Clinical Prediction Rules: Time to Sacrifice the Holy Cow of Specificity?

For those of us practicing orthopaedic manual physical therapy (OMPT)—and, in fact, for all evidence-based manual medicine practitioners—it truly is an exciting time. Only a few years ago, systematic reviews and meta-analyses were indicating a lack of evidence to support the use of manipulation for patients with acute and chronic low back pain1; similarly they noted no or, at best, inconclusive evidence for manipulative interventions if not combined with active exercise in patients with mechanical neck pain2–4. Now in short succession research has provided us with a number of clinical prediction rules to guide our OMPT diagnosis and manipulative intervention in patients with mechanical back and neck pain5–9.

Why are these clinical prediction rules so different from the earlier negative or inconclusive meta-analyses and systematic reviews? Clinical prediction rules are decision-making tools that contain predictor variables obtained from patient history, examination, and simple diagnostic tests; they can assist in making a diagnosis, establishing prognosis, or determining appropriate management10. As Childs and Flynn11 pointed out, if studies included in a systematic review or meta-analysis use no patient classification other than the broad category of non-specific low back (or neck) pain, the resultant heterogenous study samples pretty much preclude finding real effects of any specific intervention. In contrast, the recent clinical prediction rules all aim to identify a more homogenous diagnostic subgroup of patients that is expected to respond to manipulative intervention. As such, these prediction rules are part of what seems to be a paradigm shift currently occurring within OMPT. The once predominant mechanism-based classification system that is based on the premise that impairments identified during examination are the cause of musculoskeletal pain and dysfunction12 is increasingly being replaced by treatment-based classification systems13–15. In the treatment-based system, a cluster of signs and symptoms from the patient history and physical examination is used to classify patients into subgroups with specific implications for management13.

What do these recent clinical prediction rules tell us? Flynn et al5 developed a clinical prediction rule consisting of five predictor criteria to identify a subgroup of patients with non-specific low back pain who were likely to benefit from thrust manipulation. This rule was subsequently validated by Childs et al6, who calculated an adjusted odds ratio of 114.7 at the 1-week follow-up and one of 60.8 for a positive functional outcome at the 4-week follow-up for patients who were positive on the rule (≥4 criteria present) and received manipulation versus those patients who were negative on the rule and received exercise. Fritz et al7 derived a subsequent two-factor rule from this prediction rule and reported a positive likelihood ratio of 7.2 for a positive outcome in patients with low back pain positive on both predictor variables and treated with manipulation.

Tseng et al8 identified six predictor variables for an immediate positive response to cervical manipulation in patients with neck pain including patients diagnosed with cervical spondylosis with or without radiculopathy, cervical herniated disk, myofascial pain syndrome, and cervicogenic headache. An increasing number of predictor variables present led to progressively higher positive likelihood ratios of an immediate positive response to manipulation: 4 predictor variables present yielded a likelihood ratio of 5.33 and an 89% probability of a successful manipulation8. Cleland et al9 derived six predictor variables in patients with mechanical neck pain without neurological involvement, indicating a likely positive response to a combination of three different thoracic thrust manipulations, one simple cervical range of motion exercise, and patient education. They suggested using a criterion of 3/6 variables present as a sufficient research-based indication for the use of thoracic manipulation in patients with mechanical neck pain: 3 of 6 variables present yielded a positive likelihood ratio of 5.5 and an 86% probability of a successful outcome. Table 1 provides the predictor variables in the various clinical prediction rules.

View enire article