The search retrieved 13,186 records, of which 22 studies were included (Fig. 1). A list of studies excluded at full text are presented in the ESM. Studies were published between 1976 and 2021. All studies were published in English. Sixteen studies had some form of industry sponsorship. Sample sizes ranged from 50 to 511 participants. Study characteristics are summarised in Table 1.

PRISMA (preferred reporting items for systematic reviews and meta-analyses) flow chart. WHO ICTRP World Health Organization International Clinical Trials Registry Platform

Most studies focussed on participants with low back pain (16 studies [29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44]), followed by osteoarthritis (five studies [45,46,47,48,49]), and one study enrolled both populations [50]. Nine studies focussed on participants with acute symptoms [29,30,31, 38, 43, 44, 46,47,48], including three with acute symptoms on chronic presentation (osteoarthritis “flare up”) [46,47,48], one study of subacute symptoms (10–42 days) [40] and the remaining studies with chronic symptoms [32,33,34,35,36,37, 39, 41, 42, 45, 49, 50]. Of the low back pain studies, only one study included participants with intervertebral disc disease/nerve root entrapment [33]; other studies actively excluded participants with symptoms or neurologic deficits in the lower extremities [29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44, 50], or did not state they were included [35,36,37,38,39,40,41,42, 44, 50].

The mean age of participants was 52.8 years (SD 11.0; 15 studies, range 37.0–60.1 years). Studies involving participants with low back pain were, on average, younger (49.3 years, SD 11.4; 12 studies) than studies with participants with osteoarthritis (63.8 years, SD 7.4; 3 studies). Participant characteristics are summarised in Table 1.

Forty percent of risk of bias scores were marked as unclear because of insufficient available information. This was driven by the lack of detail on random sequence generation, allocation concealment and blinding in studies found through clinical trial registries when no peer review publication was associated with the trial [33,34,35,36,37, 50]. These studies also scored poorly in the remaining domains. Eight studies scored a low risk of bias in at least half the domains [29, 30, 32, 39, 41, 42, 44, 49] (Fig. 2).

All but one trial [43] delivered combination therapy orally (e.g. via tablets or capsules). Paracetamol plus tramadol was the most frequent combination assessed (ten trials [31, 32, 39,40,41,42, 45, 48,49,50]), with only one study investigating an extended-release version of the combination [32]. These trials included both low back pain and osteoarthritis populations.

Studies assessing paracetamol plus another opioid analgesic included hydrocodone [34,35,36,37], or oxycodone [33], all compared to placebo, and in chronic low back pain. Paracetamol combinations with an NSAID included aceclofenac [47], etodolac [29, 46] or ibuprofen [30, 38], with only one of these studies compared to a placebo [30]. Other paracetamol combinations included a muscle relaxant of orphenadrine [43], and one study combining paracetamol with a chlorpheniramine, aspirin and caffeine preparation [44]. Monotherapy comparators were noted to only be assessed in studies with participants having acute presentations [29, 38, 40, 43, 44, 46, 47]. Paracetamol combinations are summarised in Table 2.

All studies reported pain outcomes, 18 studies used the visual analogue scale or numerical pain rating scale, three studies use categorical measures [30,