Article Text

Download PDFPDF

Risk factors for severe acute pain and persistent pain after surgery for breast cancer: a prospective observational study
  1. Ashraf S Habib1,
  2. Miklos D Kertai2,
  3. Mary Cooter1,
  4. Rachel A Greenup3 and
  5. Shelley Hwang3
  1. 1 Department of Anesthesiology, Duke University, Durham, North Carolina, USA
  2. 2 Department of Anesthesiology, Vanderbilt University, Nashville, Tennessee, USA
  3. 3 Department of Surgery, Duke University, Durham, North Carolina, USA
  1. Correspondence to Ashraf S Habib, Department of Anesthesiology, Duke University Medical Center, Durham 27705, NC, USA; ashraf.habib{at}duke.edu

Abstract

Background and objectives There are few prospective studies providing comprehensive assessment of risk factors for acute and persistent pain after breast surgery. This prospective observational study assessed patient-related, perioperative, and genetic risk factors for severe acute pain and persistent pain following breast cancer surgery.

Methods Women presenting for elective breast cancer surgery completed State Trait Anxiety Inventory, Beck Depression Inventory, and Pain Catastrophizing Scale questionnaires preoperatively. Diffuse noxious inhibitory control and mechanical temporal summation were assessed. A blood sample was obtained for genetic analysis. Analgesic consumption and pain scores were collected in the post-anesthesia care unit, and at 24 and 72 hours. Patients were contacted at 1, 3, 6, and 12 months to assess persistent pain. Primary outcome was maximum acute pain score in first 72 hours and secondary outcome was persistent pain.

Results One hundred twenty-four patients were included in analysis. Increased duration of surgery, surgeon, and higher pain catastrophizing scores were associated with increased severity of acute pain, while preoperative radiotherapy was associated with reduced severity. Persistent pain was reported by 57.3% of patients. Postdischarge chemotherapy (OR 2.52, 95% CI 1.13 to 5.82), postdischarge radiation (OR 3.39, 95% CI 1.24 to 10.41), severe acute pain (OR 5.39, 95% CI 2.03 to 15.54), and moderate acute pain (OR 5.31, 95% CI 1.99 to 15.30) were associated with increased likelihood of persistent pain.

Conclusions Increased duration of surgery, higher pain catastrophizing score, and surgeon were associated with increased severity of acute pain. Preoperative radiation was associated with lower acute pain scores. Postsurgery radiation, chemotherapy, and severity of acute pain were associated with increased likelihood of persistent pain.

Trial registration NCT03307525.

  • breast surgery
  • acute postoperative pain
  • persistent pain
  • quantitative sensory testing

Statistics from Altmetric.com

Introduction

Breast surgery is one of the most frequently performed procedures in the USA. A substantial number of women who undergo surgery for breast cancer develop chronic pain, with an incidence of about 50% or more in some studies,1–4 leading to limited functionality and psychological distress.5 6 Such chronic pain continues to be a major humanitarian and socioeconomic burden with slow progress in the development of new preventive and therapeutic options.

Similar to other types of surgery, the severity of acute pain after breast surgery is a significant predictor for the development of persistent pain.7 8 While several studies have investigated risk factors for pain after breast surgery,4 9–12 few have provided a comprehensive evaluation of all risk factors in a prospective manner across the entire continuum of perioperative care. Further, few studies have evaluated potential genetic predictors of pain in this patient population. Indeed, it has been shown that certain genetic variations could potentially influence both the onset of acute pain and the development of persistent postsurgical pain.13 Thus, a better understanding of individual variability in patient characteristics of postsurgical pain (eg, genetic, surgical, and psychological) will help identify a high-risk subset of patients who are likely to develop severe acute pain and persistent pain, and also promote the emerging concept of the role of precision medicine in postoperative and long-term pain management of patients who underwent breast surgery. Targeting those high-risk women with more aggressive perioperative multimodal analgesic management strategies might lead to improvement in the management of acute pain and a potential reduction in the incidence of persistent postsurgical pain. For instance, in patients undergoing total knee arthroplasty, pregabalin initiated preoperatively and continued for 14 days postoperatively was effective in reducing the incidence of chronic neuropathic pain compared with placebo.14

A number of previous studies have attempted to investigate quantitative sensory testing of patients’ baseline pain perception to identify those at risk for acute postoperative pain and chronic pain.15 16 A systematic review by Werner et al reported that quantitative sensory testing may predict 4% to 54% of the variance in acute postoperative pain experience.15 The role of quantitative sensory testing has not however been widely studied in women undergoing breast surgery.

