Some surgeons use wound glue. Some use staples. Others use a subcuticular closure and strips, or some other kind of suture technique. Barbed stitches, monofilament absorbable materials, nonabsorbable materials; the list goes on. We have at our disposal a vast array of dissimilar approaches to skin closure, and they all seem to do okay.
If almost all of our surgical incisions heal uneventfully, as seems to be the case, why would a meta-analysis on the topic justify publication, much less a feature in the Spotlight here at Clinical Orthopaedics and Related Research®? If there were important differences in infection between one wound-closure technique and another, wouldn’t we already know it?
It turns out, probably not. Three relatively recent meta-analyses on exactly the same topic [5, 6, 12] disagreed as much as three such studies can: One concluded in favor of sutures, the second in favor of staples, and the third was a statistical tie. And as interested as many of us are in knotless, barbed sutures, the work on these is just too preliminary to talk about; the most-recent meta-analysis I found on that topic in orthopaedic patients included only about 600 patients from five studies , not nearly enough to evaluate the kinds of endpoints that matter, like serious infections, which occur less than 1% of the time. But as I have discussed before in this space, the absence of serious complications is not the same thing as safety . An earlier meta-analysis raised concerns about superficial infections with these new suture materials , and large, single-center analysis published in CORR® associated their use with an increased risk of arthrotomy dehiscences . Based on all this, it seems clear enough that the jury is still out in terms of the best way to close surgical incisions.
In this month’s issue of CORR, we are fortunate to publish the highest-quality meta-analysis I’ve seen on this topic . In it, Janet Martin BSc(Phm), PharmD, MSc(HTA&M) and her colleagues at the University of Western Ontario report that staples were associated with a more-than twofold increase in the relative risk of surgical-site infection (SSI) compared to sutures. Their study included a large and diverse assortment of randomized trials involving arthroplasty, trauma, and general-orthopaedic procedures. As important as their finding on SSI seems, the increased SSI risk was not even the most-concerning finding in this analysis of nearly 2500 patients’ operations.
The finding that troubled me most in their study was that a subanalysis removing the lower-quality randomized trials found the conclusion about SSIs to be fragile, meaning that only a few more infections in one group or the other could have resulted in the effect’s direction changing entirely. And related to that, I’m chagrined by the fact that we still don’t have answers about deep infections. Again—and this is really important—the fact that we didn’t get the answer in a meta-analysis of 17 trials and more than 2000 patients doesn’t mean there isn’t an important answer to be found. To speak convincingly to differences in risk of deep infection would call for a meta-analysis of tens of thousands of patients. But if that is the case, then how much can this really matter?
It may matter tremendously. More than 6 million patients will have orthopaedic surgery next year in the United States alone , and every one of them will end with the closure of an incision. Even small between-treatment differences in how we perform those closures would change the lives of tens of thousands of patients. Those differences absolutely are worth looking for, and the fact that we have not seen them does not mean they haven’t seen us.
Join me as I discuss this serious issue with a serious meta-analyst, Dr. Martin, senior author of “Is the Risk of Infection Lower with Sutures than with Staples for Skin Closure After Orthopaedic Surgery? A Meta-analysis of Randomized Trials,” in the Take 5 interview that follows.
Take 5 Interview with Janet Martin BSc(Phm), PharmD, MSc(HTA&M) senior author of “Is the Risk of Infection Lower with Sutures than with Staples for Skin Closure After Orthopaedic Surgery? A Meta-analysis of Randomized Trials”
Seth S. Leopold MD:Congratulations on this thoughtfully conducted meta-analysis. I think surgeons—particularly arthroplasty and spine surgeons—may worry more about deep infections than they do SSIs, many of which are easy to treat and do not seem to have lasting sequelae. Why should surgeons pay attention to a study whose endpoint focused on SSIs, which represents (as you wrote) a pooling of “minor events with rarer, more severe events”?
