Neck pain is common, is experienced by approximately one third of adults over the course of one year [1], can be Expert Panel: Helge Kasch, MD - Neurology, Aarhus University Hospital, Aarhus, Denmark; Jan Hartvigsen, DC, PhD, University of Southern Denmark; Odense M, Denmark; Marc White, MD, PhD - Family Medicine, Massage Therapy, Canadian Institute for the Relief of Pain and Disability, Vancouver, Canada; Margareta Nordin, DMedSci, New York University Hospital for Joint Diseases, New York, USA; Justin Kenardy, Clinical Health Psychologist, PhD, The University of Queensland, Royal Brisbane and Women’s Hospital, QLD, Australia; Norm Buckley, MD - Anesthesia, McMaster University, Hamilton, Canada

severely disabling and is contributing to rising socio-economic costs and societal burden [2]. Patients most commonly seek care from medical doctors, and physicians typically prescribe pharmacological interventions [3]. A variety of medications and medicinal injections are used to reduce transient, recurrent or persisting neck pain and disability in the acute or chronic stages of the disorder. Physicians may chose from various classes of medications (see APPENDIX 1) including: non-opioid analgesics, oral and topical NSAIDs, opioids, muscle relaxants, benzodiazepines, tricyclic antidepressants and GABA derivatives. Medicinal injections might also be used including: corticosteroids, anesthetics, and neuromuscular paralytic agent (botulinum toxins).

The choice of a specific agent often considers the mechanism of action of the specific drug [4-6], it presumed efficacy and adverse events, the individual patient, including past therapies tried, and should also be informed by evidence that the chosen intervention will lead to the therapeutic objective in that patient population. For non-opioid analgesics like acetaminophen (e.g. Tylenol^®) the mechanism of action remains unclear but may include inhibition of cyclooxygenase (COX) enzymes; it typically well tolerated with limited adverse effects. NSAIDs act by blocking cyclo-oxygenase (COX) enzymes 1 & 2. It’s thought that blocking COX-2 decreases pain and inflammation while also reducing the risk of gastrointestinal adverse effects (ulcers, bleeding) that result from COX-1 inhibition on the GI mucosa and platelets, whereas controversy remains over the role of NSAIDs and cardiovascular adverse effects [4]. Common oral NSAIDS include ibuprofen (e.g. advil^®, motrin^®), naproxen (e.g. aleve^®, naprosyn^®), diclofenac (e.g. voltaren^®) with celecoxib (celebrex^®) an example of a COX-1 sparing NSAID. Opioid medications’ analgesia is obtained principally through mu-opioid receptors, with opioids most often used for chronic pain refractory to other therapies. Opioids can produce respiratory depression, nausea, vomiting, dizziness and addictive behavior in susceptible individuals. Societal concerns of diversion also limit their use. Tricyclic antidepressants increase serotonin and norepinephrine and commonly produce drowsiness, dry mouth, blurred vision, constipation, urinary retention and weight gain. Selective serotonin and norepinephrine reuptake inhibitors (e.g. duloxetine, venlafaxine) antidepressants increase serotonin and norepinephrine and have common side effects of nausea and dizziness. Anticonvulsants such as gabapentin or pregabalin decrease excitatory neurotransmitters such as glutamate and their side effects commonly include drowsiness, dizziness, unsteadiness and unclear thinking. Topical anesthetics such as lidocaine block sodium channels and are well tolerated as a 5% topical gel. Injections include botulinum toxins and specifically Type A that are used for the treatment of muscle pain disorders and act presynaptically through inhibition of acetylcholine synthesis or its release. This blocks neuromuscular transmission at the neuromuscular junction, causing paralysis of the injected skeletal muscle. It’s presumed that injecting an overactive muscle will decrease its level of contraction and allow improved reciprocal motion, improving movement and the ability to exercise. Side effects are generally minor and temporary but rare allergic reactions can occur. Finally, corticosteroid injections are administered intraarticularly and intramuscularly with the thought that they reduce inflammation (pain and swelling) at the injury site. Short term problems with injections include flushing, transient hypertension, and serum glucose fluctuation. Chronic corticosteroid use can lead to hypergycemia, insulin resistance, hypertension, weight gain, osteoporosis, anxiety, depression and cataracts.