We therefore performed this prospective observational study to evaluate risk factors for severe acute pain and persistent pain in women undergoing surgery for breast cancer, collecting detailed patient and treatment variables, perioperative variables, and targeted allelic data. We specifically hypothesized that quantitative sensory tests would predict the severity of acute pain and the occurrence of persistent pain in this patient population.

Methods

After Institutional Review Board approval, consecutive English-speaking women presenting for elective breast cancer surgery were recruited into the study from June 2012 to November 2015. Patients with recurrent breast cancer, metastasis, previous breast surgery, history of chronic pain conditions or regular intake of opioid analgesics, and substance abuse history were excluded from the study.

During the preoperative screening visit, participants were asked to complete the State Trait Anxiety Inventory (STAI),17 Beck Depression Inventory (BDI),18 19 and the Pain Catastrophizing Scale (PCS)20 questionnaires. Diffuse noxious inhibitory control (DNIC) was measured using the Medoc-TSAII (Medoc, Ramat-Yishai, Israel) according to the method described by Granot et al.9 This involved assessing pain scores in response to a hot thermode applied to the patient’s dominant forearm and repeating the test after immersion of the contralateral hand in hot water. Temporal summation was evoked by Von Frey filaments using a 225.1 g (No. 6.45) filament applied 11 times to the volar aspect of the dominant forearm, using the method described by Weissman-Fogel et al. 21 Mechanical temporal summation score was calculated by subtracting the pain rating (0–100 verbal numerical pain scale) after the first touch from the pain rating after the 10th touch. If the rating of the 10th touch was higher than that of the first touch, this was considered mechanical temporal summation positive. Details of both tests are provided as online supplementary material 1 (Supplemental Digital Content 1).

Supplemental material

Data collection and endpoint definition

We collected data on age, race, education, marital status, presence and severity of preoperative breast pain, surgery type and duration, axillary dissection, sentinel node biopsy, surgeon performing the surgery, and history of chemotherapy or radiation therapy. Details of the anesthetic technique and analgesic consumption in the operating room and post-anesthesia care unit were collected. All opioids were converted to morphine milligram equivalents for analysis.22 Verbal rating pain scores (0–10 scale, 0=no pain, 10=worst possible pain) and details of analgesic consumption were also collected at 24 and 72 hours after end of surgery. Patients were contacted by phone at 1, 3, 6, and 12 months after surgery to collect information about persistent pain using a validated questionnaire for breast cancer persistent pain which details site, severity (0–10 scale) and frequency of pain, treatment used, and associated sensory disturbance or discomfort.1 We also assessed the interference of pain with activities of daily living using the Brief Pain Inventory short form at each of those time points.23

Candidate gene genotyping

A blood sample was obtained from all patients enrolled in the study for genotyping of single-nucleotide polymorphisms (SNPs). Genomic DNA was isolated from the individual blood samples using standard procedures. Genotyping was performed using TaqMan SNP genotyping assays. Given the size of our study cohort and the concern of multiple testing, we only selected five SNPs within a set of two candidate genes, CACNA2D3 (rs1851048, rs6777055) and CACNG2 (rs2284015, rs22841017, rs4820242), based on the current understanding of the role of these candidate genes in the development of persistent pain after surgery.13

Statistical analysis

We estimated that the sample size of 125 would have 80% power with alpha=0.05 to detect a Pearson correlation coefficient of at least 0.245 between the continuous predictors of magnitude of DNIC and mechanical temporal summation, and the severity of acute pain. With this sample size, we also estimated that the logistic regression test of no effect of acute pain (alpha=0.05, two sided) on chronic pain will have 80% power to detect an OR of 1.84 with an incidence of chronic pain of 35% and OR of 1.74 with an incidence of chronic pain of 50%.

