Skip navigation

Shoulders to the Wheel

Maastricht rules again

Those prolific Dutch systematic reviewers have been at it again. This time they have turned their attention to steroid injections for shoulder disorders [1].

Shoulder problems in primary care

They say that 10% of people have one or more episodes of shoulder pain and/or stiffness during their life, and that 5% of all primary care consultations are about shoulder problems. Of new consultations, they estimate that 23% resolve within a month, 51% within 6 months and 59% within a year.

The figures for usage of injections to treat these shoulder problems seem very high. 'Twelve per cent of all patient-physician contacts for shoulder disorders involve local steroid injections' and 'In the Netherlands, injection therapy is given in 20% of all episodes of shoulder disorders'. Bandolier has a sneaky suspicion that these high proportions reflect payment for item of service.

The way in which steroids work to mend the joint has never been clear to Bandolier . The biology is a bit akin to the label on a can of WD-40 - stops squeaks, drives out moisture, cleans and protects, loosens rusted parts and frees sticky mechanisms. Steroid injections can, however, cause problems, such as the direct hazards of injection, bleed or infection, and effects of the steroid, dermal atrophy and long-term deleterious effects on the joint and surrounding structures. These problems have led to recommendations that steroids should not be injected more than once every six weeks, and not more than three times a year. We should therefore have clear evidence of efficacy if we are going to do these injections.

22 studies

The reviewers found 22 studies which met their inclusion criteria. They put these studies through their tough quality scoring system, and no study scored more than 60 out of the maximum of 100, with only three scoring more than 50. This is a clear signal that definitive conclusions are unlikely to be possible from the available studies.

The second complexity is that the studies looked at many different treatments, not just at injections. Only three studies compared steroid injection with saline injection, and five compared steroid injection with injection of local anaesthetic. The studies also used different outcome measures.


Not surprisingly then the reviewers decided that discretion was the better part of valour, and did not pool the data from the studies. They did not mince their words in their conclusions - 'the evidence in favour of the efficacy of steroid injections for shoulder disorders is scarce. ... The few studies that appear to be credible do not provide conclusive evidence about which patients at what time in the course of shoulder disorders benefit most from steroid injections.'

Onward, ever onward

As ever brave on your behalf Bandolier did probe the results a little further. Taking the crude criterion of success at four weeks or later after injection, the NNT for such success with steroid injection compared with saline injection (three studies) was 17, with a confidence limit which includes no benefit to any patient. For steroid plus local anaesthetic injection versus local anaesthetic alone (five studies), the NNT for success at four weeks or later was 33, again with a confidence limit which includes no benefit to any patient. On this basis one patient in 17 would achieve 'success' with a steroid injection compared with an injection of saline, and one patient in 33 would achieve 'success' with a steroid plus local anaesthetic injection compared with an injection of local anaesthetic alone.

Where does this leave us?

Our Dutch colleagues' conclusions seem fair. Those (and there is one on the Bandolier team), who might be pejoratively described as hawks because of their needle-waving tendencies, have a problem. The evidence that steroid injections for shoulder problems are worthwhile is less than compelling. The onus is on those who wish to continue to offer steroid injections to the shoulder to produce convincing evidence - a starting point would be to use a study design which scored closer to the maximum on the Dutch scale.

Chasing the evidence for our favourite interventions does sometimes result in a negative like this. The reaction to the review can become a hostile reaction to the reviewers - 'Very threatening' - 'if they go on like this I'll be out of a job'. Surely the professional agenda should be to provide a package of care which includes interventions which do more good than harm. Don't blame the messenger.

A last chastening point is that the fact that 40% of the patients still had their shoulder problem after a year means that a considerable proportion of these (common) shoulder problems are not self-limited. We need good control of these symptoms if we (or time) cannot cure them.


  1. CJM van der Heijden, DAW van der Windt, J Kleijnen, BW Koes, LM Bouter. Steroid injections for shoulder disorders: a systematic review of randomized clinical trials. British Journal of General Practice 1996 46: 309-16.

previous or next story in this issue