Pharmacotherapy for the Prevention of Chronic Pain after Surgery in Adults: An Updated Systematic Review and Meta-analysis

This systematic review was conducted according to the original study protocol,⁵  and in a consistent manner with the original review.⁴  Procedures were guided by Cochrane Collaboration recommendations⁶  and followed the principles of Preferred Reporting Items for Systematic Reviews and Meta-analysis⁷  and A Measurement Tool to Assess Systematic Reviews.⁸ 

Using the originally published search strategy (Supplemental Digital Content 1, appendix A, https://links.lww.com/ALN/C628),⁴  the following databases were searched for trials since the previous review (July 17, 2013, to July 1, 2019): Cochrane Central Register of Controlled Trials, MEDLINE, and EMBASE. We conducted hand searches of trial registries using each intervention as the key word (e.g., ketamine and pregabalin, among others) and filtered results by interventional studies, age group (18 to 65+ yr), and outcomes (e.g., chronic pain OR persistent pain OR persistent postsurgical pain). No limits were placed regarding date, language, or status of the publications. Backward reference searching was conducted by screening reference lists of included studies and relevant systematic reviews. Authors of included studies and experts were asked about recent or forthcoming studies that fit our eligibility criteria.

We included double-blind, placebo-controlled, randomized controlled trials that involved participants 18 yr and older undergoing a planned surgical procedure, that evaluated one or more drugs administered systemically immediately before, during, or after the procedure by any dose, route, or frequency, and that included data on a patient-reported measure of pain 3 or more months postsurgery. This review only included randomized controlled trials because “randomization is the only way to prevent systematic differences between baseline characteristics of participants in different intervention groups in terms of both known and unknown (or unmeasured) confounders.”⁶ 

The following was extracted for each study: drug name; trial methods; trial registration; participant demographics; preoperative pain status and analgesic use; type of surgery; dosing including route, timing, and duration; dropouts due to treatment-emergent adverse effects; concomitant standardized analgesic approach; planned dichotomous outcomes; proportion of patients reporting any pain (more than 0 out of 10) or moderate to severe pain (greater than or equal to 4 out of 10) at 3, 6, and 12 months postsurgery. We reviewed trial registries when available, and in the case of secondary publications, original papers were reviewed. If a study reported parametric measures of pain intensity but not dichotomous measures of proportions of participants reporting pain, we contacted corresponding authors for supplementary data. Extraction was performed by M.E.C. and I.G. by reading each included study and completing the data extraction form.

Eligible studies were evaluated independently by two reviewers (M.E.C., I.G.) for risk of bias using the Cochrane risk of bias tool.⁹  Any discrepancies could be resolved by a third coauthor (E.V.); however, this did not occur. Attrition bias was assessed as “low-risk” for studies where the dropout rate was less than 20%.¹⁰  Studies with higher dropout rates that included intention-to-treat analyses were assessed as “unclear” or “high risk of bias.” Chronic pain was rarely the prespecified primary outcome and most included trials were underpowered for this outcome; therefore, “other potential sources of bias” were assessed as high-risk in studies that had fewer than 50 participants per arm.¹¹  While it could be argued that, for pain prevention trials, this number should even be higher than 50 participants per arm, there is currently no consensus for a specific higher threshold for trial size in this setting.¹² 

The primary outcome for the review was the proportion of participants reporting any pain at the anatomical site of the procedure or pain referred to the surgical site, or both, 3 months or more after the surgery.²  Secondary outcomes were the number of participants reporting moderate to severe pain at the anatomical site of the procedure or pain referred to the surgical site—or both—6 months or more after surgery, as well as the number of participants who dropped out of the study due to treatment-related adverse effects. All results reported represent aggregate data from the 2013 and current review, unless otherwise specified.

Comparing the study drug(s) with placebo was the primary objective. Studies were grouped if they evaluated the same drug(s) administered in a similar manner (i.e., dosage, route of administration, and treatment duration). Given the potential effect on outcome of surgical procedure and underlying condition, timing of outcome measurement, and duration of the intervention, subgroup analyses were conducted according to these parameters. Given the diverse features of the studies included in the review, not all were necessarily represented in a meta-analysis.