At the conclusion of follow-up, we examined the long-term response patterns among study participants. For missing data, we imputed the nearest available subsequent response from the 3, 6, or 12 months’ follow-up, and if no subsequent response was available, we used the nearest prior response. We performed follow-up sensitivity analysis to ensure our imputation strategy did not bias our conclusions.

We analyzed maximum patient-reported acute pain score over the first 72 hours as a numeric outcome (primary outcome) to study factors associated with increased acute pain following breast cancer surgery. We also studied persistent pain as a secondary dichotomous outcome. We defined persistent pain as reported pain score ≥3 or pain that impacts daily life as indicated by a score >0 on any of the seven questions assessing impact of pain on daily living in the Brief Pain Inventory short form at either the 6-month or 12-month follow-up.

We summarized patient and clinical characteristics according to the maximum acute pain level (mild (maximum acute pain score 0–3), moderate (maximum acute pain score 4–6), and severe (maximum acute pain score 7–10)) using mean (SD) or median (Q1, Q3) for numeric variables and count (%) for categorical variables. The univariate association of factors with the numeric acute pain outcome was assessed via Spearman correlation for numeric factors and Wilcoxon rank-sum tests for categorical factors.

Patient and clinical factors found to be associated with acute pain score in univariable analysis at the 0.10 level were considered for inclusion in a stepwise variable selection algorithm to build a multivariable linear regression model. Model residual plots and diagnostics were used to evaluate modeling assumptions. When we evaluated the functional from of our factors for linearity, we identified a non-linear pattern for PCS score. After investigating the distribution of PCS scores and their relationship with maximum acute pain, we chose to define PCS as a categorical variable. We explored a number of alternate cut points for analyzing PCS (such as 5-point intervals, 10-point intervals, median split, quartile split) and then selected the optimal cut points that minimized the model Akaike information criterion and maximized the adjusted R2 value. The selected cut points were scores of 10 and 20, which divided the patients into three levels of PCS scores for analysis.

For the secondary outcome, persistent pain at 6 or 12 months postsurgery, we again summarized patient and clinical characteristics by outcome group (persistent pain or no persistent pain). To identify baseline patient and clinical factors, and any postsurgery factors that may be associated with the persistent pain outcome, we performed univariable t-tests or Wilcoxon rank-sum tests for numeric factors, and χ2 or Fisher’s exact tests for categorical factors as appropriate. Patient and clinical factors found to be associated with persistent pain in univariate analysis at the 0.10 level were considered for inclusion in a stepwise variable selection algorithm to build a multivariate linear regression model. Model residual plots and diagnostics were used to evaluate modeling assumptions. When we evaluated the functional form of our factors for linearity, we identified a non-linear pattern for maximum acute pain score. For the persistent pain analysis, we categorized acute pain into mild (pain score 0–3), moderate (pain score 4–6), or severe pain (pain score 7–10) categories and considered it as a factor that could be associated with the persistent pain outcome.

We found that postdischarge treatments were significantly associated with our secondary persistent pain outcome. In order to account for the impact of postdischarge treatment on the incidence of persistent pain when estimating the association of baseline factors with the outcome of interest, we forced terms for these treatments into our multivariate model for persistent pain.

Following the additive model for minor allele frequency, allelic associations with acute pain (primary outcome) and persistent pain (secondary outcome) were analyzed in a univariable analysis using Spearman correlation or Wilcoxon rank-sum tests for each of the five SNPs of the two candidate genes. Subsequently, separate multivariable linear regression analyses were implemented to adjust genetic association tests for patients’ demographic characteristics, and clinical and procedural covariates. Potential susceptibility SNPs of the candidate genes with two-tailed p value <0.05 were considered statistically significant.

