Clare, et al 10 used 2 physical therapists trained in the McKenzie method to examine 25 patients with cervical pain. They found good inter-examiner reliability (IER) (kappa, [k] = 0.63 and 93% agreement) for the assessment procedure.

No studies were identified that have addressed the validity of centralization signs in the cervical spine.

Hubka and Phelan 18 assessed the IER of palpation for tenderness between 2 practitioners in 30 patients with unilateral neck pain. They found good IER (k = 0.68). Jull, et al 19 assessed IER of segmental palpation using 7 examiners and 40 subjects with or without neck pain and headache. The criteria for a positive test were based on resistance to joint movement and pain provocation in response to palpation. Kappa values indicated excellent to perfect IER (k = 0.78–1.00) in 6 instances, fair to good (k = 0.45–0.65) in 14 instances and poor (k = 0.25–0.34) in 5 instances. They point out that, in the instances of poor agreement, the raw data indicated that the examiners had agreed on 13 of 14 decisions. But the calculations of k were vulnerable because 12 of the 13 agreements were in the same cell of agreed negative finding. Marcus, et al 20 used 4 physical therapists to examine 72 headache patients and 24 controls. The therapists examined all subjects for "cervical synovial joint abnormalities" in the same manner as described in the study by Jull, et al 19. They found good IER (k = 0.63) between examiners. McPartland and Goodridge 21 assessed IER of "TART" exam, described as segmental palpation that focused on three parameters: tissue texture change, restriction of vertebral motion and zygapophyseal (z) joint tenderness. They found the IER of examination that considered all three parameters was poor (k = 0.35 for asymptomatic subjects, k = 0.34 for symptomatic subjects). But for the parameter of tenderness alone, IER improved (k = 0.529). Van Suijlekom, et al 22 used 2 neurologists to examine 24 headache patients and found IER for segmental palpation to be slight to fair (k = 0.14 to 0.37). However, the palpation method was poorly described in this study. Also, it is not known as to whether the difference between the findings of this study and those of the other studies reported here relate to the fact that the "negative" IER studies used neurologists, whereas the "positive" IER study used chiropractors or physical therapists. Cleland, et al 23 used 2 examiners and 22 subjects and found highly variable IER between 2 physical therapists for palpation for pain provocation, with k ranging from -.52 to .90, depending on the segment involved. They speculated that this high variability related to the clinicians not agreeing on the segmental level being examined, as opposed to lack of agreement on the findings.

Jull, et al 24 used diagnostic blocks to identify the presence and location of symptomatic z joints in 20 patients with cervical related pain. The patients were examined by a manipulative physiotherapist who also attempted to identify the presence and location of symptomatic z joints. The definition of a symptomatic joint as determined by palpation was based on abnormal "end feel", increased resistance to motion and reproduction of pain. They found that the SE and SP were both 1.00. That is, the examiner was able to identify 100% of the symptomatic segments as well as all of the subjects whose pain was not abolished by diagnostic block. This study used single, rather than double blind, diagnostic blocks. Regardless, as will be discussed below, the use of diagnostic blocks as a Gold Standard for the presence of z joint pain has been questioned 25. Treleaven, et al 26 assessed 12 patients with postconcussion headache with segmental palpation. The method of palpation was the same as that used by Jull, et al 24. They found complete agreement between the examiner and independent report of the patient as to which segments were painful and almost complete agreement as to which segment was most painful. Sandmark and Nisell 27, calculated the SE, SP and PPV and negative predictive value (NPV) of segmental palpation in the cervical spine relative to reported neck pain. They found these values to be 0.82, 0.79, 0.62 and 0.91 respectively. Lord, et al 28, used a double blind anesthetic block to determine the prevalence of pain arising from the C2-3 z joint in patients with the complaint of chronic headache after cervical trauma. These authors demonstrated that the prevalence of C2-3 z-joint pain was 53%, and the only sign that was associated with these patients was tenderness to palpation over the C2-3 z joint. They calculated that palpation had SE of 0.85, a positive likelihood ratio (PLR) of 1.7 and a negative likelihood ratio (NLR) of 0.3. The precise method of palpation was not described. Zito, et al 29 using the palpation method found to be reliable by Jull et al 19 found a significantly higher incidence (p < 0.05) of hypomobile and painful z joints in the upper cervical spine of patients classified according to the International Headache Society criteria as having cervicogenic headache compared to those classified as having migraine with aura. King, et al 30 used "controlled, diagnostic blocks" as a Gold Standard against which segmental palpation that was described as being similar to that of Jull, et al 24. They found the SE to be 0.88, SP to be 0.39 and PLR to be 1.3. Again, using diagnostic block as a Gold Standard may be questionable 25, leaving open the issue of what should be the Gold Standard for segmental palpation signs. Further work in the area of establishing a true Gold Standard for the identification of zygapophyseal joint pain may be needed before definitive statements regarding the presence or absence of pain from this structure can be made.

