was associated with more severe pain and worse quality of life compared to nonneuropathic pain.

The simple linear regression coefficient (β) was used to assess the association between perioperative variables and continuous pain outcomes (0 to 10). Multivariable linear regression models were then developed to assess the impact of several explanatory variables. Covariates were chosen for inclusion based on clinical likelihood and biologic plausibility, and inserted into the model using simultaneous entry. We followed the recommended rule of 10 events per variable in regression modeling.¹⁸  To control for the influence of multicollinearity among variables, the variance inflation factor was calculated, with an inclusion cutoff of less than 2. Power calculations performed before data unmasking confirmed that the sample size achieved greater than 90% power to detect small, medium, and large effect sizes using an α level of 0.05. To assess the robustness of our main analysis, we performed multiple imputations using chain equations with 10 imputations in a sensitivity analysis to impute both missing outcomes (pain scores) and missing covariates simultaneously. Finally, in an exploratory analysis to identify factors associated with chronic pain of predominately neuropathic origin, we used multiple linear regression to determine variables associated with S-LANSS scores at 3 months (continuous outcome, 0 to 24). All data analyses were performed using STATA 14.1 (StataCorp LP, USA), with a P value <0.05 considered statistically significant.

A total of 1,059 patients were enrolled in the study (fig. 1). The questionnaire response rates at 3, 6, and 12 months after surgery were 63% (n = 663), 59% (n = 625), and 57% (n = 605), respectively. A total of 767 patients (72%) provided complete outcome data at one or more follow-ups and were included in the complete case analysis.

Table 1 reports the baseline characteristics of the included sample. Patients were on average (mean ± SD) 61 ± 13 yr old and predominantly male (77%) with a mean body mass index of 29 ± 11 kg/m². Hypertension was the most prevalent comorbidity (60%), followed by osteoarthritis (40%) and coronary artery disease (37%). Approximately 15% of patients had a history of previous cardiac surgery, 7% were active tobacco smokers, 11% consumed alcohol, and more than a third of patients reported baseline chronic pain (38%) with an average pain intensity of 4.4 ± 2.2. The most common pain locations were low back (27%), lower extremities (22%), and head and neck (17%), with only 7% reporting chronic pain in the chest or thorax. Approximately 39% of patients reported using pain medication before surgery, with the most common drugs being acetaminophen (14%), ibuprofen (8%), and Tylenol #3 (4%; Janssen, Toronto, Canada).

The two most common cardiac procedures were valvular replacement surgery (455 [59%]), followed by CABG surgery (287 [37%]). Saphenous vein grafts were used in 239 patients (83%) who underwent CABG surgery. Approximately 7% of patients underwent aortic repair, and 6% had a procedure on their myocardium. No patients in the cohort underwent heart transplantation. The average surgical procedure lasted 210 ± 80 min, and the average time on cardiopulmonary bypass was 101 ± 45 min.

None of the patients received an epidural for postoperative pain, and only one patient received a regional anesthetic block. All patients were followed by the acute pain service team after surgery, with a mean follow-up of 2 ± 1 days. Approximately 37% of patients were provided patient-controlled analgesia pumps with either intravenous hydromorphone or morphine. Most patients received acetaminophen (99%) after surgery, and 25% of patients received nonsteroidal anti-inflammatory drugs. The mean pain severity across the first 5 postoperative days was 3.8 ± 1.8, with 63% reporting average scores in the moderate–severe range (4 or more out of 10). The mean opioid consumption during the first 5 days after surgery was 240 ± 323 morphine milligram equivalent.

Table 2 summarizes the incidence, timing, and severity of chronic chest and leg pain at 3, 6, and 12 months after surgery. At 3 months, 191 patients (29%; 95% CI, 26 to 32%) reported pain (greater than 0 out of 10) in the chest area, which decreased to 19% (95% CI, 16 to 22%) at 6 months, and 15% (95% CI, 12 to 18%) at 12 months postoperatively (fig. 2A). However, despite a decreasing prevalence of chest pain over time, the intensity of pain remained relatively stable across the 3-, 6-, and 12-month follow-ups (pain scores 3.2 ± 1.9, 3.2 ± 2.0, and 2.9 ± 1.9, respectively, P > 0.05). Likewise, the unpleasantness of pain at the chest incision site remained relatively unchanged over time (pain scores 3.4 ± 2.1, 3.3 ± 2.1, and 3.4 ± 2.3 at 3, 6, and 12 months, respectively, P > 0.05). However, among those reporting any pain, the incidence of moderate-to-severe pain (4 or greater out of 10) declined slightly over time from 34% (n = 65) at 3 months to 35% (n = 41) at 6 months and 27% (n = 24) at 12 months. At 3 months, 63% of patients described their pain to be deep within the muscles or joints of the chest, 57% had pain at the incision site, and 37% experienced pain in the skin around the incision. Most participants experienced pain episodes once a week (44%) or once an hour (40%), with each episode lasting seconds or minutes (65%). The use of analgesics was comparable at 3 (39%), 6 (40%), and 12 (42%) months after surgery. Approximately one in five patients scored positive for poor self-efficacy (Pain Self-Efficacy Questionnaire less than 40) at any one follow-up timepoint. The Pain Self-Efficacy Questionnaire scores remained stable over time (49 ± 13 at 3 months, 49 ± 13 at 6 months, and 51 ± 10 at 1 yr; P > 0.05).

