Comparative Effectiveness of Muscle Relaxants in Acute Cervicalgia and Low Back Pain

Low back pain has been the leading cause of disability worldwide for years now.^1,2 An estimated 619 million people lived with low back pain in 2020, and projections push that figure toward 843 million by 2050.¹ Years lived with disability attributable to LBP rose by more than 50% between 1990 and 2015 — a trajectory driven mostly by population growth and aging, not by any single clinical failure.² Roughly 80% of adults will experience low back pain at some point in their lives, and a meaningful slice of them will end up in a clinic for an acute episode.²

Cervicalgia is the quieter sibling, but the burden is comparable. Annual prevalence of neck pain runs between 30% and 50% in adult populations.⁴ The 2020 global age-standardized prevalence was around 2,450 per 100,000.³ Both conditions hit women harder, peak in mid-life, and travel with an entourage of comorbid musculoskeletal, mood, and sleep disorders.^4,5 In the United States, the two together generated $134.5 billion in healthcare expenditures in 2016 — more than any other health condition that year.⁶

What does evidence-based care look like? The American College of Physicians places non-pharmacologic therapies — heat, massage, acupuncture, spinal manipulation — at the top of the list, with NSAIDs and skeletal muscle relaxants as the preferred pharmacologic options when drugs are wanted.⁹ Layered on top is the opioid stewardship imperative: the 2022 CDC clinical practice guideline urges clinicians to maximize non-opioid options for acute musculoskeletal pain and explicitly flags that opioid initiation for acute LBP correlates with prolonged use down the line.^10–12 Yet prescribing patterns remain stubbornly heterogeneous, reflecting persistent uncertainty about which agents work best for whom.^7,8

The evidence base for muscle relaxants is honest about this. A Cochrane review concluded that non-benzodiazepine muscle relaxants are effective for acute LBP — with a real adverse-event tax, mostly drowsiness and dizziness.¹⁵ A 2023 Cochrane overview reaffirmed moderate-certainty evidence that antispasmodic muscle relaxants improve short-term pain and physical function versus placebo.¹⁶ But the most rigorous emergency-department RCT — Friedman and colleagues, JAMA 2015 — found that adding cyclobenzaprine to naproxen produced no improvement in pain or function at one week or three months versus naproxen plus placebo.¹³ A follow-up trial showed similar null results for orphenadrine and methocarbamol.¹⁴ So which is it? Do these drugs work in routine practice, and if they do, are different agents within the class actually different in ways that matter at the bedside?

Real-world evidence from large EHR networks can speak to questions RCTs can't easily answer — actual prescribing patterns, more diverse populations, downstream events like repeat visits and opioid escalation.^23,24 This study uses propensity-score-matched real-world data to compare three commonly prescribed muscle relaxants — tizanidine, cyclobenzaprine, and baclofen — against NSAID monotherapy in adults with acute cervicalgia or low back pain.

Study Design and Data Source

Retrospective cohort study using the TriNetX federated health research network, which aggregates de-identified EHR data from more than 130 healthcare organizations across multiple countries.^23,24 Reporting follows the STROBE statement.²⁷

Population

Inclusion criteria:

• Adults aged ≥18 years
• Diagnosis of acute cervicalgia (ICD-10-CM M54.2) or low back pain. ICD-10-CM coding for low back pain was revised effective October 1, 2021, with deletion of M54.5 and replacement by M54.50 (Low back pain, unspecified), M54.51 (Vertebrogenic low back pain), and M54.59 (Other low back pain). Patients meeting any of these codes during the respective active periods were eligible
• Initiation of pharmacologic therapy within 48 hours of index diagnosis

Exclusion criteria:

• Chronic pain diagnoses or pain documented for >12 weeks before index
• Spine surgery within the previous 6 months
• Baseline opioid use (any opioid prescription within 90 days before index)
• Multiple sclerosis, spinal cord injury, or cerebral palsy — to keep on-label baclofen indications out of the comparison

Of 5,842 patients meeting these criteria, 2,136 entered the matched analysis (534 tizanidine, 536 cyclobenzaprine, 533 baclofen, 533 NSAID monotherapy).