Results

Two hundred forty-four subjects were approached for participation in the study, 103 declined participation, 12 withdrew consent prior to surgery, and 4 were screen failures. One hundred twenty-five patients were enrolled in the study, one withdrew after surgery and 124 contributed postoperative pain scores for our primary and secondary outcome analyses (figure 1). Of those, 86 (69%) were Caucasians, 81 (74%) had a college or higher degree, 22 (18%) reported preoperative breast pain, and 14 (11%) were using regular analgesics. Surgery was bilateral in 25 patients (20%). Fifty-seven percent of patients underwent a partial mastectomy, 23% simple mastectomy, 10% modified radical mastectomy, and 10% mastectomy with reconstruction. Five surgeons performed the surgeries, and as three of them treated 94% of patients, we combined the two low-count surgeons into a single group to create a four-level surgeon variable for analysis. Axillary dissection was performed in 19 patients (15%) and sentinel node biopsy in 73 patients (59%). Thirty-five (28%) received preoperative chemotherapy and seven (6%) received preoperative radiotherapy. Sixty-seven patients (54%) received a paravertebral block. Thirty-seven patients (30%) had further surgery, 67 (54%) received radiotherapy, and 29 patients (23%) received postsurgery chemotherapy.

Acute pain analysis

Over the first 72 hours, 40 patients complained of mild pain, 40 of moderate pain, and 44 patients complained of severe pain. Patient demographics, response to preoperative questionnaires and quantitative sensory tests, and intraoperative anesthetic and surgical details are summarized in table 1. Except for age (r=−0.24; p=0.006), there were no significant associations of patient demographics with maximum pain score over the first 72 hours, with younger patients experiencing higher severity of acute pain (p=0.006). There was no association of preoperative breast pain with maximum acute pain score, but there was a trending association with preoperative receipt of radiotherapy (p=0.06) and chemotherapy (p=0.08). While there was no evidence of an association of acute pain with mechanical temporal summation, DNIC scores, or STAI scores, there were significant associations with BDI (r=0.23; p=0.009) and PCS (r=0.20; p=0.024). Intraoperative details were significantly associated with maximum acute pain with significant differences found in site, type, and duration of surgery. Maximum acute pain scores were also significantly different by anesthetic details with regards to the use of general anesthesia, paravertebral blocks, and intraoperative opioid consumption (table 1).

Table 1

Patient characteristics by maximum acute pain category

The following terms, which were univariately associated at the 0.10 level, were considered in building the multivariable acute pain outcome model: previous chemotherapy, previous radiotherapy, type of surgery, site of surgery, duration of surgery, operating surgeon, dose of intraoperative opioids in mg morphine equivalents, paravertebral block, general anesthesia, BDI score, and total PCS score. In the final model (R2=0.293), duration of surgery, surgeon, preoperative radiotherapy, and PCS scores were significantly associated with the severity of acute postoperative pain (table 2). Patients with longer surgery duration or higher PCS scores had higher maximum acute pain scores, and patients with previous radiotherapy had lower maximum pain scores. We also found significant variation in maximum acute pain scores between surgeons. We investigated all two-way interactions and found evidence of significant interaction between surgeon and surgery duration on maximum acute pain score indicating that the impact of surgery duration varied by surgeon. In the model including the interaction, preoperative radiotherapy and PCS scores were still significantly associated with the severity of acute postoperative pain.

Table 2

Multivariable model of predictors of maximum acute postoperative pain score

Persistent pain analysis

For the persistent pain analysis, we imputed data for 22 patients (17%). Those patients could not be contacted at one or more of the follow-up phone calls within the follow-up time windows. After imputation, we identified 71 patients (57.26%) with persistent pain. Of the patient demographics and clinical characteristics we studied (table 3), we found that only use of general anesthesia (65% with vs 45% without pain, p=0.030) and maximum acute pain (median score 5 with vs 4 without pain, p=0.005) were significantly associated with persistent pain. In addition, postdischarge chemotherapy (76% with vs 52% without pain, p=0.021) and radiation (66% with vs 47% without pain, p=0.040) were found to be associated with the incidence of persistent pain.

