Paracetamol did not beat placebo in the trial built to test it for back pain

For acute low back pain, paracetamol performs the same as placebo: in the only large trial designed to test it, median time to recovery was 17 days on regular dosing, 17 days on as-needed dosing and 16 days on placebo [s1]. For osteoarthritis of the hip and knee it does something, but the most recent head-to-head trial could not show it was as good as an anti-inflammatory, and a larger synthesis put it behind topical NSAIDs [s2] [s3].

That is an awkward result for a drug that was, for years, the recommended first-line analgesic for both conditions largely on the grounds that it was safe rather than that it had been shown to work.

The back pain trial

PACE was a multicentre, double-dummy, randomised, placebo-controlled trial run across 235 primary care centres in Sydney, Australia, between 11 November 2009 and 5 March 2013 [s1]. Patients with acute low back pain were allocated 1:1:1 to up to four weeks of regular paracetamol three times a day (equivalent to 3,990 mg per day), as-needed paracetamol taken for pain relief (maximum 4,000 mg per day), or placebo [s1]. Everyone received best-evidence advice and was followed for three months. The primary outcome was time until recovery, defined as a pain score of 0 or 1 on a 0–10 scale sustained for seven consecutive days [s1].

550 patients were assigned to the regular group, 549 to as-needed and 553 to placebo [s1]. Median time to recovery was 17 days (95% CI 14–19) with regular dosing, 17 days (15–20) as needed, and 16 days (14–20) on placebo [s1]. The hazard ratios for recovery were 0.99 (0.87–1.14) for regular versus placebo and 1.05 (0.92–1.19) for as-needed versus placebo, with an adjusted P value of 0.79 across groups [s1].

Two details close off the usual escape routes. Adherence was essentially identical — a median of 4.0 tablets per participant per day in the regular group, 3.9 as-needed and 4.0 on placebo, out of a maximum of six [s1] — so this was not a trial where nobody took the drug. And adverse events were equally common in all three arms: 99 participants (18.5%) on regular dosing, 99 (18.7%) as-needed, and 98 (18.5%) on placebo [s1].

The American College of Physicians guideline published three years later recommends, for patients with acute or subacute low back pain who want a drug, non-steroidal anti-inflammatory drugs or skeletal muscle relaxants [s4]. Paracetamol is not among them.

Osteoarthritis: a smaller failure

The osteoarthritis evidence is not the same null result, but it is not a vindication either.

RETHINK was a multicentre, randomised, double-blind, parallel-group non-inferiority trial in patients aged 65 or older with osteoarthritis-related pain of the hip and knee, run across five institutions in Japan and registered as jRCTs071200112 [s2]. It compared acetaminophen at 1,800 mg per day against NSAIDs — loxoprofen 180 mg per day or celecoxib 200 mg per day [s2]. Of 400 patients enrolled, 191 were in the acetaminophen group (mean age 73.6, 83.2% female) and 197 in the NSAID group (mean age 73.3, 74.6% female) [s2].

The primary endpoint was change in worst pain on the Brief Pain Inventory from baseline to week 8. Least-squares mean change was −1.79 with acetaminophen and −1.94 with NSAIDs, a between-group difference of 0.14 (95% CI −0.33 to 0.61) [s2]. Both groups improved, and the two were close. But the trial's own conclusion is that non-inferiority was not demonstrated [s2] — the confidence interval was too wide to rule out a meaningful disadvantage. On safety, gastrointestinal disorders occurred more frequently with NSAIDs and were the most common cause of stopping treatment [s2].

A larger network meta-analysis of 122 randomised trials in 47,113 participants with knee osteoarthritis places acetaminophen behind topical NSAIDs for function (standardised mean difference −0.29 favouring topical NSAIDs, 95% credible interval −0.52 to −0.06), with topical NSAIDs statistically indistinguishable from oral ones (SMD 0.03, −0.16 to 0.22) [s3]. Topical NSAIDs also carried a lower risk of gastrointestinal adverse effects than acetaminophen in the trials (risk ratio 0.52, 95% CrI 0.35–0.76) [s3].

The observational safety numbers need a caveat

The same paper reports a propensity-matched cohort analysis in which topical NSAID users had lower one-year risks than acetaminophen users of all-cause mortality (hazard ratio 0.59, 95% CI 0.52–0.68), cardiovascular disease (HR 0.73, 0.63–0.85) and gastrointestinal bleeding (HR 0.53, 0.41–0.69), across 22,158 participants per group [s3].

A hazard ratio of 0.59 for death from any cause, attached to a topical gel, is not a plausible pharmacological effect. It is much more likely to reflect who receives which drug: oral paracetamol is prescribed to frailer, sicker and older patients, and propensity matching on recorded variables does not fix confounding by the things that are not recorded. The randomised part of that paper is the part worth weighting.

Where this leaves the drug

Two honest statements can be made. Paracetamol has failed the best available test for acute low back pain [s1]. For osteoarthritis it produces a real reduction in pain — the RETHINK acetaminophen arm improved by 1.79 points on the Brief Pain Inventory worst-pain item over eight weeks [s2] — that is probably smaller than an anti-inflammatory's and has not been shown to be equivalent [s2] [s3].

The comparison that puts all of it in perspective comes from a network meta-analysis of 152 randomised trials in 17,431 participants with knee or hip osteoarthritis, which found no difference between exercise therapy and oral NSAIDs and paracetamol for pain at 4, 8 or 24 weeks, or for function at any of those time points [s5]. The authors' reading is that exercise has similar effects to the drugs with an excellent safety profile, and should be given more prominence in clinical care, particularly in older people at higher risk of drug-related adverse events [s5].

What to watch

Whether guidelines outside back pain follow the ACP in dropping paracetamol from first-line recommendations, and whether any adequately powered non-inferiority trial in osteoarthritis succeeds where RETHINK did not.

This article is informational and is not medical advice. Doses given here are trial parameters, not guidance; decisions about analgesics belong with a reader and their clinician.

Sources

• Efficacy of paracetamol for acute low-back pain: a double-blind, randomised controlled trial — The Lancet, 2014-07-23
• Comparison of the efficacy and safety of acetaminophen versus NSAIDs for the treatment of chronic pain in older adults with osteoarthritis of the hip and knee: the RETHINK study — Osteoarthritis and Cartilage Open, 2026-07-06
• Comparative efficacy and safety of acetaminophen, topical and oral non-steroidal anti-inflammatory drugs for knee osteoarthritis — Osteoarthritis and Cartilage, 2021-06-24
• Noninvasive Treatments for Acute, Subacute, and Chronic Low Back Pain: A Clinical Practice Guideline From the American College of Physicians — Annals of Internal Medicine, 2017-02-13
• Comparative efficacy of exercise therapy and oral non-steroidal anti-inflammatory drugs and paracetamol for knee or hip osteoarthritis — British Journal of Sports Medicine, 2023-01-02