J Manipulative Physiol Ther 2008 (Mar); 31 (3): 172–183 ~ FULL TEXT
Haymo W. Thiel, DC, PhD, Jennifer E. Bolton, PhD
Anglo-European College of Chiropractic,
Bournemouth, BH5 2DF England, UK
OBJECTIVE: Patients with nonspecific musculoskeletal disorders may vary in their response to treatment. This study set out to identify the predictors for either improvement or worsening in symptoms for which cervical spine manipulation is indicated.
METHOD: A large prospective study recorded details on patients, their presenting symptoms, and type of treatment. At the end of the consultation, any immediate improvement or worsening in presenting symptoms was noted. At the follow-up visit, information was collected on the patients' self-reported improvement.
RESULTS: Data were collected from 28,807 treatment consultations (in 19,722 patients) and 13,873 follow-up treatments.
The presenting symptoms of:
“shoulder, arm pain”,
“reduced neck, shoulder, arm movement, stiffness”,
“upper, mid back pain”, and
“none or one presenting symptom” emerged in the final model
as significant predictors for an immediate improvement.
The presence of any 4 of these predictors raised the probability for an immediate improvement in presenting symptoms after treatment from 70% to approximately 95%.
With regard to immediate worsening, "neck pain," "shoulder, arm pain, "headache," "numbness, tingling upper limbs," "upper, mid back pain," and "fainting, dizziness, light-headedness" emerged as predictors; and the presence of any 4 of these raised the probability for immediate worsening from 4.4% to approximately 12%. For global improvement, only 2 predictors were identified; but these did not enhance the postprediction probability.
CONCLUSIONS: This study is the first attempt to identify variables that can predict immediate outcomes in terms of improvement and worsening of presenting symptoms, and global improvement, after cervical spine manipulation. The predictor variables were strongest for immediate improvement.
Predictors for the Immediate Responders to Cervical Manipulation
in Patients with Neck Pain
Man Ther. 2006 (Nov); 11 (4): 306–315
From the Full-Text Article:
Mechanical neck disorders are very common; approximately 15% to 70% of the adult population can recall experiencing an episode of neck pain or stiffness in their lifetime.  In most cases, neck pain defies the identification of any pathological cause, leading to the term nonspecific. It is likely that within this group of nonspecific patients, a number of subgroups exist and that these groups may differ in their responses to treatment. Although there is evidence [2, 3] that spinal manipulation is an effective treatment of nonspecific neck pain, this evidence is not always consistent; one explanation may be the inclusion of different subgroups of patients in clinical trials giving overall effects that mask individual subgroup responses.
It is becoming clear that among patients where subgroups probably exist, there is a need to identify those patients who are likely to respond to treatment. Predicting outcome of treatment is a clinical tool that has the potential not only to improve the design of clinical trials but also the efficiency of the management of these patients. These so-called clinical prediction rules generally consist of combinations of variables obtained from self-reported measures and/or clinical examinations and can assist with identifying subgroups of patients that can be expected to benefit from treatment.
Some work has been done on clinical prediction rules as guides to the manipulative treatment of neck and back pain patients.  Flynn et al  developed a set of 5 predictor variables, including duration and pattern of symptoms, spinal mobility and hip range of motion, and patient attitudes, which identified a subgroup of back pain patients likely to benefit from spinal manipulation. A later study  refined these variables to just 2, namely, duration and pattern of symptoms, which were able to predict those patients most likely to improve after manipulation. In the first steps toward generating a clinical prediction rule, Tseng et al  were able to identify a set of 6 variables that predicted neck pain patients most likely to respond to cervical spine manipulation.
Spinal manipulation, like most interventions, can be associated with adverse effects in which there is a transient worsening of the presenting complaint(s). For example, increased stiffness and soreness and headache are common in clinical practice. [8–11] It would therefore be clinically useful to predict not only those patients likely to respond to manipulative treatment, but also those likely to experience adverse effects. The aim of this study was therefore to develop prediction rules to help identify patients likely to either benefit or worsen from cervical spine manipulation. To do this, a large-scale, prospective cohort study was conducted to establish and document the incidence and nature of beneficial outcomes and/or adverse effects after neck manipulation treatments administered by chiropractors.
This was a large-scale study investigating potential predictors of response to cervical spine manipulation. Although most of the patients in this study presented with either neck pain or reduced neck movement, these were not the only inclusion criteria. Hence, patients with other presenting symptoms for which cervical spine manipulation was indicated were included. Patients in whom a pathological cause is most usually absent are notorious for a mixed response to treatment; some respond well, whereas others do not. Being able to predict in advance of treatment those who will likely improve and those who, in the short term at least, may worsen is an extremely attractive proposition for clinicians in their management of patients.
The study by Tseng et al  appears to be the only one to date that has established clinical prediction models for immediate responders to neck manipulation. These investigators identified 6 variables that significantly predicted an immediate response, including self-reported measures and examination findings. They also established that the presence of 4 or more of these predictors increased the probability of success with manipulation from 60% to 89%. For the present study, a similar methodological approach was adopted to establish prediction models based on patient demographics, treatment, and presenting symptoms.
The most robust predictive model was identified for immediate improvement in presenting symptoms after treatment. In the final model, the presenting symptoms of “neck pain,” “shoulder, arm pain,” “headache,” and “upper, mid back pain” all significantly predicted an immediate improvement in presenting symptoms after cervical spine manipulation. The strongest predictor for immediate improvement after manipulation was the presenting symptom of “stiffness.”