Table 3

Patient characteristics by persistent pain

To build the multivariable model for persistent pain (table 4), we forced postdischarge chemotherapy and radiation into the model and then considered the following terms which were univariately associated at the 0.10 level, in stepwise selection: age, acute pain score (as three-level categorical variable), general anesthesia, and 24-hour acetaminophen use. In the final model (C-index=0.76), the only additional term selected after forcing inclusion of postdischarge chemotherapy and radiation was maximum acute pain category. The odds of experiencing persistent pain were 2.5 times higher for patients who had postdischarge chemotherapy (OR 2.52, 95% CI 1.13 to 5.82; p=0.026) and 3.4 times higher for patients who had postdischarge radiation therapy (OR 3.39, 95% CI 1.24 to 10.41; p=0.023). Patients who experienced severe acute pain had 5.4 times higher odds for developing persistent pain (OR 5.39, 95% CI 2.03 to 15.54; p=0.001), and patients who experienced moderate acute pain had 5.3 times higher odds for developing persistent pain (OR 5.31, 95% CI 1.99 to 15.30; p=0.001) compared with patients with mild acute pain.

Table 4

Final multivariable model of predictors of persistent postoperative pain

We next performed a sensitivity analysis for the persistent pain outcome using only the observed follow-up data, rather than the nearest available when data were missing. We used the multivariable model defined in the previous analysis and applied it to the sensitivity cohort of 102 patients of whom 65 (63.7%) reported persistent pain. The model indicated that all the effects found to be significant in our imputed data analysis remained significant in the observed data model, and the model continued to fit well (C-index 0.76). The odds of experiencing persistent pain was elevated for patients receiving postoperative chemotherapy (p=0.041) and/or radiation (p=0.015). Those with moderate or severe maximum acute pain also had higher odds of persistent pain compared with patients with mild maximum acute pain (p=0.003 and 0.023, respectively).

Univariable analysis indicated that none of the SNPs tested were significantly associated with either acute pain (online supplementary table 1/SDC 2) or persistent pain (online supplementary table 2/SDC 3). For the acute pain analysis, adding terms to the final multivariable model found no evidence of any SNP associated with the outcome. For persistent pain, adding the SNPs to the final multivariable model indicated that SNP rs2284017 in CACNG2 gene showed the strongest association with persistent pain, but failed to reach statistical significance (OR 1.61, 95% CI 0.96 to 2.78; p=0.08).

Supplemental material

Supplemental material

Discussion

In this observational study, 68% of women complained of moderate to severe pain in the first 72 hours after surgery, and 57% developed persistent pain. Baseline quantitative sensory tests did not independently predict the severity of acute pain or the occurrence of persistent pain. Longer duration of surgery, higher PCS scores, and surgeon were associated with increased severity of acute pain while preoperative radiotherapy was associated with lower acute pain scores. In addition, postsurgery radiation therapy, chemotherapy, and the severity of acute pain were associated with increased likelihood of persistent pain.

The variation in pain experience after similar types of surgery might be due to differences in central processing of peripheral pain stimuli. Tests of conditioned pain modulation such as DNIC are suggested to assess the efficiency of the descending inhibitory pathways. Lower DNIC efficiency was reported in patients with persistent pain after thoracotomy.24 Results of other studies assessing the use of conditioned pain modulation as a predictor of postsurgical pain have however produced inconsistent results, with some also showing negative associations,25 26 while others reporting no associations with acute or chronic pain.27–30 We also found no evidence that DNIC was associated with the severity of acute pain or the occurrence of persistent pain in our study. Furthermore, the use of DNIC was time consuming and was challenging to perform in this group of women with high level of preoperative anxiety.