In a previous 2009 overview, we found an insufficient evidence base on the benefits and risks of most pharmacological interventions for neck pain including the commonly used injections, which limits the ability to provide strong clinical guidance on appropriate use [7]. We found limited evidence supporting methylprednisone for acute whiplash, intramuscular lidocaine for chronic neck pain and epidural methylprednisone and lidocaine for chronic neck pain. We recommended against botulinum toxin-A as it was not found superior to saline for chronic neck pain.

There have been further clinical trials in neck pain patients since our 2009 review and since neck pain continues to be a common potentially disabling clinical condition, we wanted to update the evidence on oral, topical and injected medications for neck pain. The purpose of this overview is to systematically review existing reviews of randomized controlled trials published after 2006 and to consider establishing evidence-based recommendations on medicines and medicinal injections for neck pain, regardless of pain duration across several diagnostic groups (e.g. including non-specific and specific neck pain with and without cervicogenic headache, radiculopathy or associated whiplash injury), while also considering varying duration of study patient follow-up (short- and long-term) and differing control groups (placebo control or active treatment) in these trials. Clinical trial outcomes we considered of primary interest were pain, function, disability, work related function, patient satisfaction, global perceived effect and quality of life as well as adverse effects of these medicines.

Our systematic overview process included comprehensive computerized search strategies from January 2000 to August 2010: MEDLINE, EMBASE, CINAHL, ILC, CENTRAL, and LILACS, with selection criteria listed in Table 1 and at least 2 independent reviewers selecting articles, performing methodological quality assessment using the AMSTAR tool [8] qualitative assessment of the strength of evidence. We used a team consensus approach to qualitatively assess the totality of the evidence, using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology [9]. Further details on the methodological approach are provided in our ICON methods report [10] including search terms and strategies. Two separate searches were performed, one for treatment benefits and one for harms. The protocol for this overview was not registered.

We supplemented our computerized search strategy by identifying on-going systematic reviews in the grey literature nearing completion (e.g. Cochrane Reviews) up to July 2012, by asking our expert panel to identify ongoing reviews and by scrutinizing the reference lists of all of the primary studies.

Data extraction was performed using forms we piloted - first by one reviewer and then checked by a second. Disagreements were resolved through review of data extraction forms, discussion and consensus. We systematically extracted data from selected systematic reviews and produced evidence tables. Key factors extracted from the original reviews included the following three items: 1) descriptive features of the original review (e.g. authors, publication year, disorder, symptom duration, intervention, and comparator used such as placebo, no care, usual care, and other treatment), as well as noting the authors of primary studies included in the review; 2) methodological details of the original review (e.g. search period, AMSTAR score, quality ranking system, evidence statement and final GRADE). If the original review did not report using GRADE methodology, an estimate for GRADE was made by us based on the methodological details in the systematic review; and 3) data on benefits and risks (e.g. effect size, effect direction, duration of follow-up, reports of harm). Summary statements on harms included information reported in the original reviews as well as information obtained from the second literature search on harms.