Notably, among patients reporting any pain, the incidence of neuropathic pain, identified on the S-LANSS questionnaire, increased over time from 34% at 3 months to 64% at 12 months. Compared to nonneuropathic pain, patients with neuropathic pain reported more intense pain (4.0 vs. 2.8, P = 0.0004, fig. 2B) and had poorer self-efficacy at 3 months (34% vs. 17%, P = 0.012). Further, 49% of patients experiencing neuropathic pain at 3 months continued to report chronic pain of any nature at 1 yr, whereas only 30% of patients with nonneuropathic pain at 3 months had persistent chronic pain of any nature at 1 yr (P = 0.042).

Of the 239 patients who underwent saphenous vein grafts, the incidence of pain and numbness around the leg scar at 3 months was 12% and 28%, respectively. By 1 yr, 13% of patients still experienced pain in their legs, and 28% still experienced numbness. Among participants reporting leg pain, the rate of concurrent chronic chest pain decreased from 57% at 3 months to 29% at 1 yr.

In the univariable analysis, patients with chronic pain at 3 months were younger (P = 0.001), more likely to be female (P < 0.001), and more likely to have undergone previous heart surgery (P = 0.006). Both groups were similar in clinical measures including body mass index and other comorbidities such as smoking history, diabetes, and arthritis. Patients with pain at 3 months more often reported a history of angina (P = 0.003), baseline chronic pain (P < 0.001), and higher preoperative pain scores (P < 0.001). With respect to psychologic factors, patients with greater presurgical anxiety and depression (per 1 unit increase in HADS-A or HADS-D, P < 0.001) and greater pain catastrophizing scores (P < 0.001) were at a higher risk of chronic pain. Similarly, patients who developed chronic pain were more likely to have been using psychotropic drugs at baseline (anxiolytics [P = 0.006] and antidepressants [P = 0.003]). Patients with chronic pain had greater acute pain scores (P < 0.001), had higher cumulative opioid consumption (P = 0.004), and were more likely to receive patient-controlled analgesia for self-administration of opioids (P < 0.001). All unadjusted explanatory factors of chronic pain at 3 months, except angina and preoperative chronic pain, continued to be associated with chronic pain outcomes at 6 and 12 months.

In the multivariable models, six perioperative factors were associated with postsurgical pain scores at 3 months including female sex (β = 0.65, P < 0.001); previous cardiac surgery (β = 0.55, P = 0.004); pre-existing chronic pain (β = 0.46, P = 0.001); baseline HADS depression subscores (β = 0.08, P = 0.003); pain catastrophizing scores (β = 0.02, P = 0.004); and moderate-to-severe acute pain (4 or greater out of 10) within 5 postoperative days (β = 0.43, P = 0.005, table 3). The same factors, with the exception of previous cardiac surgery and baseline chronic pain, continued to be correlated with chronic pain scores at 6 months (appendix 1). Only three factors remained associated with chronic pain scores at 12 months, namely female sex, baseline chronic pain, and higher HADS depression subscores (appendix 2).

In order to untangle the contribution of postoperative factors from preoperative factors that may offer greater utility for preventative measures, both postsurgical factors, namely acute postoperative pain and opioid use after surgery, were removed from the multivariable models. Their exclusion had no impact on the statistical significance of the remaining preoperative factors such as baseline pain and presurgical depression, which continued to be associated with chronic postoperative pain scores at 3, 6, and 12 months (appendix 3). Finally, in an exploratory analysis to identify factors associated with the S-LANSS score, we found that female sex (β = 2.79, P = 0.009) and baseline HADS anxiety subscores (β = 0.36, P = 0.016) were associated with higher neuropathic pain scores at 3 months.