The standard neurodynamic test in the cervical spine is the brachial plexus tension test (also known as the upper limb tension test 31). Wainner, et al 32 found good to excellent IER of this test (k = 0.76 to 0.81). They also found good to excellent IER of several historical questions of patients with documented cervical radiculopathy (k = 0.53 to .082). They found varying IER of neurologic exam findings, but good to excellent IER of Spurling's test (which they described as bending the seated patient's head toward the side of symptoms, rotating and extending slightly, and applying downward pressure), the cervical distraction test and Valsalva's maneuver. The kappa values for these tests ranged from 0.60 to 0.88.

Wainner, et al 32 provide data on the SE, SP PLR and NLR of a variety of historical factors and examination procedures. They found that the cluster of 4 tests – Spurling's test, the upper limb tension test, the cervical distraction test and limited rotation toward the side of symptoms secondary to pain – carried the greatest diagnostic accuracy as compared to the Gold Standard of electromyography. When 3 of these tests were positive, there was a 65% probability of the presence of cervical radiculopathy the SE and SP were 0.39 and 0.94, respectively and a PLR of 6.1. When all 4 tests were positive, there was a 90% probability of the presence of cervical radiculopathy. The SE and SP were 0.24 and 0.99 respectively and the PLR was 30.3.

Shah and Rajshekhar 33 also used Spurling's test, the description of which was the same as that in the Wainner, et al study 32, and found it to be useful in identifying "soft disc prolapse" as opposed to "hard disc" (i.e., osteophyte). They calculated the SE and SP to be 0.90 and 1.00, respectively compared to the Gold Standard of operative findings. The PPV was calculated to be 1.00 and the NPV to be 0.71. In patients treated non-surgically, they used MRI as the Gold Standard and calculated the SE and SP to be 0.90 and 0.93, respectively. The PPV was calculated to be 0.90 and the NPV to be 0.93.

Marcus, et al, in the same study cited above 20 found good to perfect IER of TrP palpation in the cervical spine (k = 0.74), head (k = 0.81) and shoulder (k = 1.00). van Suijlekom, et al 22 in the study cited above, found variable IER (k = 0.0 – 1.00) of TrP palpation in patients with headache. As was the case with segmental palpation, the method of TrP examination was poorly described. Gerwin, et al 34 performed 2 different experiments to assess IER. In the first, 4 examiners assessed 20 different muscles on each of 25 patients with various symptom presentations. They used a general observer-agreement statistic called the "S_av", which they defined as "a generalized version of the Cohen's kappa which reports pairwise judge agreement corrected for chance agreement." They found poor IER (S_av = 0.0–1.0). They then repeated the study after spending a 3-hour session in which the examiners discussed positive findings and palpation techniques. They found good to excellent IER (S_av = 0.65 – .95) after the training session. Sciotti, et al 35 found good IER (Generalizability coefficient = 0.83–0.92) between 2 examiners looking for latent trigger points (TrPs) in the upper trapezius muscle. However, the subjects were asymptomatic. On the other hand, Lew, et al 36 found poor IER for TrP palpation in the upper trapezius, although the subjects in that study were also asymptomatic.

The validity of muscle palpation signs is unknown, largely due to lack of an appropriate Gold or reference standard.

Early studies 6869 failed to demonstrated adequate IER of the McKenzie assessment in the lumbar spine. For example, Riddle and Rothstein 68 looked at 363 patients with LBP and used 49 physical therapists at 8 different clinics and found poor IER (k = 0.26) of the classification systems of McKenzie. Postgraduate training in the system did not improve IER. However, these studies have been criticized on the grounds that minimally trained therapists were used, the study failed to consider the classification of patients into subsyndromes and, in the case of Kilby, et al 69, the protocol included elements that are not a standard part of the McKenzie system 10. More recent studies have attempted to improve upon the methodology of these earlier studies. Werneke, et al 70 used 5 physical therapists who assessed 289 patients with LBP or neck pain and found IER that ranged from k = 0.917 to 1.0. Fritz, et al 71 used 40 physical therapists in practice and 40 physical therapy students and had them watch a video of 12 examinations using the McKenzie method. They found IER coefficients ranging from k = 0.763 to 0.823. Razmjou, et al 72 used 2 trained McKenzie therapists and 45 patients with acute, subacute or chronic LBP and found good IER (k = 0.70). Kilpikosk, et al 73 looked at 39 patients with low back pain examined by 2 physical therapists trained in the McKenzie method. They found good agreement for the presence of the centralization sign (k = 0.7) and excellent agreement for direction preference (k = 0.9). Clare, et al 10 found perfect IER (k = 1.0) between 2 examiners in 25 patients with LBP.