The preprediction probability of immediate improvement in presenting symptoms after treatment was 70%. In other words, randomly manipulating individuals in this sample, without any attempt at prediction, may result in success in about 70% of the time. Our findings suggest that if at least 2 of the identified predictors are present, there is an 85% probability of achieving immediate improvement in presenting symptoms after manipulation. With 4 predictor variables present, the probability of achieving immediate improvement is raised to 95%. Paradoxically, with all 5 predictor variables in the model, the postprediction probability fell to 60%, that is, to a level lower than that of the preprediction probability for improvement. The most likely reason for this is that the model was constructed purely on the basis of an additive format and that, in reality, a linear relationship does not apply. It is quite possible that there are inverse relationships in the interactions between some of the predictor variables and that, by the nature of their combination, they actually lower the postprediction probability for achieving immediate improvement. In clinical practice, the presence of multiple symptoms is generally regarded to be a nonfavorable prognostic indicator. As such, it is conceivable that there could be a “critical” number of presenting symptoms or a combination of certain symptoms that, once reached or present, results in a lowering of the probability for improvement.
Using a similar analytic model, 8 variables were found to significantly predict immediate worsening in presenting symptoms after manipulation. Of these, “fainting, dizziness, light-headedness” and “numbness, tingling upper limbs” were the strongest. The preprediction probability of immediate worsening was 4.4%; and with 4 predictors present, this was raised to 12%. As was the case with the model for immediate improvement, adding more than 4 predictors resulted in a lowering of the postprediction probability. Again, the most likely explanation lies in the relationships in the interactions between variables.
When considering immediate response to neck manipulation, it is apparent that some of the same predictors can either predict improvement or worsening. This was the case for “neck pain,” “shoulder, arm pain,” “headache,” and “upper, mid back pain.” Although, for both of the outcomes, the predictor variables had achieved statistical significance, the ORs were much more favorable in terms of the prediction of immediate improvement and, as such, would tend to indicate stronger clinical significance. Furthermore, this situation reflects clinical practice, where there is often an absence of well-defined borders and where the same symptoms can bring a different meaning to a clinical situation and must be interpreted in view of the complexity of the clinical presentation. For example, a combination of neck pain together with upper limb symptoms is likely to indicate a different prognosis when compared with neck pain combined with headache.
These results must be considered in context of 2 other studies that used multivariate analysis and effect estimates in the form of ORs and CIs for the prediction of adverse effects related to neck manipulation. The study by Cagnie et al11 established headache as the most common postmanipulative reaction (in 19.8% of 280 patients) and, after multivariate analysis, concluded that upper cervical manipulation, use of medication, female sex, and age were all significant predictors of headache after spinal manipulation. In the University of California, Los Angeles, Neck Pain Study, Hurwitz et al9 reported a number of predictors for adverse reactions after manipulation, including those of neck pain, stiffness/soreness, radiating pain/discomfort, tiredness/fatigue, headache, and neurologic symptoms (this category included among others symptoms of dizziness, nausea and vomiting, impaired vision, tinnitus, and lower limb symptoms). Their reported risks for these symptoms are not dissimilar to the ones obtained in the present study.
Finally, this study attempted to predict global improvement based on the data available from 13873 follow-up treatments. Only the variables of “neck pain” and “toggle manipulation” were combined to significantly predict global improvement. However, in the accuracy analyses for these 2 predictors, the PLRs were approximately 1.0; and as such, it was not possible to identify any predictors for global improvement after cervical spine manipulation.
This study had a number of limitations. Despite being a prospective study with large numbers of data, these data were collected in routine clinical practices throughout the United Kingdom and as such are liable to issues of accuracy, reporting bias by the chiropractors and patients, and selection (of patient) bias. Patients participating in this study may also have been treated concurrently by other health care professionals. We relied on chiropractors recruiting treatment consultations in a consecutive manner. As the chiropractors were not blinded to the purpose, any deviation from this protocol may have introduced further bias.
This study is the first attempt to identify variables that can predict immediate outcomes in terms of improvement and worsening of presenting symptoms, and global improvement, after cervical spine manipulation.
From the findings, it was possible to identify some predictors of immediate improvement in presenting symptoms after cervical spine manipulation.
Patients presenting with symptoms of “reduced neck, shoulder, arm movement, stiffness,” “neck pain,” “upper, mid back pain,” “headache,” “shoulder, arm pain,” and/or “none or one presenting symptom only” are likely to report immediate improvement in these symptoms after treatment. Patients presenting with any 4 of these symptoms were shown to have the highest probability of immediate improvement. This finding may enhance clinical decision making for selecting cervical manipulation in the treatment of patients with one or more of these complaints. Although it was possible to identify a number of predictor variables for immediate worsening in presenting symptoms and global improvement after cervical spine manipulation, these failed to provide a robust predictive model for clinical application.
Predicting responses to treatment may aid in clinical decision making.
Predictors can be obtained from patient presenting symptoms.
Robust predictors could be identified for immediate improvement after neck manipulation.
Patients presenting with symptoms of “reduced neck, shoulder, arm movement, stiffness,” “neck pain,” “upper, mid back pain,” “headache,” “shoulder, arm pain,” and/or “none or one presenting symptom only” are likely to report immediate improvement after treatment.
Patients presenting with any 4 of these symptoms in the study had the highest probability of immediate improvement.
These findings may enhance clinical decision making for selecting cervical manipulation in the treatment of patients with one or more of these complaints.