Statistical analyses were conducted using Review Manager v5.3.¹³  Dichotomous data were analyzed using Mantel–Haenszel fixed-effects model for risk ratio with 95% CI. Heterogeneity was evaluated by visual examination of forest plots and use of the I² statistic. In cases of moderate to considerable heterogeneity (i.e., 30 to 100%) the random-effects model was employed.⁶  For studies with multiple intervention arms, we split the “shared” (placebo) group into two or more groups with smaller sample size, and included two or more (reasonably independent) comparisons.⁶  Sensitivity analyses were conducted to evaluate robustness of a result by omitting studies considered to be outliers with respect to study quality, drug dose and duration, or pain measurement scales.

The search identified 6,709 citations, with first level screening based on title and abstract yielding 115 studies for full text review, of which 70 new studies fulfilled the inclusion criteria (fig. 1). The majority of the 45 excluded studies did not follow participants for at least 3 months (n = 15), were not placebo controlled (n = 9), were not double-blinded (n = 7), were not relevant to the prevention of chronic postsurgical pain (n = 6), or did not evaluate drugs administered systemically (n = 4). Full details regarding the excluded studies are summarized in Supplemental Digital Content 2 (appendix B, https://links.lww.com/ALN/C629). Our trial database searches yielded 46 ongoing and unpublished studies. Ongoing studies are evaluating ketamine (n = 12), pregabalin (n = 11), IV lidocaine (n = 8), dexamethasone (n = 4), gabapentin (n = 3), dexmedetomidine (n = 2), magnesium (n = 2), acetyl-salicylic acid (n = 1), cannabinoids (n = 1), clonidine (n = 1), duloxetine (n = 1), lamotrigine (n = 1), meloxicam (n = 1), midazolam (n = 1), propranolol (n = 1), sevoflurane (n = 1), and tramadol-paracetamol (n = 1). A summary of the 46 ongoing studies is included in Supplemental Digital Content 3 (appendix C, https://links.lww.com/ALN/C630).

Characteristics of the 110 included studies (70 new plus 40 from the previous review)⁴  are summarized in Table 1 and Supplemental Digital Content 4 (appendix D, https://links.lww.com/ALN/C631). Studies (new and from previous review) involved various surgeries including breast (n = 19), total hip or knee arthroplasty (n = 16), thoracotomy (n = 14), spine (n = 14), abdominal or pelvic (n = 12), heart (n = 8), limb amputation (n = 5), thyroidectomy (n = 5), inguinal herniorrhaphy (n = 4), caesarean section (n = 3), carpal tunnel (n = 2), brain (n = 1), mandibular fracture (n = 1), and a combination of surgeries (n = 6) (table 1).

Of all the new and previous studies, only 37 studies included patients that were free of pain before surgery. Patients taking various analgesics were excluded from 36 trials. Preoperative pain or analgesic use was unclear in 11 studies. Patients with preexisting pain were included in 26 studies (table 1).

Studies received financial support from research granting agencies (n = 28), institutional and/or departmental sources (n = 18), pharmaceutical companies (n = 10), and granting agencies and pharmaceutical companies (n = 1); 10 studies stated that no funding was received; and the source of funding was not reported for 43 studies (Supplemental Digital Content 4, appendix D, https://links.lww.com/ALN/C631). Insufficient reporting prohibits further investigation of possible correlations between sources of financial support and study outcomes and it is beyond the scope and preplanned objectives of the current review. Seventy-nine of 110 (71.8%) included studies had at least four of seven items that qualified as low risk of bias (Supplemental Digital Content 5, appendix E, https://links.lww.com/ALN/C632). Most studies were of small sample size having fewer than 50 participants per arm (n = 70 [64%]), greater than or equal to 50 and fewer than 100 per arm (n = 29 [26%]), and greater than or equal to 100 per arm (n = 11 [10%]).