Treatment Groups

• Tizanidine (including brand-name Zanaflex)
• Cyclobenzaprine
• Baclofen — included to reflect real-world prescribing in acute musculoskeletal spine pain. Worth flagging upfront: baclofen is not FDA-labeled for acute non-spastic LBP. Its presence here captures an off-label pattern that exists in the data, not a recommendation
• NSAID monotherapy (reference)

Outcomes

Primary outcome. Day-7 change in NRS (0–10) pain score, in the subset with documented baseline and follow-up values (n=812; 38.0% of the matched cohort). Baseline characteristics in this subset matched the full cohort across age, sex, index diagnosis, and comorbidity burden (all SMD <0.10), which supports treating it as broadly representative rather than a self-selected slice.

Secondary outcomes (full matched cohort):

• Any opioid prescription within 30 days of index
• Repeat pain-related ambulatory or emergency visits within 30 days
• Documented adverse events within 30 days, identified by ICD-10-CM codes mapped a priori to predefined categories: sedation/somnolence (R40.0), dizziness (R42), neurological adverse effects (G93.4, G25.9, G92.x), gastrointestinal effects (K30, R10.x), and hypotension (I95.x). Events were attributed to the index treatment if recorded between Day 0 and Day 30 with no occurrence in the preceding 6 months

Statistical Analysis

Propensity scores were estimated by logistic regression on a pre-specified set of covariates: age, sex, race/ethnicity, BMI, smoking status, index diagnosis (cervicalgia vs LBP), Charlson Comorbidity Index, prior NSAID use, prior physical therapy, baseline depression and anxiety diagnoses, and number of healthcare encounters in the prior 12 months.²⁵ One-to-one nearest-neighbor matching was performed without replacement using a caliper of 0.2 standard deviations of the logit of the propensity score. Post-matching balance was checked using standardized mean differences, with SMD <0.10 the cutoff for adequate balance.²⁶

Continuous outcomes: paired t-tests. Categorical outcomes: McNemar's test. Logistic regression yielded adjusted odds ratios (OR) with 95% CI. All comparisons were pre-specified; p<0.05 was treated as statistically significant. Analyses ran inside the TriNetX analytics platform.

Baseline Characteristics

After matching, baseline characteristics were well-balanced across treatment arms (all SMD <0.10).

Primary Outcome — Pain Intensity (NRS Subset, n=812)

All three muscle relaxant groups produced greater pain reduction than NSAID monotherapy. Confidence intervals across the muscle relaxant arms overlapped — meaning the agents performed roughly similarly on pain itself.

Opioid Utilization (30 Days)

Only tizanidine cleared the significance threshold against NSAIDs. The absolute risk reduction was 7.9 percentage points, which translates to an NNT of about 13 to prevent one opioid prescription within 30 days.

Repeat Pain-Related Visits (30 Days)

All three muscle relaxant groups had significantly fewer repeat pain-related visits than NSAID monotherapy.

Adverse Events

Cyclobenzaprine showed the highest rate of documented sedation/somnolence (8.4%) — not surprising given its tricyclic structure and well-known anticholinergic activity.²⁰ Baclofen produced more neurological adverse effects (6.1%), including dizziness and altered mental status, in line with its central GABA-B mechanism.²² Tizanidine sat between the two on sedation (5.7%), with most events transient. Adverse events in the NSAID arm (3.2%) were predominantly gastrointestinal.

In a propensity-matched real-world cohort of more than 2,000 adults with acute cervicalgia or low back pain, muscle relaxants outperformed NSAID monotherapy on short-term outcomes — modestly, but consistently. The headline result is not that they worked. It is how they differed.

Day-7 NRS reductions were close across tizanidine, cyclobenzaprine, and baclofen. All three did somewhat better than NSAIDs alone, with overlapping confidence intervals among muscle relaxants. So far, an unsurprising story. The interesting divergence lives in what happened next: only tizanidine was linked to a statistically significant drop in 30-day opioid prescribing (OR 0.76; NNT ≈ 13). Cyclobenzaprine and baclofen pointed in the same direction but did not get there. The mechanistic picture fits — tizanidine is a central α2-adrenergic agonist, and its action on dorsal-horn interneuron excitability gives it antinociceptive effects independent of muscle relaxation per se.^18,19 In other words, less downstream pressure to reach for an opioid.

