Authors: Karen T Bjørnholdt, Birgitte Brandsborg, Kjeld Søballe, Lone Nikolajsen
Categories: Register Studies
Source: Acta Orthopaedica
Persistent postsurgical pain is a well-recognized problem after various types of surgery such as amputation and thoracotomy. The prevalence of persistent pain, and the extent to which it involves neuropathic pain, is highly dependent on the type of surgery. We investigated the prevalence of, characteristics of, and risk factors for persistent pain 1–2 years after shoulder replacement.
A questionnaire was sent to patients who underwent primary shoulder replacement between April 2011 and April 2012, and whose data were recorded in the Danish Shoulder Arthroplasty Register. Patients who had undergone reoperation or bilateral replacements were excluded. Persistent pain was defined as constant or daily pain within the last month, which interfered much or very much with daily activities. Multivariate logistic regression was used to assess risk factors.
538 patients were available for analysis. The prevalence of persistent pain was 22% (CI: 18–25), and the prevalence of presumed neuropathic pain was 13% (CI: 10–16). Persistent pain was more frequent in fracture patients (29%) than in osteoarthritis patients (16%), while the prevalence of neuropathic pain was similar. Severe pain during the first postoperative week increased the risk of persistent pain. Risk also increased with hemiprosthesis (as compared to total prosthesis) in osteoarthritis patients, and with previous osteosynthesis and pain elsewhere in fracture patients.
Persistent pain after shoulder replacement is a daily burden for many patients. Further studies should address patient and prosthesis selection, postoperative pain management, and follow-up of these patients.
There is a substantial amount of literature documenting that there is a possible risk of persistent pain after almost any surgical procedure (Macrae 2001, Johansen et al. 2012). The prevalence rates are highly dependent on the type of surgery, and vary from 5% to 85% (Kehlet et al. 2006, Macrae 2008). The consequences of chronic or persistent postsurgical pain are significant, not only in terms of suffering and reduced quality of life for the individual patient, but also with regard to the subsequent costs to healthcare services and social services. Many authors have reported putative risk factors for persistent pain, including genetic factors, age, psychosocial factors, type of anesthesia, pain elsewhere than the surgical site, other comorbidities, preoperative pain, and acute postoperative pain (Althaus et al. 2012, VanDenKerkhof et al. 2013). Intraoperative nerve damage and the extent of surgery are also important risk factors (Katz and Seltzer 2009). In fact, many patients with persistent postsurgical pain present with characteristic symptoms of neuropathic pain in the affected area (Kehlet et al. 2006).
There is a scarcity of data on persistent postsurgical pain after orthopedic surgery. To our knowledge, previous studies focusing on persistent postsurgical pain in orthopedic patients have concerned mainly amputation or hip or knee replacement (Nikolajsen et al. 2006, Lundblad et al. 2008, Beswick et al. 2012, Liu et al. 2012, Jansen et al. 2014). Trials of shoulder replacement surgery have more commonly reported pain relief, or a composite score including pain, rather than the prevalence of pain at follow-up. There has been very little research on predictive factors for persistent postsurgical pain following shoulder replacement, but the general outcome has been shown to be associated with diagnosis and prosthesis type (Radnay et al. 2007, Fevang et al. 2013) and with previous shoulder surgery, age, and preoperative Short Form-36 mental score and DASH functional score (Simmen et al. 2008). Identification of subgroups at increased risk is important in order to establish interventions to prevent or minimize the impact of persistent postsurgical pain.
We investigated the prevalence of, the characteristics of, and risk factors for persistent pain 1–2 years after more than 500 shoulder replacements performed in Denmark.
This was a cohort study in which the baseline data were retrieved from the Danish Shoulder Arthroplasty Register (DSR) (Rasmussen et al. 2012) and the follow-up data were obtained using a patient questionnaire.
All patients were recruited from the DSR (see description below). For this study, the inclusion criteria were primary shoulder replacement between April 1, 2011 and April 1, 2012, and age above 18 years. Exclusion criteria were prosthesis in the contralateral shoulder and any type of reoperation. Patients who met these criteria were sent a questionnaire on May 30, 2013. A reminder was sent after 25 days to all patients who did not reply. Thus, the length of follow-up was 14–26 months. In order to reduce response bias, patients were strongly urged to respond regardless of whether they had experienced pain or not. The size of the study was determined indirectly by the number of patients in the registry who matched our criteria.
