CPR for the Lumbar Spine: Why Is Evidence So Variable?

About 10 years ago, the physical therapy community adopted Clinical Prediction Rules (CPRs) for treatment and diagnosis of their patients. In the United States, if you weren’t using, thinking, incorporating, bathing in, or breathing in clinical prediction rules, you weren’t performing evidence-based therapy. Multiple clinical prediction rules (CPRs) were created to diagnose challenging conditions, manipulate the low back, tape the patella, determine immediate improvement after manipulation of the low back, and various other dubious treatment or examination. The fervor for CPRs led to the premature adoption of these tools into clinical practice, and, in some cases, the incorporation of them into clinical practice guidelines.

CPRs are algorithmic decision tools that use parsimonious clinical findings to aid clinicians in determining a diagnosis, prognosis, or likely response to an intervention. Each tool combines stand-alone clinical findings into clusters. The clusters improve the specificity of the rule (if positive findings are present) or improve the sensitivity of the rule (if negative findings are present). The first formally reported publication on CPRs was by Wasson and colleagues in 1985. Since then, there have been a myriad of CPRs that are specific to physical therapists.

CPRs in Practice

Prescriptive CPRs identify characteristics related to a specific intervention, which are then related to a good outcome (or, if used incorrectly, a bad outcome). The design of prescriptive CPRs requires a control group to determine differences in outcomes. A good example is the CPR by Wright and colleagues.¹ They report that patients with unilateral hip pain, age of ≤58 years, pain of ≥6/10 on a numeric pain rating scale, 40-meter self-paced walk test time of ≤25.9 seconds, and duration of symptoms of ≤1 year had better outcomes when seen by physiotherapy, compared with controls who did not receive physiotherapy, and controls who lacked these characteristics. Prescriptive CPR studies are more difficult to design, frequently misunderstood, and often accepted outside the context of this clinical pearl.

One early CPR (lumbar spine manipulation CPR) that was clinically derived and validated in separate studies involved manipulation of the lumbar spine when individuals met baseline characteristics. The baseline criteria for individuals involved in those studies was:

1. Duration of symptoms <16 days
2. Hip internal rotation of at least 35 degrees
3. Lumbar segmental hypomobility tested with a spring test
4. No symptoms distal to the knee
5. Score of <19 on the work subscale of the Fear-Avoidance Beliefs Questionnaire

Four of five positive findings increased the odds of a short term positive response from manipulation by 25 fold and demonstrated better outcomes than the control group that received exercises.

Three other studies explored the concepts of the lumbar spine manipulation CPR. None of them found the same results as the original studies, and one found no relevance at all. It is important to recognize that these studies did not involve the same parameters (patient sample, type of intervention, length of care, etc.) as did the original derivation and validation studies.

Our group was one of the three groups that evaluated this work.^2,3 We found different results as well. We suggested that meeting the CPR was prognostic, in that, if the rule was met, you were going to get better no matter which intervention you received. The CPR was not specific to manipulation (the mobilization group also improved), and it was not sensitive enough to help all individuals who would benefit from manipulation.

Making Sense of CPRs

Why do we find variable findings with CPRs? Interestingly enough, it’s the same reason we find variations in randomization controlled trials: different populations lead to changes in findings. Slight alterations in how the interventions are implemented can also affect results. Small sample sizes lack stability when a study is repeated. Nearly all CPRs use a dichotomous outcome measure, which can lead to unstable results as well. The modeling used can also lead to greater deviations across studies, with differences found in the predictive variables after minuscule changes to minimally important points.

In most cases, CPRs are not usable on your day-to-day clinical population. One must use common sense and clinical reasoning to determine the risks and benefits of applying each dedicated care sequence. There are presently no rules in practice that do not deserve additional vetting. However, if the finding fits your patient population, and passes your clinical reasoning filter, then it may add additional information to your thought processing.