We employed the following a priori triage rules to facilitate decisions on including/excluding reviews:

1. Type of treatment (analgesic, NSAID, opioid, etc.) reviews by drug class. See APPENDIX 1 for a complete list of medications and injections considered by medicine treatment category.
2. Within a treatment drug class we grouped data by type of comparator (placebo, active treatments, etc.).
3. We prioritized the highest quality reviews based on the rules below, PER grouping.
   1. If there were few reviews, we retained them all.
   2. If there were several reviews reporting on the same treatments and comparators, we retained the highest quality reviews, using the approach recommended by Whitlock [11]. Whitlock has suggested considering the following factors: i. Year of publication. We selected the most recent reviews when the data was similar across reviews and there was no loss of studies contained in the older reviews. We further ensured consistency among reviews’ conclusions before eliminating older reviews. Inconsistency and discordance were highlighted and potential reasons for differences were discussed; ii. AMSTAR- risk of bias. We prioritized reviews with a low risk of bias. Reviews that scored 8 or higher on the 11-point AMSTAR scale were considered at low risk of bias; moderate risk of bias was considered for scores between 5 and 7; and a high risk of bias was assigned to scores of 4 or less. These various reviews were then further summarized in a “Summary of Findings” table to facilitate incorporating this information into clinical practice. Inconsistency and discordance were highlighted and discussed across reviews; iii. Effect size estimates: We considered effect sizes as the primary summary measure. Within our defined groups of treatments and comparators, we selected a review that best represented the treatment effect sizes (including through meta-analysis) although we also report the range of estimates from the other included reviews. In cases where there was a large discordance between reviews, we reported our own analysis using the individual studies included in the reviews. Additional data on magnitude of effects such as number-needed-to-treat (NNT) and weighted mean difference (WMD) were also extracted when possible. Further we also considered the clinical importance of these effects using several guiding principles. We considered the published data on the minimal detectable change and the minimal clinically important difference for that outcome. We used a change from baseline of ≥ 15% to represent the MCID when it was not otherwise published. We also considered the magnitude of the treatment effect (represented by WMD, NNT, absolute benefit, treatment advantage), the evidence for a dose-response gradient, and evidence on the duration of effect (See APPENDIX 2) [12-15] in our assessment of clinical relevance.
4. Strength of Evidence using GRADE approach: We considered the same prioritized systematic review to represent the body of evidence for any treatment and assigned an overall GRADE on the strength of evidence. If the selected (prioritized) reviews already reported a GRADE table, we used that. As a reminder, the GRADE approach assessing the quality of evidence from primary trials considers information on design (randomized controlled trials or RCT), information on timing of outcomes (immediately post treatment or IP to long term or LT follow-up); risk of bias; imprecision based on sample size; inconsistency across trials; indirectness and reporting bias.

From 10, 055 reviews that were screened and 117 eligible reviews relating to neck pain filters and adverse events filters, 43 reviews related to medicines were ultimately considered for inclusion in this report. A total of 26 treatment reviews and 6 harm related reviews were included for this topic (see Fig. 1 - PRISMA diagram [16]). Excluded reviews are presented in Appendix 3, along with an accompanying rationale for their exclusion. Using the selected (prioritized) reviews, we report on trial findings by “overall quality of evidence” using the GRADE approach and stratified by the pre-determined treatment category in the Summary of Findings table (Table 2) [17-90]. We report on conflicting evidence across reviews in Table 3 [91, 92]. Our final recommendations are summarized in the Evidence-based Recommendation table and provided in Table 4.The AMSTAR assessment (see Table 5) revealed that the most common methodological limitations among the included reviews were incomplete reporting on: publication bias; conflict of interest; and full reporting of excluded studies. Details on risk of bias (AMSTAR scoring) are available in the companion methods paper by Santaguida et al. APPENDIX 2 to this report provides the details on why 17 medicines reviews were excluded. Table 2 provides summary findings by treatment category and includes the primary trials and related systematic review(s) used to compile recommendations in this systematic review. The evidence tables, ‘Characteristics of Included Studies and GRADE rationale’, underpinning the summary provided in Table 2 are available from the authors. Table 6 [93-98] summarizes the findings on harms. The primary reviews included in our analyses considered the following medicines and medical injection therapies: anti-inflammatories and analgesics in combination, anti-inflammatories alone, analgesics alone, anesthetics such as lidocaine intramuscular (IM and topical nerve blocks), muscle relaxants, neurotropic multivitamins (IM), psychotropic agents, sterile water (IM, subcutaneous and intracutaneous), subcutaneous insufflation, botulinum toxin-A (IM), and corticosteroids (intra-articular, intravenous, epidural).