Janet Martin BSc(Phm), PharmD, MSc(HTA&M): Quite rightfully, surgeons, patients, and healthcare managers will worry more about deep infections and PJIs more than they do about superficial SSIs. However, in an era of antibiotic resistance and overuse, it is also important not to trivialize superficial SSIs. Most clinical trials and meta-analyses of interventions to reduce SSI report on all infections as a composite endpoint (in compliance with CDC recommendations for reporting SSI ) and assume that any beneficial impact on the composite of SSI may be more likely to indicate potential impact on the subcomponent of serious infections. In any case, it is difficult to power randomized controlled trials to address serious infections as a primary outcome, given their rarity in orthopedic surgery. Because of the importance of this question, I asked my coauthor Dr. Rampersaud to contribute here, and he said: “Although superficial wound infections in isolation are not as serious as deep infections, they still have the potential to have detrimental effects for both the patient and the healthcare system. The most obvious is that they can turn into deep infections. However, even isolated superficial (that is incisional) infections can increase patient morbidity and may lead to delays in rehabilitation and achievement of functional outcomes. They also have the potential to lead to increased healthcare visits, and prolonged time away from work. Additionally, from a health systems perspective, superficial infections are likely to increase the number of physician visits, diagnostic tests, and even readmission rates to hospital. These events have substantial direct healthcare costs associated with them. As many healthcare systems move to performance-based funding models, superficial infections may ultimately decrease healthcare efficiency.”
Dr. Leopold:It also seems possible that surgeons and healthcare systems might favor staples or knotless barbed sutures (which may not heal as well as conventional suture[1, 11]) for reasons of efficiency and cost, while patients—if offered that same choice—might make exactly the opposite choice out of concerns for infection based on a study like yours. How should surgeons make a choice in this setting that is ethical, just, and evidence-driven?
Dr. Martin: Effective decision-making is an art that is informed byscience. Best-possible evidence from clinical trials, or preferably from meta-analyses, should serve as the foundation for decisions. But, of course, evidence from clinical trials is not the only thing that needs to be incorporated into decisions. Other contextually relevant issues should inform decisions commensurate with their importance and potential impact. In this case, the evidence does not convincingly prove that sutures outperform staples, though there is some indication in the direction of less SSI with sutures (albeit, driven mostly by superficial events), which may be clarified in future trials. In other words, there is uncertainty about which is best, from a clinical perspective. When the evidence fails to demonstrate clinically meaningful differences between treatment approaches, patient preferences and physician preferences become even more important to incorporate into the decision. Patients may prefer sutures or staples, based on other outcomes not yet reported the clinical trials, such as rates of wound healing, short-term discomfort under dressings or clothing, requirement for removal or ease of removal, and longer-term scar patterns, or cosmetic results. Surgeons may prefer one type of closure over the other for reasons of familiarity, availability, ease of application, and overall efficiency. Healthcare managers and administrators may simply prefer the closure device with the lowest acquisition cost. Given the potential for opposing preferences between patients, surgeons, and healthcare managers, this is an excellent open question for further research on preferences and is especially relevant for a cost-effectiveness analysis or discrete choice experiment taken from various perspectives (patient perspective, hospital perspective, health system perspective). In the meantime, shared decision-making involving patients and providers may the best interim solution.
Dr. Leopold:How far are we from getting the answer we most need, which is the difference in the risk of serious infection (PJI and/or deep SSI resulting in reoperation), and what will it take—that is, exactly how large a meta-analysis—to get us there?
Dr. Martin: This is an excellent question, though at first blush, the answer may seem somewhat dispiriting. If we assume that deep infection or PJI requiring reoperation occurs in 0.5% of patients, we would need about 20,000 additional patients randomized to staples versus sutures to detect a 50% reduction in risk (that is, a reduction from 0.5% down to 0.25%). If we assume that serious infections occur in 1% of patients, we would need about 11,000 additional patients randomized to staples versus sutures to detect a 50% reduction in risk. Both scenarios reflect realistic baseline rates. Of course, this number could vary considerably based on assumptions we make about baseline likelihood of serious infection and would be influenced heavily by attrition (loss to followup), incomplete data, and patient withdrawals. This large sample-size requirement should really be cause for celebration since it is evidence for the progressive successes of modern-day surgery in high-resource countries, where serious PJIs have declined precipitously. It is interesting that even if we could prove an absolute difference in serious infections as hypothesized in the above sample-size-calculation scenarios, the number needed to treat in order to prevent serious infection in one patient would likely be about 200 to 400. Clearly, in an environment of very low rates of serious infections, we are quibbling over differences that are very small, if they exist at all. However, in settings where baseline rates of serious infections are higher than 1%, such as in resource-restricted settings where the risk of infection is much higher, the impact of type of skin closure on serious infections might reasonably be considered an important research priority.