Thirteen new studies (n = 1,283 participants)^14–27  evaluated ketamine or (S)-ketamine (total, 27 studies; n = 2,757).^14–41  Nine of 27 studies reported prevalence of any pain at 3 months,^{16,20,22,26,29,37,39–41}  16 studies at 6 months,^{14–17,22,24–26,30,33–37,40,41}  and five studies at 12 months.^{14–16,28,30}  Prevalence of any pain at 3 months ranged from 5.6 to 72.2% (mean, 35.0%) in the placebo arm and 5.6 to 83.3% (mean, 31.5%) in the ketamine arm. No treatment effect of ketamine was observed on prevalence of any pain regardless of outcome timing, duration of drug administration, or surgical procedure (fig. 2). Forest plots for studies evaluating ketamine are included in Supplemental Digital Content 6 (appendix F, https://links.lww.com/ALN/C633). In 2013, subgroup analysis based on duration of treatment suggested a significant effect of ketamine compared to placebo (odds ratio, 0.37 [95% CI, 0.14 to 0.98]; two studies; 135 participants) on the prevalence of any pain at 3 months for studies evaluating ketamine treatment for more than 24 h; however, the current review did not demonstrate a similar treatment effect (risk ratio, 0.83 [95% CI, 0.58 to 1.18]; five studies; 331 participants).

Two studies reported prevalence of moderate to severe pain at 3 months (placebo: range, 14.7 to 16.7%; mean, 15.7; ketamine: range, 9.1 to 32.3%; mean 20.7),^16,22  six studies at 6 months (placebo: range, 0.0 to 39.1%; mean, 17.9; ketamine: range, 3.2 to 26.7%; mean, 12.2),^{14,16,22,33,35,37}  and two studies at 12 months (placebo: range, 7.1 to 26.1%; mean, 16.6; ketamine: range, 0.0 to 12.5%; mean, 6.3).^14,16  No treatment effect of ketamine was observed on prevalence of moderate to severe pain regardless of outcome timing, duration of drug administration, or surgical procedure (fig. 2). Only two of the 27 ketamine studies provided data regarding dropouts due to treatment-related adverse effects. Of those, 4 of 70 (5.7%) received ketamine and 4 of 70 (5.7%) received placebo. Adverse events included hallucinations, delayed emergence, dizziness, diplopia, and confusion.^19,25 

Ketamine has been evaluated in three recent reviews for orthopedic surgery,^42,43  and thoracotomy.⁴⁴  Consistent with the current review, the majority (two of three) indicated results to be inconclusive.^43,44  In disagreement, one narrative systematic review evaluating various interventions for adults receiving primary total knee arthroplasty concluded a treatment effect of ketamine claiming “good-quality evidence for a small benefit”⁴² ; however, their conclusion was based on one small randomized controlled trial.¹⁴ 

Twenty-one new studies (n = 3,184)^21,45–61  evaluated pregabalin (total, 26 studies; n = 3,693).^21,45–67  Nineteen of 26 studies reported prevalence of any pain at 3 months,^{21,45,47–51,53,57,58,61–63,65–67}  six studies at 6 months,^{45,48,54,58,63}  and two studies at 12 months.^54,65  Prevalence of any pain at 3 months ranged from 3.1 to 80.0% (mean, 39.5%) in the placebo arm and 3.7 to 88.0% (mean, 31.9%) in the pregabalin arm. Subgroup analyses resulted in a statistically significant treatment effect of pregabalin 3 months after cardiac surgery (three trials; risk ratio, 0.25 [95% CI, 0.13 to 0.50]), and 3 months after total knee arthroplasty (three trials; risk ratio, 0.75 [95% CI, 0.58 to 0.97]). No treatment effects were observed for any pain evaluated at 3, 6, or 12 months when drug administration was for 24 h or less or more than 24 h or for other types of surgical procedures (fig. 3). Forest plots for studies evaluating pregabalin are included in Supplemental Digital Content 7 (appendix G, https://links.lww.com/ALN/C634). In 2013, only one study evaluated the prevalence of any pain at 6 months therefore no subgroup analyses were performed; in the current review, six studies were included in meta-analysis and did not demonstrate a treatment effect of pregabalin when drugs were administered for more than 24 h (risk ratio, 0.78 [95% CI, 0.47 to 1.28]).

