The AAOS is committed to developing treatment guidelines to improve patient care and to identify gaps in our clinical knowledge base. As Chair of the rotator cuff guidelines work group, I was honored to work with a knowledgeable and experienced group of surgeons and researchers in this endeavor. Given that experience, I would like to provide my personal perspective on the guidelines process to facilitate proper utilization by our key stakeholders: patients, providers, and payors. This commentary reflects my personal opinions. It is not intended to represent the opinions of the AAOS or other members of the guidelines committee.
Orthopaedic surgeons strive to provide the best possible care, with minimal harm. We want to maximize the cost-effectiveness of health-care expenditures, but our foremost duty is to individual patients. Given this framework, the fundamental objectives of evidence-based medicine (EBM) are laudable. However, interpretation and application of evidence-based guidelines will surely vary, according to the lens through which one views research data and according to one's understanding of the guidelines process itself.
Consider this simple hypothetical statement: There is no evidence that Intervention X has greater benefit than Intervention Y. To surgeons, this statement might indicate that it is appropriate to offer our patient either procedure. To payors, the same statement might infer coverage denial for both interventions, pending better evidence. Patients might take this equivalency as confusion, leading to a lack of trust in the surgeon. To me, the statement simply means that there is not sufficient proof, according to EBM rules, to make a declarative statement. Nothing more.
Solid understanding of research methodology is critical for appreciation of guidelines implications and limitations. There are very specific study inclusion and exclusion criteria. For example, the guidelines process excludes basic science research and excludes most studies that are based on surrogate outcomes, such as clinical imaging. Deference is given to patient-based clinical outcomes. Does that mean that other valid information is irrelevant or unimportant? Of course not. Interested stakeholders should examine the conditions and prerequisites of the EBM process and should balance associated findings against other scientific and clinical information.
The practice guidelines algorithm is highly structured, and the process continues to evolve. The terminology used to wordsmith guidelines is pre-scripted (to give some inference about the associated level of evidence for each recommendation), which makes some statements awkward, noncommittal, and bland. All questions posed at the beginning of the process must be addressed by a written guideline at the end (even when no evidence exists), tending to dilute the impact of the resultant document. Expert opinion is discouraged.
The absence of Level-I or II evidence does not prove that an intervention does not work. Aggregate and compelling Level-III, IV, and V evidence can (and should) sway clinical routines. However, according to the guidelines process, higher-level evidence trumps lower-level evidence. Put another way, several Level-I studies outweigh a thousand Level-IV studies. Unfortunately, Level-I research designs and study populations are not automatically relevant to daily practice, and subjective inclusion criteria can substantially affect research outcomes1. Currently, practice guidelines are not sufficiently sophisticated to replace the critical elements of expert surgical management: knowledge, judgment, experience, technique, and ethics. In my opinion, integration of available information, including sage advice from thoughtful experts and one's personal experience, is still the best way to define optimum treatment options for an individual patient. Eventually, we may find that EBM and clinical practice guidelines are most valuable for optimizing the process of informed consent and not for improvement of health per se2.
An important objective of the clinical guidelines process is identification of gaps in our knowledge base, which can stimulate high-quality outcomes research. But an obvious corollary question follows: Should we strive to answer all clinical questions with randomized controlled trials (RCTs)? This question touches fundamental medical ethics, highlighting the friction between duty to individual patients and population-based objectives of maximizing health-care resources.
The question stimulates examination of the concept of clinical equipoise. For a surgeon-researcher to enroll patients in an RCT, it is widely accepted that a state of “clinical equipoise” must exist3. Clinical equipoise describes a state where “there is no consensus within the expert clinical community about the comparative merits of the alternatives to be tested” with “honest professional disagreement among…clinicians.”4 In the context of equipoise, let us consider Guideline 10.a from the clinical practice guidelines: “It is an option for surgeons to attempt to achieve tendon to bone healing of the cuff in all patients undergoing rotator cuff repair.” The guideline used the term “option” because the level of evidence was weak. But would anyone endorse an RCT for this question? I think not. There is uniform acceptance that surgeons should try to fix the tendon to bone, realizing that it does not always heal if surgery is done for a repairable cuff tear. Why would we not try? A state of “clinical equipoise” simply does not exist for this issue, and therefore an RCT will never be done.
For an RCT to be ethical, each treatment arm should minimize the likelihood of a reversal of the risk-benefit ratio5, in other words: primum non nocere. This principle evokes concerns about the design and implementation of an RCT of chronic, symptomatic rotator cuff tears, which was the topic of Guideline 2 (again, weak evidence). There are two ethical considerations. First, there is a large body of published information and broad clinical experience that demonstrates strong likelihood of decreased pain and improved function following surgery for symptomatic rotator cuff tears for which nonsurgical treatment has failed. Is it ethical to randomize patients to prolonged periods of persistent suffering in order to compare outcomes with a surgically treated group? Second, basic science and clinical literature demonstrate that prolonged nonsurgical treatment is associated with progressive tendon retraction and adverse changes in muscle architecture, which can lead to an irreparable rotator cuff tear. Is it ethical to randomize patients to known and potentially permanent morbidity?
Certainly there are rotator cuff topics that are highly appropriate for randomized trials. Examples include arthroscopic versus open cuff repair, single-row versus double-row cuff repair, and suture anchors versus bone tunnels for cuff repair. Such questions are associated with clinical equipoise (opinions vary across surgeons), and the risk:benefit ratios of the treatment alternatives are ethically acceptable. Still, on the back end, such studies must be evaluated carefully to ensure that the patient populations, surgical methods, and requisite surgical skills are relevant to the general orthopaedic population.
In summary, stakeholders must be careful about the use, and the potential abuse, of clinical practice guidelines. These are serious and laudable efforts, but like all research, there are substantial limitations. A guideline does not replace expert surgical judgment. Lack of Level-I evidence does not mean that treatments are ineffective, irrational, or unsafe. It simply means that an RCT has not been published. For some clinical questions, Level-I investigation is not the best answer.