Summary of main results

This review included 61 studies comparing a topical NSAID with placebo, another topical NSAID, or an oral NSAID. In total, 5311 participants were treated with a topical NSAID, 3470 with placebo, and 220 with an oral NSAID. There were 63% more participants than in the previous version of this review. Conditions treated were sprains, strains, and contusions, mainly resulting from sports injuries, and overuse injuries such as tendinitis.

Formulations of topical diclofenac, ibuprofen, ketoprofen, piroxicam, and indomethacin demonstrated significantly higher rates of clinical success than matching topical placebo lacking the NSAID; benzydamine did not. Three drug and formulation combinations had NNTs for clinical success below 4. For diclofenac, Emulgel® had the lowest NNT of 1.8 (1.5 to 2.1) in two studies using at least 50% pain intensity reduction as the outcome (high quality evidence). Diclofenac plasters other than Flector® also had a low NNT of 3.2 (2.6 to 4.2) based on good or excellent responses in relatively recent studies (high quality evidence). Ketoprofen gel had an NNT of 2.5 (2.0 to 3.4) from five studies in the 1980s, some with less well defined outcomes (moderate quality evidence). Ibuprofen gel had an NNT of 3.9 (2.7 to 6.7) from two studies with outcomes of marked improvement or complete remission (moderate quality evidence). All other drug and formulation combinations had NNT values above 4, indicating lesser efficacy.

These results are better than alternative topical products that might be used for acute musculoskeletal pain. There is no evidence to support the use of topical salicylate rubefacients (Derry 2014).

Treatment with a topical NSAID was not associated with an increase in local adverse events (skin reactions) compared with placebo (inert carrier), or in withdrawals due to adverse events (high quality evidence). The inert carrier was sometimes associated with mild skin irritation, but this rarely led to cessation of treatment, and quickly resolved. Systemic adverse events were uncommon and did not differ between topical NSAID and placebo (high quality evidence). Two participants experienced serious adverse events with diclofenac plaster and diclofenac gel, but it is unlikely that these were related to the study medications.

There were insufficient data directly comparing a topical NSAID with the same oral NSAID to draw conclusions about efficacy. Based on very limited data for oral NSAIDs, there were fewer systemic adverse events with topical than oral treatment. There were sufficient data only for topical piroxicam compared with topical indomethacin to compare one topical agent with another. These limited data suggested that piroxicam was more effective than indomethacin, and was less likely to cause local adverse events. It is worth noting here that topical indomethacin did not give significantly better pain relief than placebo in two of the three studies in this analysis.

Overall completeness and applicability of evidence

There is a tension between pooling studies to produce analyses with larger numbers and the subsequent large increases in clinical and statistical heterogeneity on the one hand, and using the approach of clinical homogeneity with subsequent smaller numbers of participants on the other hand. Previous reviews have taken the former approach; that is useful in demonstrating that topical NSAIDs 'work' by being significantly better than placebo. Because of the substantial increase in the amount of data available, in this review we have chosen to seek greater clinical homogeneity; this produces results that are more relevant to patient and prescriber choice.

There were too few studies comparing one topical NSAID versus another, or versus the same oral NSAID, to allow meaningful direct comparisons between individual drugs or routes of administration.

The conditions treated in these studies are representative of those likely to be suitable for acute treatment with topical NSAIDs. The mean age of participants in individual studies ranged from 25 to 57 years, and the nature of recruitment in many studies meant that participants were actively engaged in sporting activities. Nevertheless, older people in their 60s to 80s were also included in some studies, and the low levels of predominantly mild adverse events means that this route of administration of NSAIDs is suitable for all age groups able to manage the application process.

Information from other sources, mainly randomised studies lasting 12 weeks or more in older populations with arthritis, tend to confirm this. A systematic review of topical NSAIDs in older adults was difficult to interpret, but suggested that the range of withdrawal rates in these studies was similar with topical and oral NSAIDs (Makris 2010). It also claimed potentiation of warfarin effects, but that was with topical salicylate, not an NSAID. In contrast, a pooled safety analysis of topical diclofenac in people aged 75 years or older reported minimal changes, with a mean reduction in haemoglobin of less than 1 g/L with topical diclofenac, and a mean systolic blood pressure reduction of almost 4 mm Hg (Roth 2012). Two large randomised 12‐week studies comparing topical with oral diclofenac in arthritis reported lower rates of gastrointestinal adverse events with topical than oral, especially severe events, but larger reductions in haemoglobin with oral diclofenac (Simon 2009; Tugwell 2004).

The available evidence was limited by numbers to comment on rare but potentially serious adverse events. One example is the potential for photo‐sensitivity reactions with topical ketoprofen. Current advice from the Medicines and Healthcare products Regulatory Agency in the UK is to avoid direct sunlight, ultraviolet (UV) rays, sunlamps, and sunbeds while using topical ketoprofen, and to see a healthcare professional or go to hospital if they experience a skin reaction to sunlight, sunlamps, or sunbeds (MHRA 2009).