Table 2.

Summary of Findings by GRADE (Quality of Evidence)

Table 4.

Evidence-Based Recommendations

Table 6.

Harms Summary of Findings with AMSTAR Score

Fig. (1). PRISMA diagram showing the flow of reviews.

There is striking lack of trials and evidence for pharmacological therapies commonly used in neck pain. For subacute or chronic WAD, the evidence strongly recommends against the use of botulinum-A to reduce pain, improve disability or global perceived effect after short-term follow-up. For chronic facet joint pain and related disability, the evidence suggests against the use of medial branch block with steroids from short- to long-term follow-up data. For chronic neck pain, the evidence supports the use of only one muscle relaxant (psychotropic agent), eperison hydrochloride. There is limited efficacy with this agent however as it will help one in 37 people achieve immediate pain relief and evidence for longer-term benefits are not available.

This section provides data favouring the use of certain oral medication and medical injections by the GRADE of evidence.

There continues to be a lack of high-quality evidence to inform recommendations on the use of medicinal injections and medicines for neck pain. Only one trial meeting the moderate quality evidence threshold and reporting evidence of benefit was uncovered. The trial by Bose et al. supported the use of an oral psychotropic agent, eperison hydrochloride, for subjects with chronic mechanical neck disorder for immediate post treatment pain and range of motion but longer term follow-up was not reported and the treatment effects were small (1 expected to benefit for 37 treated).

There was more data on evidence of no benefit. In the review by Langevin et al. [26], a meta-analysis of four trials provided a strong quality of evidence demonstrating no benefit for botulinum toxin-A over placebo for subjects with chronic non-specific neck pain for short-term pain. Additionally in the same review by Langevin et al, another meta-analysis of four trials with moderate quality evidence demonstrated no benefit for botulinum toxin-A over placebo for subacute or chronic WAD patients for short-term pain, disability or global perceived effect. Two reports of one trial by Manchikanti et al. [33, 34] did not support the use of nerve block injections with bupivacaine and varying combinations of steroid and sarapin for subjects with cervical facet joint pain for short- and long-term pain.

Although oral medication such as analgesics, anti-inflammatories, antidepressants, opioids, psychotropic, and muscle relaxants are commonly used in clinical practice, there continues to be low to very low quality evidence available for their benefits and risks in neck pain. Furthermore the existing evidence is conflicting, which limits the ability to make clear recommendations. Data on disability, function, and quality of life are rarely reported. Most pharmacological therapies would be expected to produce side effects and their balance of risks and benefits are likely to vary by the condition being treated. We speculate that physicians assume that injections and medications demonstrating efficacy for other musculoskeletal conditions such as in low back pain [100] inform their use in neck pain and that clinical trials are not being conducted or not considered necessary in the neck pain population. However we are not aware of data that suggests treatment benefits seen in the back pain population can be extended to the neck pain populatoin. Given that neck pain is common, potentially disabling and costly to society, and that benefit-risk may well vary by condition, high quality studies are still needed to understand the benefits and risks in the neck pain population. Our qualitative research suggests that the side effects associated with medication use can be very concerning for patients and patients may discontinue medications related to these concerns [101]. Specifically patient's worried about how medications would interfere with their ability to participate in normal life roles [101]. Further, patients in our qualitative study indicated that they prefer that physicians present all treatment options and not confine their recommendations to prescriptions of medications alone [101].

