Arthroscopic shoulder surgery no better than sham shoulder

Another article just came out in the Lancet finding a lack of benefit from a commonly done surgery vs sham surgery, this time for arthroscopic subacromial decompression for subacromial shoulder pain (see doi.org/10.1016/S0140-6736(17)32457-1).

Details:

--313 patients from 32 hospitals in the UK, with 51 surgeons. From 2012-15.

--mean age 53, 50% female, baseline Oxford Shoulder Score (OSS) was 29 (scale of 0-48, with 0 being worst; this scale basically sums patient symptom scores over the prior 4 weeks, including their ability to get dressed, carry things, other activities of daily living; plus other scores for anxiety/depression, pain and quality of life. At baseline, all were mostly midrange, though depression/anxiety was in the lower range (less bad)

--all patients had subacromial pain for at least 3 months, had intact rotator cuff tendons, and had completed a non-operative management program that included exercise therapy and at least one steroid injection. Diagnosis was confirmed by a consultant shoulder surgeon

--patients were randomized to arthroscopic subacromial decompression (removing bone spurs and soft tissue) vs arthroscopy only (considered the "sham surgery", and involved only irrigation and inspection of the glenohumeral joint and subacromial bursa) vs just an appointment with a shoulder specialist 3 months after study entry but no intervention (including no further PT or injections).

--primary outcome was the OSS, analyzed by intention-to-treat. Follow-up up at 6 months and 1 year. [my searching around did not reveal either what the specific OSS numbers meant in terms of disability, though I assume that a score of 29 at baseline would represent moderate disability, nor could i find any validated indication of what a minimal clinically-relevant change would be]

Results:

--though they are looking at the 6-month and 12-month progress, this is 6 months after randomization and the median surgery time was 3 months after randomization.  So, the results below really indicate an average of 3 months and 9 months after surgery [and they are lumping together patients who had surgery up to 5-6 months before with some only a couple of months before the 6-month target, perhaps affecting the utility of their results]

--all patients in all groups got progressively better over time: at both the 6-month and 12-month visits, and this improvement was equivalent (all were parallel straight lines)

--at 6 months (primary outcome):

            --no difference in OSS in the 2 surgical groups (32.7 vs 34.2 for decompression vs arthroscopy)

            --both surgical groups together did statistically but not clinically better than the nonsurgical group (benefit only 2.8 points for decompression ( p=0.0186) and 4.2 points vs arthroscopy alone (p=0.0014) [their comment that this was not clinically important]

--at 12 months (secondary outcome):

            --no difference in OSS between the 2 surgical groups (38.2 vs 38.4 for decompression vs arthroscopy)

            --both surgical groups did statistically but not clinically better than the nonsurgical group (benefit only 3.9 points for decompression ( p=0.0193) and 4.2 points vs arthroscopy alone (p=0.0193)

--but 23% of those in the decompression group vs 42% with arthroscopy only vs 12% of no surgery groups did not receive their assigned treatments at 6 month follow-up. Their analysis, however, did not show anything different from the above findings if using a per-protocol analysis (ie, looking at the patients who really had the different surgeries or medical management), or after multiple imputations to mathematically normalize the different cohorts

--no difference in adverse events (2 people in each group had frozen shoulders)

Commentary:

--the relevance here is very high: 4.5 million clinical visits annually in the US for subacromial pain, which accounts for 70% of all shoulder pain. And there is a clear anatomical cause (which would seem to justify surgery): mechanical contact between the rotator cuff tendons and the overlying acromion or bone spur, typically at the anteroinferior margin of the acromion, along with the generalized weakening of the rotator cuff with age. And this narrowed subacromial space leads to impingement of the rotator cuff, inflammation/swelling and further impingement, and symptoms.  Most patients do well with medical/PT management, but surgery is typically done in recalcitrant cases.

--This blog follows closely on the heels of one showing the lack of benefit of stenting for angina (see http://gmodestmedblogs.blogspot.com/2017/11/pci-in-angina-stenting.html ), a blog which also comments on recent studies finding no benefit for renal denervation surgery for resistant hypertension and arthroscopy for knee osteoarthritis. This latter example is in many ways very similar to this one: good mechanistic reason to do arthroscopy, highly accepted and widely utilized procedure, but 2 studies found that sham surgery was as good. [though the persistent concern is that even after these negative studies, arthroscopic surgery is still done pretty regularly]. see prior blog: http://gmodestmedblogs.blogspot.com/2015/06/arthroscopic-surgery-for-knee-oa-not-so.html for review of one of the sham surgery studies.

--these types of studies would be difficult to do in the US, given the more interventional ideology here (acceptance and expectation of medical/surgical interventions by both clinicians and patients) and the lack of a coherent health care system which is evaluating the true benefits of such interventions (and, drug and device companies are funding the vast majority of studies these days and are less interested in testing for potentially negative outcomes in a very profitable market)

--why did the surgical groups do somewhat better (but not clinically significantly so)?: just the superior placebo effect of surgery over none??? because of the postoperative physiotherapy the surgery groups received, since all patients had received exercise programs prior to the randomization but the 2 surgical groups had post-op PT, with nothing more for the medical group (though this latter point might make non-operative treatment less beneficial/skew treatment effects toward the surgery groups, given less of a medical intervention in the non-operative group)?? Would repeated shoulder injections have helped more than surgery??

--they chose to do arthroscopy as a surrogate for sham surgery, since just sham surgery might "pose recruitment issues". Would true sham surgery (superficial incisions only) have led to the same results?

--this article suggests that in the aggregate, arthroscopic therapy is not particularly useful.  But, there may be some patients who need it/get really important benefit.  The issue is that at this point there is no clear method to risk-stratify patients either based on symptoms (the patients in this study had continued pain for at least 3 months, even with PT and injection. is there some clinical indicator of benefit from surgery?), or based on radiologic findings (is there some radiologic finding that warrants surgical intervention?).

So, what do we do now since it is pretty clear from the above that surgical intervention as indicated now does not significantly improve patient symptoms in the aggregate??  Who knows??, but we will still see individual patients with debilitating symptoms…

--I will personally discourage patients from getting arthroscopic surgery, and really more aggressively encourage them to pursue non-operative management of their shoulder pain, relaying the data that overall patients tend to get better over time (pretty clear in this study).  And, many of my patients do get substantial and long-lasting relief from shoulder injections, which I will continue to pursue.  these are really easy to do, and the vast majority of patients do get good relief of symptoms with improved functionality, though i recognize that prior studies on injections have had mixed outcomes, some finding benefit and others not. but these injections are still commonly done, and were part of the inclusion criteria for this study....