The questionnaire was in Danish and was developed especially for the study, as no suitable pre-existing questionnaire was found. Based on the literature and the experience of the authors, the questionnaire was drafted and assessed by research peers. After revision, the questionnaire was piloted in a group of 10 patients who had undergone shoulder replacement at the first author’s institution. After evaluation of the responses, the final questionnaire was drawn (see the translated questionnaire in Supplementary data). The questionnaire included questions to assess (1) inclusion and exclusion criteria, (2) current pain characteristics and pain treatment, (3) neuropathic pain characteristics (DN4: Douleur Neuropathique, 4 questions) (Bouhassira et al. 2005), and (4) possible predictors of persistent pain (pain elsewhere, height and weight, preoperative and acute postoperative pain). The outcome of persistent pain was defined as pain experienced every day or constantly within the last month at a level that interfered much or very much with daily activities. Recall bias was expected for the questions concerning preoperative pain and acute postoperative pain. To minimize this bias, a verbal rating scale (none/mild/moderate/severe) was used instead of a numerical rating scale, and the period in question was limited to the week before/after surgery.
The DSR was established in 2004. The DSR included 91% and 92% of all shoulder replacements performed in Denmark in 2011 and 2012, respectively (DSR annual report 2013). It collects reports by the surgeons at the time of the operation, and patients are routinely contacted by the registry after 1 year to complete the Western Ontario Osteoarthritis of the Shoulder Index (WOOS) (Lo et al. 2001, Rasmussen et al. 2013) with 2 supplementary questions. Data extracted from the registry included age, sex, diagnosis, prosthesis, previous shoulder surgery, supplementary surgery, and patient-reported data. Patient names, updated addresses, and status (e.g. death, emigration) were retrieved from the Danish Civil Registration System and matched to the registry data by means of the civil registration number.
Data from the questionnaires were entered manually into Epidata version 3.1. They were then merged with registry data for analysis in Stata software version 12. Missing data were not constructed to expected values, and the analysis was based on the data available. Patients were not included in the final analysis if their dataset was incomplete for the grouping variables (questions 8 and 11) and the following predictive factors used in the age, sex, diagnosis, prosthesis type, pain elsewhere, and severity of acute postoperative pain. For patients who had returned their questionnaires with these crucial data missing, and who had accepted to be contacted again, we obtained the missing data by telephone interview or e-mail. Descriptive statistics are presented as counts (with %), as mean (with SD) for normally distributed data, or as median (with interquartile range (IQR)) for data that were not normally distributed. Confidence intervals (CIs) for prevalence are calculated as exact binomial 95% CI (Clopper-Pearson). Analyses for association with persistent pain in Table 2 and for generalizability were performed by t-test, chi-squared test, or Wilcoxon rank-sum test as appropriate. Any p-value of less than 0.05 was considered statistically significant. A multivariate logistic regression model was used to assess whether selected factors predicted the outcome of pain at follow-up. Factors were considered suitable for inclusion in the risk factor analysis if they could correct for unknown confounders (age, sex, body mass index) or were clinically relevant and there were enough data to allow inclusion in analysis.
The study was approved by the Danish Data Protection Agency and the committee of the DSR. Studies based on questionnaires or registers do not require approval from the regional or national Committee on Health Research Ethics in Denmark. The study was registered at http://clinicaltrials.gov/ with identifier NCT01900223.
The registry contained records of 786 patients with 1 primary shoulder replacement who were operated from April 2011 to April 2012. Replies were received from 615 patients (response rate 78%). Data from 538 patients were available for analysis (Figure 1). Mean follow-up time was 20 (14–26) months.
Figure 1. Flow chart of the patients. DSR: the Danish Shoulder Arthroplasty Register.