In the multiple imputation models accounting for the potential effect of missing data, the main analysis was replicated with a few exceptions. We continue to see the same significant association between female sex, pre-existing chronic pain, previous cardiac surgery, pain catastrophizing, baseline depression, and acute postoperative pain with chronic pain scores at 3 months. In addition, at 6 months, two additional factors that trended toward an effect in the main model, baseline chronic pain and previous cardiac surgery, reached statistical significance in the imputed model. Finally, cumulative analgesic consumption within the first 5 postoperative days (a factor that is significant univariate, but not significant multivariate in the complete-case analysis) was found to be independently associated with chronic pain scores at 1 yr (appendix 4).

This prospective cohort study followed patients at consecutive timepoints up to 1 yr after cardiac surgery. Our results confirm findings from previous studies and provide novel insights into the characteristics and impact of postsurgical pain on a sizable cardiac surgery cohort. The multivariable regression confirmed several previously identified perioperative factors associated with postsurgical pain, including female sex, baseline pain, preoperative depression, and higher acute pain scores within the first 5 postoperative days. This study has several strengths, including its prospective design, large sample size, comprehensive range of risk factors examined, long-term follow-up, and the use of validated tools for symptom assessments. Our broad inclusion criteria allow us to generalize our results to all open-heart surgery patients.

In our sample, nearly one third of patients developed pain at 3 months, decreasing to 19% at 6 months and 15% at 1 yr. This is comparable to the Canadian CARD-PAIN study, which reported pain prevalence rates of 40%, 22%, and 17% at 3, 6, and 12 months after surgery, respectively.⁹  Despite both studies reporting a 50% resolution of pain by 12 months, the 1-yr prevalence of pain is still concerning. First, these rates are significantly higher than the incidence reported after noncardiac surgeries, where an average of 3% of patients suffer from incisional pain after 1 yr.¹⁹  Second, considering that greater than 2 million patients undergo cardiac surgery annually worldwide,¹  more than 300,000 new patients will report some degree of chronic pain at 1 yr postoperatively. The increasing number of cardiac surgeries represents a major public health issue with economic consequences and poses a significant impact on patients’ quality of life.^4,20–22 

Pain intensities associated with chronic pain are highly variable. It is important to note that most patients experience mild pain (64 to 71%), and few report severe pain (2 to 5%). Further, more than half of chronic pain patients described their pain as deep to the sternal incision (muscle or joint), consistent with the high occurrence of musculoskeletal pain reported in poststernotomy patients.²³  The wide variability in the incidence and severity of chronic pain may be explained by discrepancies between the study’s methodology (e.g., the timing of outcome measurement, duration of follow-up) and the lack of a standardized definition for chronic postsurgical pain, which was only recently provided by the International Association for the Study of Pain task force in 2019.¹⁵ 

One of the objectives of this study was to identify the proportion of patients with features of neuropathic pain using the S-LANSS questionnaire. Interestingly, we found that among patients reporting any chronic pain, the proportion of pain compatible with a neuropathic phenotype increased from 34% at 3 months to 39% at 6 months, and 64% at 1 yr. This was in part due to the differential resolution of pain between patients with and without neuropathic features—only 51% of patients with neuropathic pain at 3 months had complete pain resolution at 1 yr, whereas nearly 70% of patients whose pain was not compatible with a neuropathic phenotype at 3 months were found to be pain-free at 1 yr (P = 0.042). Furthermore, patients with neuropathic pain reported higher pain intensity and poorer self-efficacy scores. These findings suggest that chronic pain with neuropathic features is more challenging compared to nonneuropathic pain in terms of severity, impact, and prognosis. Neuropathic features are indicative of neuroplastic changes that occur within the peripheral and central nervous system (e.g., activation of N-methyl-d-aspartate receptors centrally, the release of excitatory neurotransmitters along pain pathways, and spinal microglial activation), which leads to pain amplification, enhanced sensitivity to pain, and nociceptive signaling in response to non-painful stimuli.^24,25  Once neuroplastic changes occur within the nervous system, reversal is challenging. Given the refractory nature of neuropathic pain disorders, greater assessment and management efforts administered earlier in the postoperative period should be directed toward neuropathic pain. Ongoing work is exploring whether early interventions guided by Transitional Pain Services can potentially change the poor prognostic trajectory of patients who develop neuropathic symptoms in the early postoperative period.^26,27 

Several studies have attempted to establish risk factors for chronic pain in cardiac surgery patients, with variable results.^9,28–30  Our study confirms and extends a number of demographic, clinical, and psychologic factors that have been understudied. In line with previous studies, we found that female sex was associated with chronic pain scores at 3, 6, and 12 months. Sex differences in pain perception are well-established, and the underlying mechanisms contributing to this disparity are thought to be an interaction of multiple biopsychosocial mechanisms, including sex hormones, endogenous opioid function, pain coping, and gender roles.³¹  In addition, we found that a history of chronic pain was strongly correlated with chronic postsurgical pain scores across all three timepoints. This is in keeping with the well-established phenomenon that pain begets pain and relates to the aforementioned neuroplastic changes that occur with chronic pain syndromes.^32,33  To date, the CARD-PAIN study is the only prospective examination that has reported an association between pre-existing chronic pain with postsurgical pain in the cardiac surgery population.⁹  This indicates that patients with baseline pain undergoing cardiac surgery may be at increased risk for developing persistent postsurgical pain, and should be considered for interdisciplinary multimodal pain management options throughout the perioperative period.