Dr. Leopold:At CORR, we enjoy publishing meta-analyses, network meta-analyses[8, 9], and decision analyses, as we believe they can answer questions that no other study design can answer. I’m often surprised, though, by how negatively some readers react to them; it’s not merely a dislike of these study designs, it’s almost a distrust. How would you try to persuade skeptical clinicians of the value of studies that use only pencils and computers, rather than scalpels and patients?
Dr. Martin: I can assure you, no pencils were harmed during this study! Meta-analysis sometimes is feared, or outright rejected, by those who make blunt assumptions about “garbage in, garbage out”, or who assume that apples and oranges were combined to produce nonsensical results. While both of these would be good reasons to reject a meta-analysis, it is less common for meta-analyses to reach publication stage in better journals if they have nonsensically synthesized results across studies that should not be combined, or if they have overstated the results based on the quality of the underlying studies.
That said, I don’t think of meta-analysis as research involving only pencils and computers, since real patients were randomized into the trials that form the basis of the meta-analysis. Contrary to what most people assume, meta-analysis is not a blunt averaging of effects across studies. If we fail to appropriately meta-analyze all studies that have addressed the same research question, then we will lose the opportunity to gain insights available to us from those studies. The most-obvious reason to combine studies through meta-analysis is to gain power to detect differences, since individual trials often are underpowered.
But there are other reasons to meta-analyze, and they may be even more valuable. Meta-analysis allows us to place all studies reporting similar types of outcomes on the same forest plot. Even without statistically synthesizing the results across the studies, there is value in reviewing the forest plot to understand how many studies have been done, the range in sample sizes of the studies, how many have reported on the outcome(s) of interest, the number of studies that showed significant difference for each of the outcomes of interest, the heterogeneity across studies in the size of effect, and the range in event rates for the intervention and control groups. There is also opportunity to subgroup the studies according to factors that are predicted to be important (such as low-middle-income versus high-income settings; or emergency versus elective settings), to understand whether effect sizes differ by subgroup. There is tremendous power in the forest plot for understanding the state of the evidence on a topic. Without meta-analyses, clinical decision-making becomes an exercise in cherry-picking which studies to trust, often based on a biased preference for studies with conclusions that agree with one’s preconceived notions.
Dr. Leopold:I’m sometimes surprised that we don’t see more decision analyses performed to followup on meta-analyses where the main findings resulted in close calls, as occurred in your paper. It seems to me this might be a good way to honor our patients’ preferences and values. Why might this be, and, more importantly, how might we encourage more meta-analysts to take up this charge?
Dr. Martin: I fully agree that decision analyses and decision modeling remain underrepresented in clinical journals. They are more commonly embedded within formal cost-effectiveness or cost-utility analyses, and published in journals devoted to health technology assessment, health policy, or health economics. The basic components of a decision analysis include reliable probabilities that a given outcome will occur for staples and for sutures (along with ranges of uncertainty) and valuations of the different possible outcome states (that is, reliable utility values derived from published research on orthopedic patients, or derived from our own patients’ valuations of superficial, deep, or PJIs). In order to perform decision analysis as a followup to our meta-analysis, we would use the meta-analysis-derived probabilities of SSI for staples and for sutures to inform the short-term outcomes.
However, some gaps still would need to be filled by making key assumptions or by incorporating indirect evidence. For example: What is the likely base case rate of deep infection and PJI? What is the probability that superficial SSI will predispose to PJI, and to what degree are these dependent events? Do patients’ preferences for staples or sutures affect utilities? Given the short-term probabilities of SSI, how should we extrapolate probabilities over the course of a lifetime? What is the clinical impact of SSI on health utilities over the short- and long-term? Once these are estimated, the likely health utility from using staples versus sutures could be calculated. In short, while decision analyses may sometimes help to clarify a decision by modelling beyond available data across plausible ranges of assumptions, the devil is always in the details. Coherent and reliable decision analyses require more than may initially be evident.