Persistent pain, defined as pain experienced every day or constantly within the last month at a level that interfered much or very much with daily activities, was present in 117 of 538 patients (22%, CI: 18–25) (Table 1). The prevalence of persistent pain differed between the 2 predominant diagnoses, being present in 66 of 228 patients with fractures (29%, CI: 23–35) and 26 of 226 patients with osteoarthritis (16%, CI: 11–21). Neuropathic pain, assessed by the 7-item DN4 questionnaire, was present in 66 of 505 patients (13%, CI: 10–16). The prevalence of presumed neuropathic pain was similar between 32 of 212 fracture patients (15%, CI: 11–21) and 26 of 212 osteoarthritis patients (12%, CI: 8–17). There was no difference in the length of follow-up between those with persistent pain or presumed neuropathic pain and those without. Analgesics were used daily by 159 of 527 patients (30%) for pain limited to or including the operated shoulder, and a further 67 patients (13%) used analgesics less than once a day for pain limited to or including the operated shoulder. The drugs used were paracetamol (n = 161, 31%), non-steroid anti-inflammatory drugs (n = 59, 11%), opioids (n = 98, 19%), and other kinds (n = 25, 5%). Other analgesics included e.g. gabapentin, pregabalin, tricyclic antidepressants, and anti-epileptics. Non-pharmacological treatment of pain limited to or including the shoulder included physiotherapy (n = 76 of 490, 16%), use of a hot water pool (n = 36 of 490, 7%), acupuncture (n = 20 of 490, 4%) and chiropractic (n = 14 of 490, 3%). For assessment of overall improvement in pain status, patients were asked to compare their current pain to their pain before the operation. However, this was problematic in patients with fractures less than 2 weeks old, who instead often compared their current pain to their pain before the fracture. In the remainder of patients (n = 363), 80% reported being “better” or “much better”, 10% reported no change, and 11% reported being “worse” or “much worse”.
Factors associated with the occurrence of persistent pain were age, BMI, diagnosis, previous osteosynthesis, previous cuff reconstruction (marginally), duration of preoperative pain, prosthesis type, supplementary cuff reconstruction, infection, and frozen shoulder (Table 2). Other patient-reported complications included fever, kidney affection, pneumonia, hematoma, thrombosis in the arm, fistula, swelling, complex regional pain syndrome, 3–4 week paralysis of the arm, skin disorder, tight scar tissue, trapped nerve, irritated biceps, and other prosthesis-related complaints (“the prosthesis irritates me” and “it seems as if it is on its way out through the skin”). For patients with persistent pain compared to those without, all items of the WOOS (completed by 392 of the 538 patients) were highly significantly worse by 2-sample Wilcoxon rank-sum (Mann-Whitney) test (p < 0.001 for all items). Persistent pain was especially associated with the 3 emotional items (frustration/discouragement, worry, and feeling like a burden to others), pain with movement, and increased pain after activity.
As diagnosis interacted with other predictors, regression analysis was stratified for fractures and osteoarthritis (Table 3). The other diagnoses were too rare to allow separate regression models. The 2 patient-reported variables used in the regression were dichotomized from 4 categories (none/mild/moderate/severe) due to the number of patients available, so intensity of pain in the first week was dichotomized as “severe” or “other”, and pain elsewhere was dichotomized as “any pain” or “none”. Severe pain during the first postoperative week (experienced by 199 of 538 patients, 37%) was a risk factor regardless of diagnosis of fracture or osteoarthritis. For fractures, previous osteosynthesis and pain elsewhere were predictive of persistent pain. For osteoarthritis, operation with a hemiprosthesis (humeral head replacement) was associated with a higher risk of persistent pain compared to a total prosthesis. Effects of age, sex, and BMI were small or non-existent.
To assess generalizability, we compared the patients who were analyzed to the 223 patients who were not available for analysis, but presumably meeting other criteria (they were excluded due to death, missing address, non-response, or missing data). There were no statistically significant differences with regard to age, sex, side operated, previous surgery, hospital volume, or supplementary surgery. Patients who were not available for analysis were more often diagnosed with fracture < 2 weeks old (39%; p = 0.01), and less often osteoarthritis (28%; p < 0.001) and cuff arthropathy (9%; p = 0.03). The type of prosthesis was more often hemiprosthesis (62%; p = 0.01) and less often total prosthesis (4%; p = 0.01). WOOS and the 2 supplementary questions from the registry tended to be worse, but only 26% of the unavailable patients had completed WOOS.
The prevalence of persistent pain 1–2 years after primary shoulder replacement was 22%, being higher in fracture patients (29%) than in osteoarthritis patients (16%). Comparisons with other studies are complex due to varying definitions and follow-up periods. For hip replacements due to degenerative hip arthritis, 12% of patients experienced pain with moderate, severe, or very severe impact on daily life after 12–18 months (Nikolajsen et al. 2006). For knee replacements mainly due to osteoarthritis, 20% experienced considerable problems with persistent pain (rated 4–5 out of 5) after 14–23 months (Baker et al. 2007). For shoulder replacements, in a review of 40 studies of Neer-II-type shoulder prostheses including 3,584 patients with osteoarthritis mainly, 85% of patients obtained good pain relief while 9% experienced severe pain after 2–12 years (van de Sande et al. 2006). This is comparable to our result of 80% of non-fracture patients reporting pain to be better or much better than before surgery and 10% reporting pain to be worse or much worse. Given the nature of our study, response bias may lead to overestimation of prevalence. However, our generalizability analysis suggested that those who were unavailable for analysis had comparable or worse WOOS, thus supporting the validity of our estimate.