Donelson, et al 74 found that the McKenzie assessment differentiated discogenic from nondiscogenic pain (p < 0.001), using discogram as the Gold Standard. Young, et al 75 used the Donelson, et al 74 data and calculated the sensitivity (SE) and specificity (SP) to be 0.94 (95% confidence interval [CI] 0.82, 0.99) and 0.52 (95% CI 0.34, 0.69), respectively. Young, et al 75, using their own original data, calculated the SE and SP of centralization signs to be 0.47 and 1.00, respectively, also using discography as the Gold Standard. They also found that pain upon arising from a sitting position was associated with disc pain (p = .017). This historical factor may therefore be useful in identifying the "centralizer", though as will be noted below, pain when arising from sitting is also associated with segmental pain provocation signs in the sacroiliac (SI) area. Laslett, et al 76 also used discogram as the Gold Standard and calculated the SE, SP, and positive likelihood ratio (PLR) and negative likelihood ratio (NLR) for centralization signs to be 40%, 94%, 6.9 and 0.63 respectively. They also used the Roland Morris Disability questionnaire to measure disability and the Distress Risk Assessment Method to measure distress, and found these factors altered the SE, SP and PPV. In the presence of severe disability, these values were 46%, 80%, 3.2 and 0.63 respectively and in the presence of severe distress they were 45%, 89%, 4.1 and 0.61 respectively.

It is pointed out by Long, et al 77, that it is not necessary to assume a particular pain generating tissue when using the McKenzie assessment as a means of making treatment decisions. In their study, clinical decisions were made regarding exercise direction based on the findings of the end range loading examination. One group of patients were given exercise maneuvers in the direction of centralization of symptoms, another was given exercises in the direction opposite that of centralization, and a third group was given exercises that did not consider any specific direction. They found significantly greater improvement (p < 0.001) in outcome in the patients who were given exercises in the direction of centralization, suggesting that the McKenzie evaluation in the lumbar spine allows clinicians to make treatment decisions that are of ultimate benefit to patients. This may be a more important measure of "validity" than the identification of a certain pain generating tissue (e.g., using a prognostic criterion as a reference standard for the assessment method).

Centralization signs have also been found to be predictive of long term outcome. Werneke and Hart 78 found that discriminating between patients who exhibit centralization signs from those who do not allows for prediction of pain, disability and return to work at 1 year. In a separate study, Werneke and Hart 79 compared classification according to centralization signs with classification according to the Quebec Task Force (QTF) criteria 80. They found that examination for centralization signs had greater predictive validity for pain and disability at discharge from care than the QTF criteria. Werneke and Hart have also found that assessing centralization signs over the period of multiple visits allows for more accurate discrimination than a single assessment 81.

Similar to what was found for the cervical spine, palpation for movement restriction in the lumbar spine has not been shown to be reliable, though palpation for pain has. Keating, et al 82 used 3 chiropractors who examined 25 asymptomatic subjects and 21 patients with low back pain. They found marginal to good IER of palpation for pain provocation over bony structures (k = 0.19 to 0.48) and soft tissues (k = 0.10 to 0.59). The strongest IER was found for the L4-5 and L5-S1 segments. Maher and Adams 83 used 2 examiners to assess 90 subjects with low back pain, allowing each examiner to use whatever palpation method he or she chose. The examiners assessed each patient for pain and stiffness. They found that, while the IER of palpation for stiffness was low (intraclass correlation coefficient [ICC] = 0.03–0.37) the IER for pain was good (ICC = 0.67–0.72). Strender, et al 84 used 2 medical physicians and 2 physical therapists to evaluate 71 patients with low back pain. They found moderate agreement (k = 0.40) for palpation for tenderness. Lundberg, et al 85 used 2 examiners to assess 609 female subjects for segmental mobility and pain provocation through palpation. They found good IER (k = 0.67 – 0.71) for this assessment.