Nine studies reported prevalence of moderate to severe pain at 3 months (placebo: range, 4.2 to 34.0%; mean, 20.2; pregabalin: range, 0.0 to 20.0%; mean, 8.7),^{45,47,48,51,53,57,59,61,63}  and three studies at 6 months (placebo: range, 11.3 to 28.0%; mean, 17.9; pregabalin: range, 2.7 to 8.8%; mean, 5.8).^45,48,63  When pregabalin was administered for more than 24 h the overall effectiveness risk ratio showed a statistically significant treatment effect of pregabalin compared to placebo at 3 months (nine trials; risk ratio, 0.47 [95% CI, 0.33 to 0.68]), and 6 months (three trials; risk ratio, 0.29 [95% CI, 0.14 to 0.58]) for varying surgical procedures, and 3 months after total knee arthroplasty (two trials; risk ratio, 0.42 [95% CI, 0.22 to 0.81]) (fig. 3). Only eleven of the 26 pregabalin studies provided data regarding dropouts due to treatment-related adverse effects. Of those, 56 of 1,295 (4.3%) received pregabalin and 27 of 819 (3.3%) received placebo. Adverse events included dizziness, nausea, vomiting, sedation, diplopia, somnolence, visual disturbances, fainting, fatigue, constipation, and allergic reaction.^{45,47,49,56–58,62–64} 

Pregabalin has been evaluated in four recent reviews for orthopedic surgery,⁴²  thoracotomy,⁶⁸  breast cancer surgery,⁶⁹  and various surgeries.⁷⁰  Consistent with the current review, half (two of four) of these reviews did not have sufficient evidence to make a clear recommendation.^69,70  Two reviews concluded a treatment effect of pregabalin. One narrative systematic review evaluating various interventions for total knee arthroplasty⁴²  was limited to one randomized controlled trial from 2010⁶³  and the other review included nine studies for thoracotomy, seven of which were excluded from the present review due to lack of blinding, not placebo controlled, and lack of long term pain assessment.⁶⁸  Furthermore, two of the nine studies that were included in our review did not find a reduction in the prevalence of postsurgical chronic pain.^47,53  Despite the high proportion of studies lacking data on adverse events, consistent with our review adverse events included sedation,^42,70  dizziness,^68,70  drowsiness,^68,69  and visual disturbances.⁷⁰ 

Eight new studies (n = 1,367)^52,71–77  evaluated gabapentin (total, 18 studies; n = 2,166).^{38,52,71–86}  Six of 18 studies reported prevalence of any pain at 3 months,^{72,81–84,86}  four studies at 6 months,^72,73,80,84  and one study at 12 months.⁷³  Prevalence of any pain at 3 months ranged from 20.0 to 66.7% (mean, 49.9%) in the placebo arm and 12.5 to 70.2% (mean, 47.8%) in the gabapentin arm. No treatment effects were observed for any pain evaluated at 3 or 6 months (fig. 3). Forest plots for studies evaluating gabapentin are included in Supplemental Digital Content 8 (appendix H, https://links.lww.com/ALN/C635). Consistent with the 2013 review, meta-analyses of studies evaluating gabapentin failed to demonstrate statistical significance upon comparison to placebo at three or six months.