For medicinal injections, the therapies with the most supporting evidence continue to be IV methylprednisolone for acute whiplash, IM-lidocaine for chronic MND, and epidural methylprednisonlone with lidocaine injection for radiculopathy. It remains unclear if all corticosteroids or local anesthetics are equally effective or if there is a dose response for these therapies. Replication in larger, high quality trials is needed for these injections. If subsequent trials were positive, efforts to promote widespread adoption would be indicated. Anti-inflammatory drugs warrant further study particularly since several of them such as ibuprofen and naproxen may be available over the counter without a prescription; we note that they are also frequently used as co-intervention with other physiotherapy management approaches. Oral psychotropic agents classified as muscle relaxants, such as cyclobenzaprine, diazepam, and tetrazepam continue to require further study to clarify their benefits and harms. There were no studies of tricyclic antidepressants and one very low quality trial on opiate analgesics in chronic neck pain. In this regard, little has changed since the 1996 [102] and 2007 [20] systematic reviews.

Another challenge in making more definitive recommendations was the lack of high-quality clinical trials that addressed meaningful outcomes in a standardized way. Many studies focus on pain alone with function being reported to a lesser extent. Even when function is reported, different outcomes are used. Consistent use of pain and disability outcomes would facilitate cross study comparisons and inform future metaanalyses. A number of reporting and design issues in neck pain clinical trials have been detailed in Goldsmith et al. [103]. There are design features that would improve the quality of future clinical trials in neck pain. In particular, future research should ensure adherence to CONSORT guidelines, look beyond the basic two group design (active vs placebo or active vs active) that is commonly used and the should consistently report standardized impairment and disability outcomes [104]. A core set of patient reported outcomes and key participation indicators (such as return to work) are needed. Further an accurate prospective collection of adverse events is also fundamental. Finally, studies that compare medicines or injections to other commonly used therapies, such as physical therapies and manual therapies are needed to understand whether some therapies should preferentially be recommended.

The ability to generalize our findings across the entirety of the neck pain population is limited from at least four perspectives. First, there are many disorders that can have neck pain as an associated feature, such as migraine, tension type headache and trigeminal neuralgia. A careful differential diagnosis and its reporting are important when describing neck pain patients included in any trial. This report specifically applies to a narrow type of mechanical neck pain with or without cervicogenic headache or radiculopathy. Second, when considering botulinum toxin injections, identification of the dystonic muscle may need EMG guidance to ensure the right region or muscles are injected; this was not the case in the trials included in this review. Third, once neck pain has been identified, the mechanisms underpinning the pain experience need to be identified where possible. Mechanisms underlying peripheral or central neuropathic pain will be treated differently than nociceptive mechanisms such as chemical [inflammatory from tissue (bradykinin, prostaglandin, serotonin), nerve (NGK, neurokinins, noradrenaline), immune cells (macrophages, cytokines)], ischaemic (SP, CGRP) and mechanical. Understanding the pain experience could help to optimize different medical strategies. Finally, understading why some patients respond and other do not would also allow the most appropriaat therapy to be directed to the right patients. Clinical decision rules may then help guide practice.

Our approach to summarizing the literature has several strengths. We used a comprehensive, librarian-assisted search of multiple databases. We used two independent reviewers to determine article relevance and assess review quality using the AMSTAR tool. We used at least two people to verify data extraction. We used a group and an external panel consensus approach to validate the GRADE of evidence and recommendations. We avoided the professional bias inherent in having a single professional group evaluate its own literature. The largest limitation is that new trials will have been published following the publication of many of these reviews. Given the state of the literature, one large trial with a low risk of bias could change the direction of benefit (positively or negatively) as well as the magnitude of that benefit. We did consider the potential for selection bias and we examined the grey literature of reviews as part of our search strategy.

The vast majority of trials did not lead to firm recommendation, since few injections and medicines results have been replicated by large, high quality trials. Although we did find some meta-analyses (the highest level of evidence), the quality of studies for neck pain continues to be limited. Adverse effects and associated costs of treatment are widely under-reported and when they are reported are they often in narrative form rather than quantitatively expressed. There is evidence of major catastrophic adverse events with all of the injection procedures. Approximately 1 in 5 patients will experience minor transient adverse events with transforaminal or interlaminar epidural steroid injections. No clear estimate of the overall incidence rate for any injection procedure is available due to a lack of reporting. Although we did not find any specific reviews reporting harms associated with oral medication for neck pain, we can be informed by reviews that report adverse events for other musculoskeletal pain disorders for some guidance.