Seffinger, et al 86 systematically reviewed the literature regarding the IER of palpatory diagnosis in both neck and back pain. They concluded that palpatory procedures for pain provocation generally have acceptable IER (k = 0.40 or greater) and that 64% of studies looking at pain provocation found acceptable IER.

With regard to the SI area, the earliest study of IER was that of Potter and Rothstein 87. They did not use the kappa statistic, but they found that tests that attempt to determine movement abnormality had poor reliability (less than 70% agreement) but the 2 tests that relied on patient response had agreement of 70–90%. Carmichael 88 also found poor IER (k = 0.314) of an SI test that assessed for mobility. Freburger and Riddle 89 found poor reliability (k = 0.18) of the measurement of SI joint position using handheld calipers. Robinson, et al 90 evaluated the reliability of various pain and SI joint dysfunction tests. The palpation test for joint play showed very poor reliability (k = -0.06). Other pain provocation tests demonstrated moderate to good reliability (k = 0.43–0.84). When clustered results of three to five pain provocation tests were used there was also good reliability (k = 0.51–0.75). A study by Vincent-Smith and Gibbons 91 evaluated the IER and intra-examiner reliability of the standing flexion test for SI joint dysfunction. Intra-examiner reliability was moderate (k = 0.46) while IER was very poor (k = 0.052).

Tong, et al 92 tested the hypothesis that combining the test results of various measures of SI joint dysfunction would yield greater reliability than individual tests. They established three methods to be evaluated; Method 1: using the test result with the highest IER; Method 2: requiring at least one test result to be abnormal for the variable to be abnormal, and; Method 3: requiring all test results to be abnormal for the variable to be abnormal. Kappa scores were 0.47, 0.08, and 0.32 using Method 1 for the sacral position, innominate bone position, and side of sacroiliac joint dysfunction, respectively. For Method 2 the values were 0.09, 0.4, and 0.16. For Method 3 the values were 0.16, 0.1, and -0.33.

Laslett and Williams 93 used 2 examiners to evaluate 51 patients using 6 tests designed to identify a painful SI joint. They found moderate to high IER (k = 0.69 to 0.82), of several tests. Dreyfuss, et al 94 found moderate IER (k = 0.61 to 0.64) for 3 SI pain provocation tests. Kokmeyer, et al 95 found good IER (k = 0.70) of a cluster of 5 SI pain provocation tests. Studies that have evaluated tests of SI mobility have generally found poor IER 96.

Young, et al 75 found a correlation between abolishment of pain with facet joint blocks and the absence of a historical report of pain when standing from a sitting position. Revel, et al 97 found that the following characteristics were associated with patients whose pain was relieved by 75% or more with facet joint blocks: age over 65, pain not exacerbated by coughing, pain not worsened by hyperextension, pain not worsened by forward flexion, pain not worsened by rising from forward flexion, pain not worsened by extension-rotation and pain well relieved with recumbency. Similar findings have been found by other authors 9899. Laslett, et al 100 found that these criteria had low SE (< 0.17), though they did have high SP (0.90). Laslett, et al 101 found that 4 or more out of the following 7 signs carried a SE of 1.00 and SP of 0.87 as compared to single facet joint blocks: Age ≥ 50, symptoms best walking, symptoms best sitting, onset pain is paraspinal, Modified Somatic Perception Questionnaire score > 13, positive extension/rotation test, and absence of centralization signs. So, as will be seen in the SI joint area, ruling out centralization signs is necessary to increase the diagnostic yield in identifying segmental pain provocation signs.