Two studies reported prevalence of moderate to severe pain at 3 and 6 months,^72,76  however results were not pooled given heterogeneity of timing and duration of administration. When drug administration was for 24 h or less, the prevalence of moderate to severe pain at 3 months was 21.1% in the placebo group and 22.2% in the gabapentin group and 10.5% and 16.7% at 6 months, respectively.⁷⁶  When drug administration was for more than 24 h, the prevalence of moderate to severe pain at 3 months was 13.5% in the placebo group and 12.8% in the gabapentin group and 8.1% and 16.7% at 6 months, respectively.⁷²  Only five of the 18 gabapentin studies provided data regarding dropouts due to treatment-related adverse effects. Of those, 32 of 506 (6.3%) received gabapentin and 18 of 401 (4.5%) received placebo. Adverse events included severe sedation, dizziness, nausea, syncope, paresthesia of the legs, and elevated serum creatinine.^{72–74,83,84} 

Gabapentin has been evaluated in two recent reviews for breast cancer surgery.^69,87  One review concluded low- to very-low–quality evidence that preoperative use of gabapentin does not reduce the rate of chronic postsurgical pain.⁶⁹  One review concluded that “preoperative use of gabapentin was able to reduce acute and chronic postoperative pain.”⁸⁷  However, seven of nine studies were excluded from the current review; six due to follow-up for less than 3 months (range, 12 h to 1 month), and one was a clinical trial with one arm that combined topical analgesia and gabapentin. It is unclear why two of five studies were included in their meta-analysis evaluating chronic pain given their short timeline for follow-up (i.e., 24 h and 7 days).^88,89  Furthermore, it is unclear why two studies included in the meta-analysis by Jiang et al.⁸⁷  show a treatment effect of gabapentin: Amr et al.⁷⁸  did not report dichotomous results for the incidence of chronic pain and concluded “gabapentin had no effect on chronic pain,” and Fassoulaki et al.⁸¹  reported no difference in the proportion of chronic pain between gabapentin 12 of 22 (54.5%) and pregabalin 14 of 24 (58.3%).

Nine new studies (n = 808)^{18,51,90–96}  evaluated IV lidocaine (total, 10 studies; n = 844).^{18,51,90–97}  Six of 10 studies reported prevalence of any pain at 3 months,^{51,90,91,93,94,97}  three studies at 6 months,^90,93,96  and no studies at 12 months. Prevalence of any pain at 3 months ranged from 17.4 to 79.2% (mean, 41.6%) in the placebo arm and 11.8 to 92.3% (mean, 32.7%) in the IV lidocaine arm. One study could not be pooled in meta-analysis due to duration of drug administration for more than 24 h during colectomy.⁹⁰ Subgroup analyses of prevalence of any pain at 6 months based on duration of treatment being 24 h or less showed a statistically significant treatment effect of IV lidocaine after breast surgery (two trials; risk ratio, 0.43 [95% CI, 0.23 to 0.80]). No treatment effect of IV lidocaine was observed at 3 months after breast surgery or when the drug was administered for 24 h or less (fig. 4). Forest plots for studies evaluating IV lidocaine are included in Supplemental Digital Content 9 (appendix I, https://links.lww.com/ALN/C636).

Two studies reported prevalence of moderate to severe pain at 3 months (placebo: range, 10.0 to 20.8%; mean, 15.4; IV lidocaine: range, 4.7 to 7.7%; mean, 6.2),^51,93  and two studies at 6 months (placebo: range, 3.4 to 22.2%; mean, 12.8; IV lidocaine: range, 3.2 to 8.8%; mean, 6.0).^93,96  No treatment effect of IV lidocaine was observed for this outcome regardless of timing of outcome measurement or surgical procedure (fig. 4). Only 1 of the 10 IV lidocaine studies provided data regarding dropouts due to treatment-related adverse effects. Of those, 1 of 22 (4.5%) received IV lidocaine and 0 of 22 (0.0%) received placebo. One patient in the IV lidocaine group developed convulsions during injection of the loading dose.⁹² 

Intravenous lidocaine has been evaluated in two recent reviews for breast cancer surgery,⁹⁸  and various surgeries.⁹⁹  Both reviews were cautiously optimistic in support of IV lidocaine for preventing chronic postsurgical pain. However, higher quality evidence from large, definitive, multicenter clinical trials was called for before a widespread change in practice could be justified.⁹⁹ 