Presurgical depression was strongly correlated with postsurgical pain scores at 3, 6, and 12 months. While psychologic factors have been found to be associated with chronic pain outcomes after orthopedic and breast surgeries, this is the first study to demonstrate an association between presurgical depression and chronic pain after cardiac surgery. Overall, considering the vast amount of literature that has shown the role of psychologic factors in the development of persistent pain after surgery,³⁴  effective management of preoperative psychologic vulnerabilities may be crucial for preventing chronic pain. Interventional trials are needed to confirm whether psychologic counseling, education, and medical treatment before surgery can improve long-term outcomes in relevant patients.

Our study identified that moderate-to-severe acute pain in the first 5 postoperative days was associated with chronic pain scores at 3 and 6 months. Worse acute pain scores after surgery have been consistently linked to worse long-term pain outcomes and have been replicated in many surgical cohorts, including hernia, breast, and thoracic surgery.³⁵  Considering that every chronic pain was once acute,³⁵  and that neuropathic pain tends to be persistent, these observations may reflect an early manifestation of nerve injury with associated immediate hyperalgesia. Therefore, our results suggest that inadequate management of acute pain may increase the risk of chronic pain, and efforts to better manage acute postoperative pain, including multimodal opioid-sparing regimens, early involvement by the acute pain service, and follow-up with the transition pain clinic may be beneficial in chronic pain prevention.

It is important to note that acute postoperative pain score is a “late” risk factor and potentially less useful compared to preoperative risk factors that can be utilized in presurgical risk stratification to help determine an individual’s risk and need for preventative measures. We recognize that severe acute postoperative pain may be attributed to a combination of baseline susceptibilities and inadequate postoperative pain management, and therefore may not constitute a true “predictive” risk factor but rather is an early manifestation of chronic postsurgical pain outcomes. Therefore, we ran multivariable models excluding pain and opioid use in the acute postoperative period and found that tenably modifiable factors such as presurgical depression continued to remain associated with long-term postsurgical pain scores and may be of interest in future prediction tools.

There are several limitations to this study. Although we attempted to examine the association of relatively unexplored aspects of perioperative care with chronic postsurgical pain scores, many specific demographic, clinical, psychologic, and genetic factors remain unaddressed. Moreover, it would have been informative to assess the prevalence of postoperative opioid use, which could be used to help understand the relationship between opioid consumption and postoperative pain scores. Additionally, despite enrolling slightly more than a thousand patients, a lower sample was included in the analysis due to incomplete data and loss to follow-up. This raises the possibility of chance findings, and reduced statistical power, although we have performed a sensitivity analysis to impute both missing outcomes and covariates. Furthermore, this study relied on patients’ self-reports on mailed surveys, which may be prone to misinterpretation. Additionally, survey responses introduced the possibility of selection bias, where only those who are capable (fewer postoperative complications or less debilitating pain) would have responded. Finally, clinical evaluations, such as in-person examinations or quantitative sensory testing, were not performed, which would have assisted in ruling out other causes of chest wall pain and obtaining clinically definitive features of neuropathic pain (e.g., evidence of allodynia or hyperalgesia).

In summary, our study utilizing a large cardiac surgery sample and a robust process for imputing missing data demonstrated that persistent pain is a common and problematic complication of cardiac surgery. Several of the consistently reported risk factors have been confirmed in our study including female sex, baseline chronic pain and depression, and increased acute postoperative pain. Intuitively, optimization of patients undergoing cardiac surgery with baseline risk factors in addition to improved postoperative pain management may mitigate the risk of chronic pain outcomes. However, this effect will need to be explored in future randomized controlled trials in this patient population.

Support was provided solely from institutional and/or departmental sources.

Dr. Katz is the Canada Research Chair in Health Psychology (Canadian Institutes of Health Research, Ottawa, Ontario, Canada). Dr. Rao does consultation work for Medtronic (Oakville, Canada), Abbott Medical (Mississauga, Canada), and Gore Medical (Mississauga, Canada). The other authors declare no competing interests.

Questionnaire: Postoperative Pain after Cardiac Surgery, https://links.lww.com/ALN/D158