The prevalence of neuropathic pain (as assessed by DN4) was 13%. In a recent article concerning evaluation of failed shoulder replacement, nerve injury was estimated to occur in 0.6% to 4% of cases (Wiater et al. 2014). Our estimate of neuropathic pain is considerably higher, but it should be considered with caution. DN4 is a screening tool that has not been validated in the Danish language—or in a population of shoulder prosthesis patients.
Chronic or persistent postsurgical pain has been defined as pain that develops after surgery, persists for more than 2 months, and cannot be attributed to causes other than surgery (Macrae and Davies 1999). For certain types of surgery the healing period is longer, and the definition should be adjusted accordingly (Kehlet and Rathmell 2010, Wylde et al. 2013). In the present study, the definition was different, as identification of the underlying causes of persistent pain would require a thorough physical, radiological, neurophysiological, and biochemical examination of each patient. However, this would be necessary to more closely estimate the prevalence of persistent postsurgical pain as defined by Macrae and Davies (1999) and the prevalence of neuropathic pain as defined by the International Association for the Study of Pain (pain caused by a lesion or disease of the somatosensory nervous system). After our study, and in light of the high prevalence of pain found, it would be desirable in further studies to apply the definition of Macrae.
In the present study, complications that may reflect a neuropathic pain state were mentioned and could be investigated further, e.g. complex regional pain syndrome, paralysis of the arm, tight scar tissue, and trapped nerve. Some of the patients experiencing persistent pain would most likely benefit from such an evaluation, including assessment of the possibility of revision or alternative analgesic treatment. A surprisingly high number of the patients used opioids as analgesics (19%), and this also calls for a further assessment of the pain patients. In a review article focusing on complications after 4,010 shoulder replacements, 23% of the patients experienced 1 or more complications within a mean follow-up time of 6 (2–25) years. Many of the complications reported in the review were possibly avoidable or would lead to revision (Gonzalez et al. 2011).
Severe pain in the first postoperative week was associated with a markedly increased risk of persistent pain, although this result may have been influenced by recall bias. The association between acute and chronic postsurgical pain has also been found in other studies. The causal relationship has not been fully established, but the rather high prevalence of 37% of patients experiencing severe pain in the first week identifies a need to treat acute postsurgical pain more aggressively, regardless of the possibility of increased risk of persistent pain. Future intervention studies may determine whether better pain control is attainable, or whether preoperative assessment concerning the risk of severe acute and/or chronic pain could improve patient selection. In osteoarthritis patients, there was a higher risk with hemiprostheses than with total prostheses. This result is in accordance with other studies indicating a superior outcome with a total prosthesis compared to a hemiprosthesis (Bryant et al. 2005, Radnay et al. 2007, Singh et al. 2010, Fevang et al. 2013). As in all registry studies, a limitation exists in the completeness and reliability of registry data such as prosthesis type and supplementary surgery, and this problem is not easy to quantify. Also, the questionnaire developed did not undergo testing of reliability and validity beyond the method described.
In conclusion, persistent pain after shoulder replacement is a daily burden to many patients. Further prospective studies are required to confirm our results and to evaluate the causes of persistent pain and the treatments or preventive measures required. Studies should not only concentrate on prosthesis selection and surgical complications, but also involve (1) preoperative assessment to effectively improve patient selection, and (2) improvement of postoperative pain management and its effect on the development of persistent pain.
The Questionannaire is available at www.actaorthop.org, identification number 7739.
LN came up with the idea. KTB collected the data and, assisted by BB, performed the analysis. All the authors contributed to study design, interpretation, and critical revision of the manuscript. The final manuscript was approved by all the authors.
We thank Gerhardt Teichert and Janne Ovesen for contributing to the interpretation of data and for commenting on the manuscript, and secretary Line Jensen for proofreading the manuscript. This work was supported by the Health Research Fund of Central Denmark Region, the Hede Nielsen Family Foundation, and the Augustinus Foundation.
No competing interests declared.