How do these findings square with prior randomized trials? The most cited counterpoint is Friedman et al. (JAMA 2015), where adding cyclobenzaprine to naproxen produced no improvement in functional outcomes versus naproxen plus placebo in an ED LBP cohort.¹³ A few things probably explain the apparent gap. The Friedman trial drew from a single urban ED with acute, severe presentations; our cohort spans broader ambulatory and ED practice. Friedman's primary outcome was Roland-Morris functional change — different from the pain-intensity and downstream-utilization endpoints we used. And in real-world practice, drug choice itself reflects clinician judgment about case complexity; propensity score matching mitigates this but cannot eliminate it. The convergence with the broader Cochrane evidence base — antispasmodic muscle relaxants effective for short-term pain and global improvement — is reassuring.^15,16 But these findings should sit alongside the randomized data, not replace it.

Tolerability differences are the part of this picture clinicians will recognize fastest. The sedation signal for cyclobenzaprine is a class effect of its tricyclic-like pharmacology, and it is exactly the reason cyclobenzaprine sits on the American Geriatrics Society Beers Criteria list of medications to avoid in adults aged ≥65.²¹ Baclofen's neurological adverse-event profile — dizziness, altered mental status, occasional metabolic encephalopathy — reflects its central GABA-B agonism and is well-documented in pharmacovigilance data.²² Tizanidine's profile in this cohort, mostly transient sedation, supports its use when daytime function needs to be preserved.

Implications for spine practice

For the spine community, the most actionable signal is the opioid-sparing effect of tizanidine. A 24% relative reduction in 30-day opioid prescribing is not a small thing in 2026 — it is the kind of marginal gain that adds up across a practice. A few clinical scenarios where this matters:

• Patients with acute axial pain awaiting surgical evaluation
• Patients deferring surgery, where conservative management is expected to do real work
• Pre-procedural optimization pathways, where minimizing perioperative opioid exposure has become a quality metric in its own right
• Older adults — where cyclobenzaprine is essentially off the table per Beers Criteria²¹ and tizanidine becomes the more sensible choice within the class

The broader takeaway: muscle relaxants are not interchangeable. Choosing between them on the basis of perceived equivalence misses real differences in opioid-sparing potential and tolerability that show up only at scale, in data like this.

Limitations

Several caveats. First, this is a retrospective observational study; residual confounding by indication cannot be excluded despite propensity score matching. Second, NRS scores were available for only 38% of patients — a reflection of how heterogeneous documentation practices are across TriNetX-contributing organizations. The NRS subset was demographically representative of the full cohort, but selection effects on who actually gets a pain score recorded cannot be fully ruled out. Third, adverse-event capture relied on ICD-10-CM coding within the EHR, which under-counts milder symptoms (transient drowsiness, for example) that patients may not bring up to clinicians. Fourth, TriNetX data lean toward academic and large healthcare systems and may underrepresent rural and uninsured populations.²³ The October 2021 ICD-10-CM revision (deletion of M54.5; introduction of M54.50/51/59) introduces some heterogeneity in case ascertainment across the study period — addressed here by including all relevant codes during their active windows. And finally: associations are not causation. Randomized comparative-effectiveness trials of tizanidine versus other muscle relaxants in acute spine pain are still warranted.

In a large propensity-matched real-world cohort of adults with acute cervicalgia or low back pain, all three muscle relaxants studied produced modest short-term pain reduction — comparable to one another, and ahead of NSAID monotherapy. Tizanidine alone was associated with a statistically significant reduction in 30-day opioid prescribing (NNT ≈ 13). Tolerability profiles differed in clinically meaningful ways: more sedation with cyclobenzaprine, more neurological adverse events with baclofen. Taken together, the data argue for picking within the muscle relaxant class rather than treating these agents as interchangeable, and for considering tizanidine specifically when opioid-sparing is the goal.