Temporal summation is considered representative of the physiological wind-up phenomenon taking place at the spinal level of central pain pathways. Mechanical temporal summation was found to predict acute postoperative pain after thoracotomy,21 and persistent pain in patients undergoing hysterectomy31 and knee replacement.27 On the other hand, temporal summation to electric and heat pain was not associated with acute pain after percutaneous nephrolithotomy32 and thoracotomy.21 We also did not find that mechanical temporal summation was predictive of acute or persistent pain in women undergoing breast surgery. A recent systematic review reported no consistent association between preoperative quantitative sensory tests and pain outcomes after surgery.16 The review, however, suggested that quantitative sensory tests related to central pain mechanisms such as mechanical temporal summation and conditioned pain modulation were more frequently associated with pain outcomes that those quantitative sensory tests assessing pain detection and thresholds.

Similar to previous studies in women undergoing breast surgery33–35 and other surgical patient populations, the severity of acute postoperative pain was a significant predictor of persistent pain. This suggests that we need to focus our efforts on optimizing postoperative pain management in this patient population and improve counseling of these patients regarding the risk of severe acute pain and persistent pain as they make decisions for breast cancer surgery. Local anesthetic techniques such as paravertebral blocks might reduce the severity of acute pain, but currently there are insufficient data to assess their impact on persistent pain.36 While univariate analyses suggested that the anesthetic technique might impact acute and persistent pain in our study, those effects were not statistically significant in the multivariable analyses. Limited data on gabapentanoids suggest that they might improve acute pain management in this patient population,37 but more data are needed to assess impact on acute and persistent pain.

We have identified radiotherapy as being a significant predictor for persistent pain and PCS scores as being a significant predictor of acute pain, which is similar to what has been previously reported in this patient population.38 39 Preoperative radiation was, however, associated with lower severity of acute pain. The different effects of preoperative and postoperative radiation are likely related to differences in the extent of radiation and other treatment characteristics. In our institution, single dose preoperative radiation is usually applied as part of a clinical trial in the setting of lumpectomy of small low-risk tumors, usually with no need for extensive axillary surgery, whereas postoperative radiation involves prolonged courses sometimes associated with more extensive surgery. Although this is an interesting finding, further evaluation is needed before drawing any definitive conclusions due to the small sample size of those who received preoperative radiation. On the other hand, while some previous studies suggested that axillary dissection was a risk factor for persistent pain,38 we were surprised to find that this was not the case in our cohort. This could be related to the low frequency of axillary dissection in our cohort, as well as the relatively small sample size.

The limitations of our study relate mainly to its single-center nature and its relatively small sample size. Follow-up at 6 and 12 months was missing for some of the patients and we imputed this data for our persistent pain analysis. A sensitivity analysis revealed no difference in results when using the imputed data compared with the completed case data. Given the size of our study, we may have been underpowered to detect some smaller but clinically important associations between the risk alleles of the selected genes and persistent pain, but our results highlight the need for larger-scale genetic association studies to further delineate the potential role of genetic susceptibility in persistent pain after breast surgery. Its strengths include the longitudinal study design which allowed us to follow women across the whole continuum of perioperative care up to 1 year after surgery, and included clinical as well as psychological, genetic, and pain perception risk factors.

In summary, factors most highly associated with increased severity of acute pain included longer duration of surgery, higher PCS, and surgeon, while preoperative radiotherapy was associated with lower acute pain scores. Postsurgery radiation therapy, chemotherapy, and the severity of acute pain were associated with the development of persistent pain. Quantitative sensory tests were not predictive of acute or persistent pain in this cohort. These findings indicate that a greater focus on optimizing acute postoperative pain management for all patients undergoing breast surgery should be an explicit goal with potential long-term benefit. Additional studies should continue to assess whether women expected to have high levels of pain postoperatively could be identified preoperatively so that more aggressive multimodal analgesic interventions could be targeted to this high-risk patient population.

Acknowledgments

We recognize and thank the patients who participated in this study as well as Lina Habib and Andreea Podgoreanu for their help in data management.

References

Footnotes

  • Presented at Interim data from this work were presented at the 2016 American Society of Anesthesiologists Meeting in Chicago, IL, October 22–26, 2016.

  • Funding This study was supported by a grant from the Society for Ambulatory Anesthesia (SAMBA).

  • Patient consent Obtained.

  • Provenance and peer review Not commissioned; externally peer reviewed

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.