Five new studies (n = 451) evaluated nonsteroidal anti-inflammatory drugs (NSAID) including one celecoxib,¹⁰⁰  one dexketoprofen,¹⁰¹  one flurbiprofen axetil,¹⁰²  one parecoxib,¹⁰³  and one IV parecoxib in combination with oral celecoxib¹⁰⁴  (total, eight studies; n = 1,602).^100–107  Two of eight studies reported prevalence of any pain at 3 months,^103,104  three studies at 6 months,^102,104,106  and four studies at 12 months.^{102–104,107}  Prevalence of any pain at 3 months ranged from 48.8 to 59.1% (mean, 53.9%) in the placebo arm and 22.5 to 54.3% (mean, 38.4%) in the NSAID arm. Subgroup analysis did not show an effect of NSAIDs compared to placebo for studies evaluating treatment for more than 24 h at 3, 6, and 12 months; however, a statistically significant treatment effect was observed at 12 months when drugs were administered for 24 h or less (fig. 5). Forest plots for studies evaluating NSAIDS are included in Supplemental Digital Content 10 (appendix J, https://links.lww.com/ALN/C637).

One study reported prevalence of moderate to severe pain at 3 and 6 months and concluded no treatment effect of COX-2 inhibitors on persistent pain.¹⁰⁴  Two studies reported prevalence of moderate to severe pain at 12 months; however, results were not pooled due to heterogeneity of timing and duration of NSAID administration. When drug administration was for 24 h or less,¹⁰⁷  the prevalence of moderate to severe pain at 12 months was 3.2% in the placebo group and 0.0% in the NSAID group versus 2.4% versus 0.0%, respectively, when drug administration was for more than 24 h.¹⁰⁴  Only one of the eight NSAID studies provided data regarding dropouts due to treatment-related adverse effects. Of those, 51 of 440 (11.6%) received ibuprofen and 37 of 435 (8.5%) received placebo.¹⁰⁵ 

Three new studies (n = 1,315) evaluated corticosteroids: two dexamethasone^108–110  and one methylprednisolone¹¹¹  (total, six studies; n = 1,620).^107–113  One of six studies reported prevalence of any pain at 3 months,¹¹⁰  one at 6 months,¹¹¹  and one at 12 months.¹⁰⁷  Results were not pooled due to heterogeneity of the timing of outcome measurement.

Two of six studies reported the prevalence of moderate to severe pain at 12 months (placebo: range, 3.2 to 50.0%; mean, 26.6; corticosteroid: range, 5.4 to 72.7%; mean, 39.0).^107,109  Subgroup analysis at 12 months based on duration of treatment for 24 h or less resulted in a statistically significant treatment effect of placebo (two trials; risk ratio, 1.47 [95% CI, 1.05 to 2.06]) (fig. 5). Forest plots for studies evaluating corticosteroids are included in Supplemental Digital Content 11 (appendix K, https://links.lww.com/ALN/C638). No studies evaluating corticosteroids provided data regarding dropouts due to treatment-related adverse effects.

Fewer studies evaluated acetaminophen (two new; n = 290),^114,115  amantadine (two studies, one new; n = 82),^116,117  dexmedetomidine (one new; n = 80),¹¹⁸  dextromethorphan (one study, not new; n = 50),¹¹⁹  duloxetine (two new; n = 207),^120,121  etanercept (one new; n = 77),¹²²  fentanyl (one study, not new; n = 65),¹²³  magnesium (one new; n = 126),⁹⁴  memantine (one study, not new; n = 19),¹²⁴  mexiletine (two studies, not new; n = 175),^81,125  minocycline (two new; n = 231),^126,127  nefopam (four new; n = 307),^14,128–130  nitrous oxide (two studies, one new; n = 5,375),^131,132  valproic acid (one new; n = 128),¹³³  venlafaxine (one study, not new; n = 150),⁷⁸  and vitamin C (one new; n = 123).¹³⁴  Primary and secondary outcomes for drugs evaluated in fewer than five studies were inconclusive and shown in Supplemental Digital Content 12 (appendix L, https://links.lww.com/ALN/C639).

