Intra-operative vancomycin: to randomize or not

Today we discussed a recent paper published by our orthopedic surgeons on using powdered vancomycin in the wounds of spinal surgery to prevent surgical site infections (SSI). Two years ago I already had a post on the topic. The powder is spread deep in the wounds, on the bone and metal, before fascia and skin are closed. Yet, none of the 3 guidelines addressing prevention of surgical site infections that appeared recently recommends this intervention; WHO (2016) didn’t even include the intervention, CDC (2017) said “don’t do it”, and NICE (2019) acknowledged that the procedure is widely used without strong supporting evidence and recommended: stop doing it and do a trial.

So, now there is new evidence. In short: To determine the efficacy of intrawound vancomycin (1-2 grams of powder) in reducing the incidence of deep- and superficial SSIs in instrumented spinal surgery they performed a retrospective cohort study, including all consecutive adult patients undergoing open, posterior, instrumented spinal surgery at any level. SSI was based on the CDC criteria and chart review. The first period served as control (N=257), in which 25 (9.7%) patients developed a deep and 13 (5.1%) developed a superficial SSI. In the second period all patients received vancomycin (N=379), and then 19 (5.0%) patients developed a deep and six (1.6%) a superficial SSI. Both deep (RR: 0.52, 95% CI 0.29−0.92) and superficial SSIs (RR: 0.31, 95% CI 0.12−0.81) were significantly reduced, also in multivariable logistic regression analysis including risk factors for SSI.

Will this study change the guideline recommendations?

Henri van Werkhoven presented the 4 risks of bias in a before-after study, being (1) regression to the mean; (2) secular trends; (3) observer bias; and (4) confirmation or publication bias. With the study described I was especially worried about the observer bias. The CDC criteria can be interpreted subjectively and – to me – it seems rather difficult to determine a primary endpoint with a retrospective chart review (knowing that orthopedic surgeons are not famous for their extensive and detailed medical chart prose).

Another potential bias is the effect of time; regression to the mean may occur (such studies usually start after recognition of a problem), but also gradually improving surgical skills, or temporary absence of that one surgeon famous for SSI or …. many other things may occur.

We concluded that there might be more information in the crude data, especially on the time trends, that now will be explored. If anything, inclusion of time as a variable in the multivariable logistic regression analysis will increase uncertainty around the effect estimate (and may well push the upper boundary of the confidence interval above 1).

In the end, this study creates an ethical dilemma. Our orthopedic surgeons are convinced that this intervention is good for their patients (and that randomizing them to not receiving vancomycin would be unethical). We, the purists from epidemiology, tell them that that belief is based on very low quality evidence, which will not change guideline recommendations. So, continuation of its use may benefit our patients (epi says not sure), but will not benefit other patients treated in centers where they don’t throw vancomycin in wounds. A single well-designed RCT with an objective patient-centered primary outcome  (for instance a combination of hospital stay and physical performance) confirming the effects of the before-after study will immediately be the only available high-level evidence, will change guideline recommendations and will benefit many more patients.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s