A few systematic reviews of antidepressant side effects have been conducted and Perrot et al. (2008) note that when prescribed for painful conditions, side-effects occur in 30 to100% of patients and are often dose-dependent [105]. Reported side effects are dysuria, constipation, dry mouth, drowsiness, eye accommodation disorders, tachycardia, memory disorders and confusion, orthostatic hypotension, dizziness, weight gain, trembling, impotence, nausea, fatigue and serotonergic syndrome. Hauser et al. (2012) provide relative risk estimates for similar mild transient adverse events ranging from 0.94 (0.46, 1.68) for headache to 9.51 (1.22, 74.0) for sexual dysfunction [106].

There is a vast literature informing NSAIDS adverse events although the great majority is related to oral and compared to topical use. Zhang et al. report that gastrointestinal events are 3 to5 times higher compared to placebo, while Jones et al. report that up to 60% of all NSAID users will experience a GI event [107, 108]. Canadian guidelines for NSAID use reported that the rate of gastrointestinal hemorrhage is 1.5 to 2.0% per year for average risk patients while Jones et al. report a rate of 1.0 to 1.5% [108, 109]. The Osteoarthritis Research Society International (OARSI) guidelines reported that cardiovascular events are comparable in the COX-2 class with the traditional NSAIDS blocking both COX-1 and COX-2 (RR 1.19 95% CI 0.80, 1.75) [107], while Jones et al. argue that selective COX-2 inhibitors and NSAIDs could have different rates [108].

Opiates were considered in a recent systematic review of conservative management for low back pain. Opiate use was associated with a greater risk of headache and nausea, drowsiness, dry mouth, constipation, pruritis, vomiting, anorexia and increased sweating over placebo. The risk differential was low and ranged from 3 to 9% [110]. The OARSI guidelines report a relative risk for any adverse event related to opiate use at 1.4 (95% CI 1.3, 1.6) and specifies constipation, nausea, drowsiness, dizziness and vomiting as side effects.

A Cochrane review reported that there was no significant difference in the overall safety of Acetaminophen compared to NSAIDS [111]. Acetaminophen can be associated with GI events but at a lower rate than that of NSAIDS. Acetaminophen can lead to renal failure however with a RR up to 2.5 (95% CI 1.7, 3.6) [90].

There is a lack of trials in neck pain for common injections and medications and this leads to an inability to fuller inform the proper use of pharmacological therapies. The current state of the evidence appears to favor the muscle relaxant eperison hydrochloride for chronic neck pain. Evidence is emerging against IM botulinum toxin-A injections for chronic mechanical neck disorder or subacute/chronic WAD and against medial branch block with steroids for chronic facet joint pain. Given the limited number of trials and the low level of evidence in those that have been performed, coupled with the frequent and disability nature of neck pain, high quality trials of analgesics, anti-inflammatories, and psychotropic agents are urgently needed.

AMSTAR Assessment of the Methodological Quality of Systematic Reviews; CI Confidence Interval; COX Cyclooxygenase; GRADE Grading of Recommendation Assessment, Development, and Evaluation; ICON International Collaboration on Neck; IM intramuscular; IP immediate post; LT long-term; MCID Minimal Clinically Important Difference; MND mechanical neck disorder; NDI Neck Disability Index; NNT Number Needed to Treat; NSAID non-steroidal anti-inflammatory drug; PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT randomized controlled trial, RR relative risk; SMD_p Standard Mean Difference (pooled); VAS Visual Analogue Scale; WAD whiplash associated disorder; WMD Weighted Mean Difference.