Quality of the evidence

All included studies were both randomised and double‐blind; none was considered at high risk of methodological bias. Many were carried out in the 1980s and 1990s when methodological rigor and detailed reporting were not given such high priority and studies did not always report details of the randomisation, treatment allocation, and blinding processes. More recent studies often did report methodological details, and tended to be larger (see Figure 4 for a comparison of quality of reporting for different dates and NSAIDs). Our primary outcome of clinical success was not always well‐defined, and was measured using different scales, but again more recent studies tended to report outcomes better.

The studies were conducted in different conditions, with somewhat different outcome definitions and duration, and with different topical NSAIDs and formulations. Moreover, the small size of many of the studies is likely to result in considerable chance variation (Counsell 1994; Moore 1998b). These factors would account for the high I² values seen in several analyses. Despite these sources of potential clinical heterogeneity, most studies showed benefit of topical NSAID over placebo.

The design of studies to be able to demonstrate analgesic sensitivity is important in self limiting conditions such as strains and sprains. Too long a duration and the condition results in spontaneous resolution of painful symptoms, while too short a duration may be inadequate to show any effect. The decision by trialists to concentrate on outcomes closest to seven days of treatment appears to be prudent, and has been adopted in this and previous reviews. There are potential differences in response to treatment between strains and sprains and overuse‐type injuries such as tendinitis, and future reviews may examine this. At the present time, there are too few existing trials to explore any differences adequately.

Baseline pain may be a cause for concern. Seven studies did not report baseline pain levels (Billigmann 1996; Curioni 1985; Haig 1986; NCT01255423; NCT01272934; NCT01272947; Sinniger 1981), and a further 11 reported either mean levels of less than moderate pain or a significant proportion of individuals with less than moderate pain (Ăkermark 1990; Aoki 1984; Auclair 1989; Diebshlag 1990; Fujimaki 1985; Jenoure 1997; Linde 1985; Picchio 1981; Ramesh 1983; Sugioka 1984; Whitefield 2002), using recognised scales. Insufficient pain at baseline compromises the ability of a study to demonstrate any improvement. All the newly added studies reported baseline pain to be of at least moderate intensity.

Potential biases in the review process

There has been greater interest in topical NSAIDs in recent years, mainly because lower systemic drug levels reduce the risk of troublesome and severe adverse events, particularly in the gastrointestinal tract, and renal and cardiovascular systems. Most of the attention has been in chronic conditions such as osteoarthritis, with few studies in acute painful conditions. Low levels of serious adverse events with topical NSAIDs has been noted previously (Evans 1995), and the near absence of serious adverse events in this review is unlikely to be due to any biases in the review process.

One potential bias is that clinical trials for topical NSAIDs may not have been published. One previous review did find previously unpublished trials (Moore 1998a), but a subsequent attempt that included extensive contacts with pharmaceutical companies revealed no additional data (Mason 2004a). While some old unpublished studies of topical NSAIDs in acute painful conditions may exist, they constitute an unknown number of studies and participants whose results are unknown, and are likely to remain unknown. Furthermore, their relevance to current clinical practice may be limited as better formulations are developed. New systems of trial registration mean that we know what recent studies have been done or are ongoing; the number of studies and participants is known even if their results remain unknown. We identified in Clinicaltrials.gov three unpublished studies (612 participants) with adverse event data but no dichotomous efficacy data, 14 completed unpublished studies (4403 participants) with no results posted, and three ongoing studies (880 participants).

For the main topical NSAIDs of interest and where most information exists, about 4200 participants in this review provided data on efficacy for diclofenac, ibuprofen, ketoprofen, indomethacin, and benzydamine compared with placebo. For efficacy, there are unknown results from almost 5900 participants in studies known to have been done but essentially unpublished. Almost 6500 participants in this review provided information on local adverse events for topical NSAIDs compared with placebo. For local adverse events, the unknown results from known studies represents almost 5300 participants. It is clear that identified unpublished but unavailable study data amounts to a further potentially large increase in knowledge, over and above the 60% increase in numbers of participants already included in this updated review.

Based on efficacy data on known and available study results, unpublished trials showing no difference between any topical NSAID and topical placebo and involving 5500 participants would have to exist in order for the NNT to be as high as 9, at which point the effectiveness of topical NSAIDs would become clinically irrelevant (Moore 2006). This amount of unpublished negative data is obviously available, and while a negative result in all the identified studies is unlikely, knowledge would be greatly served by having these unpublished trial results available.

We have not yet attempted to obtain results from these clinical trials from the trial sponsors, because this takes a considerable amount of time and may not be successful. Moreover, the studies were spilt between nine different sponsors: Cerimon Pharmaceuticals (five studies), Novartis (four studies), Endo Pharmaceuticals (three studies), GlaxoSmithKline (two studies), Hisamatsu Pharmaceutical (two studies), and one each from Actavis, Pfizer, Imprimis Pharmaceutical, and Strategic Science & Technologies.

Agreements and disagreements with other studies or reviews

A review published in 2004 included some of the studies in this review and reported an NNT for all topical NSAIDs combined of 3.8 (3.4 to 4.4) for clinical success equivalent to half pain relief at seven days (Mason 2004a). That review found no difference between topical NSAID and placebo for local adverse events, as did this review. In turn, the Mason review was in broad agreement with the original systematic review on topical NSAIDs (Moore 1998a). To our knowledge, no previous review assembled sufficient trial data to analyse results by both drug and formulation, as was done here.