This update reports on an escalating number of randomized controlled trials evaluating perioperative systemic drugs for the prevention of chronic postsurgical pain. The previous review in 2013 included 40 studies and the current one adds 70 new studies in just the last 6 yr. Most studies evaluated drugs that are used to treat acute postoperative pain—namely, ketamine, pregabalin, gabapentin, IV lidocaine, and NSAIDs. Overall, meta-analyses of available studies demonstrated superiority over placebo in 0 of 15 ketamine meta-analyses, 5 of 17 pregabalin meta-analyses, 0 of 4 gabapentin meta-analyses, 2 of 8 IV lidocaine meta-analyses, and 1 of 7 NSAID meta-analyses. Treatment-related adverse effects resulting in study dropouts were reported in only 2 of 27 ketamine studies, 11 of 26 pregabalin studies, 5 of 18 gabapentin studies, 1 of 10 IV lidocaine studies, 1 of 8 NSAID studies, and 0 of 6 corticosteroid studies. Insufficient reporting on the potential harms of each of the pharmacologic interventions was an impediment to conducting quantitative assessments to weigh the benefit–risk trade-offs.

The 110 included studies were of reasonably good quality with mostly low risks of bias related to randomization and blinding. Frequent risks of bias were related to small sample size (fewer than 50 participants).^11,135  Studies which were insufficiently blinded or uncontrolled were excluded as shown in the “Characteristics of Excluded Studies” table (Supplemental Digital Content 2, appendix B, https://links.lww.com/ALN/C629).

The studies included in this review varied with respect to pharmacologic interventions (i.e., 28 different drugs and 16 drug classes); dosage, timing, and duration of drug administration; surgical procedures; participants (e.g., with and without preoperative pain); sample size; outcome measurement tools; and timing of pain assessment (e.g., 3, 6, and/or 12 months). These disparities restrict the amount of data that can be pooled in meta-analysis which presents major challenges in interpretation and applicability of the results. Therefore, caution is advised when generalizing the results beyond the boundaries of the subanalyses conducted in this review. This review should be considered in the setting of several potential limitations. Although 110 randomized controlled trials were included, only 59 studies allowed for direct comparisons in quantitative synthesis. Others were excluded due to variation in drugs evaluated, surgical procedures, pain assessment tools, and timing of pain outcome measurement. Although restriction of this review to double-blind, randomized controlled trials limits the potential for some sources of bias, the relatively small size of most of the studies (i.e., 90% with fewer than 100 participants per arm), and high levels of withdrawals in some studies contribute other sources of bias that potentially overestimate treatment effect. Also, chronic pain was not necessarily the primary outcome for all included studies. Measures of pain at 3 or more months after surgery may have been secondary outcomes which may be a source of selective reporting bias. Furthermore, detailed assessment of pain and its consequences were often not reported beyond “Yes/No” since only a limited number of studies reported relevant moderate/severe pain. However, we believe all available results be considered for inclusion. The heterogeneity with respect to surgical procedures (i.e., nerve vs. other tissue damage), participant populations (preexisting chronic pain, opioid use, and psychiatric morbidities), diverse underlying sources of pain after surgery (e.g., incisional, nerve transection/injury, lymphedema, and deep tissue, among others, occurring after breast cancer surgery), and treatment dose/duration limit interpretation. This includes the question of whether the surgery was done to treat a pain condition, or otherwise, has not been addressed sufficiently in the literature. Other limitations come from heterogeneity regarding the study intervention (e.g., drug dose [small/large], timing with respect to surgery [pre-, intra-, postoperative], and insufficient numbers of trials in each of these categories to conduct relevant subgroup analyses). Although this review did not reveal strong or consistent treatment effects for preventing chronic postsurgical pain, the observation of some statistically significant results points to the concern of multiplicity in systematic reviews where several different meta-analyses are conducted.¹³⁶  Although the Cochrane Collaboration⁶  and other investigators do not generally recommend adjusting for multiple comparisons and is not generally done in meta-analyses—which seek to estimate intervention effects rather than test for them—this is still an area for future investigation.¹³⁶  Finally, lack of access to data from studies that remain unpublished may be an important source of publication bias to consider.

However, strengths of this review should be acknowledged: (1) this is the most up-to-date review of pharmacotherapy for prevention of chronic postsurgical pain with trials published as recently as 2019; (2) we conducted a comprehensive search for eligible randomized controlled trials in any language; (3) procedures throughout the review were conducted in a way that was rigorous, transparent, and replicable; (4) this review follows definitive standard reporting criteria according to the Cochrane Collaboration,⁶  Preferred Reporting Items for Systematic Reviews and Meta-analysis,⁷  and A Measurement Tool to Assess Systematic Reviews⁸ ; (5) this is the only known systematic review in the past 5 yr that has considered all perioperative systemic drugs and was not limited by surgical procedure; (6) we reviewed a number of therapeutic agents in the same systematic manner; and (7) we used subgroup analyses according to dose/duration of treatment, surgical procedure, and timing of outcome measurements.

There is a need for better designed, large-scale, high-quality studies with adequate power to detect treatment effects of pharmacologic interventions on chronic pain outcomes 3 or more months after surgery, and focus on patient safety by reporting consistent and reliable data on withdrawals due to treatment-related adverse events. Conducting further trials of gabapentinoids for chronic pain prevention should take into consideration their apparent lack of effect for acute postoperative pain,¹³⁷  and the diminishing likelihood of effectiveness for preventing chronic postoperative pain. Researchers should consider using detailed standardized outcome measurement tools (e.g., pain intensity on a 0 to 10 numerical rating scale) that can be summarized using dichotomous outcomes (e.g., any pain [more than 0 out of 10] and moderate to severe pain [greater than or equal to 4 of 10]) assessed at multiple and consistent time points (e.g., 3, 6, and 12 months) postsurgery, along with the specific relation of pain to the operated area, and consider stratification of those with and without preoperative pain and analgesic use, as well as implementing better characterization of surgical procedure (nerve damage) and patient characteristics (high pain responders) where appropriate. Studies should focus on drug dosage and duration within the context of the procedure-specific acute pain trajectory in question. There may be little value to repeat studies on single-shot or short-term drug interventions for this multifactorial problem, with a continuous inflammatory response lasting for several days (or weeks). Finally, considering use of the drugs included in this review to prevent chronic postsurgical pain—in light of their apparently uncertain effectiveness—also requires consideration of their safety in the perioperative setting. Given the potential adverse effects of some of these drugs (e.g., COX-2 inhibitors,¹³⁸  gabapentinoids¹³⁹ ), it should be noted that safety assessment and reporting in perioperative clinical trials is sometimes inadequate.^140,141  Therefore, any future research in this area should incorporate more thorough and comprehensive safety assessment and reporting.

Consistent with our original review, and supported by nearly triple the number of studies, this review suggests again the need for larger-scale, high-quality studies to confirm or refute the effectiveness and safety of pharmacologic interventions for the prevention of chronic postsurgical pain. Based on currently available evidence, none of the drugs studied so far can be recommended for clinical use specifically for the indication of preventing chronic pain after surgery.

The authors wish to thank Joanne Abbott, M.Sc., Cochrane Collaboration, Oxford, United Kingdom, and Amanda Ross-White, B.A., M.L.I.S., Queen’s University Library, Kingston, Ontario, Canada, for their valuable assistance with searching the literature.

This review was supported, in part, by the Canadian Institutes of Health Research Strategy for Patient-oriented Research Chronic Pain Network (Hamilton, Ontario, Canada).

The authors declare no competing interests.