In the SI joint area, Broadhurst and Bond 102 compared 3 pain provocation tests with anesthetic block and found the SE of single tests ranged from 0.77 to 0.87. The SP of each test was 1.00. Slipman, et al 103 used a cluster of pain provocation tests and used the criteria of at least 3 "positive" tests in 50 consecutive patients with LBP. They compared this examination with the Gold Standard of single anesthetic blocks. They estimated the PPV of the examination to be 60%. van der Wurff, et al 104 assessed 140 patients with chronic LBP with a cluster of 5 pain provocation maneuvers for the SI joint. This cluster was the same as that used in the study by Kokmeyer, et al 95 that had found good IER. They considered that 3 out of the 5 tests being pain-producing constituted a "positive" test. They compared this regimen with the Gold Standard of double anesthetic blocks. They calculated the SE of the regimen as 0.85 (95% CI, 0.72–0.99) the SP as 0.79 (95% CI, 0.65–0.93), and the PPV and NPV as 0.77 (95% CI, 0.62–0.92) and 0.87 (95% CI, 0.74–0.99), respectively. The PLR was 4.02 (95% CI, 2.04–7.89); the NLR was 0.19 (95% CI, 0.07–0.47). Laslett, et al 105 used these same SI provocation tests and compared these to single anesthetic block. They added to the Gold Standard criteria the reproduction of concordant pain upon infiltration, followed by 80% or more reduction of pain as a result of injection. They found that the presence of 3 positive tests carried a SE of 0.94, a SP of 0.78, a PPV of 0.68, and a NPV of 0.96. Young, et al 75 also found significant (p < .001) association between the presence of 3 or more positive pain provocation tests for the SI and positive SI injection and also found positive association between positive SI injection and the following historical factors: pain when arising from a sitting position (p = .02), pain being unilateral (p = .05) and the absence of midline pain (p = .05). They also noted that patients with positive SI injection rarely had pain superior to the L5 level.

Importantly, Laslett, et al 106 found that performing SI provocation maneuvers in the context of the end range loading exam for centralization signs (see below) increases the diagnostic yield of the SI tests. The SP of the SI provocation tests rose from 0.78 to 0.87 and the PLR rose from 4.16 to 6.97.

Slipman, et al 107 compared radionuclide imaging to the Gold Standard of single SI joint block and found this test to have high SP (100%) but very low SE (12.9%).

The standard neurodynamic tests in the lumbar spine are the Straight Leg Raise (SLR), Femoral Nerve Stretch test (FNST – also sometimes referred to as the Prone Knee Bend 108) and the Slump test. Clinicians will often include Bragard's test (adding ankle dorsiflexion to the SLR) and the Well Leg Raise (WLR) test (eliciting pain on the affected side by performing a SLR on the contralateral limb) to serve as sensitizing and differentiating maneuvers for the purpose of increasing the specificity of the examination for lower lumbar nerve root pain 109.

Hunt, et al 110 assessed the IER of the SLR using 2 teams of examiners, each team consisting of one physician and one physical therapist. They found fair IER (k = 0.54 for left leg, 0.48 for right leg) but this study used asymptomatic subjects and measured SLR using a goniometer. Vroomen, et al 111, used a neurologist and a neurology resident to assess 338 patients with "sciatica". They calculated the IER of a variety historical factors and clinical tests in patients with suspected lumbar radiculopathy. For the standard SLR, they found good IER (k = 0.68) when the interpretation of the test findings included the production of "typically dermatomal pain". The k values for the Bragard's and WLR tests were 0.66 and 0.70, respectively. When historical and examination factors were taken into consideration regarding arriving at a diagnosis of nerve root pain, the k value was 0.66. The historical factors with the greatest IER were increased pain with coughing/sneezing/straining (k = 0.64), increased pain with walking (k = 0.56), coldness in the lower extremity (k = 0.56), urinary incontinence (k = 0.79) and previous back pain episodes (k = 0.67).

McCombe, et al 112 used 2 surgeons to assess 50 patients and found fair agreement for the FNST (k = 0.3–0.5). Philip, et al 113 used 6 pairs of physiotherapists to examine 93 patients using the Slump test. They found good to perfect IER (k = 0.72 to 1.00). Gabbe, et al 114 used a physiotherapist and a research student to assess 15 asymptomatic volunteers using the slump test and found excellent reliability (ICC = 0.92, 95% CI 0.77, 0.97).

Vroomen, et al 115 found that SLR was not predictive of the presence of herniated disc on MRI. They did not assess WLR or Bragard's test. They did note that the historical factors of a dermatomal distribution of pain, increase in pain on coughing, sneezing, or straining, paroxysmal pain, and predominant leg pain were predictive. Using MRI as a "Gold Standard" may be questionable because of the potential for false positive findings 116. Lurie 117 reviewed the literature on diagnostic tests for LBP and found that the SLR has generally been found to have high SE (0.78 to 0.97) but low SP (0.10 to 0.52) in identifying patients with disc herniation. The opposite is found for WLR test, which has been found to have low SE (0.22 to 0.52) and high SP (0.85 to 1.0). He does note, however that "much of the literature is limited by methodological flaws". Many clinicians feel that the combination of the SLR and WLR, along with Bragard's test and other "localizing" and "sensitizing" maneuvers improves the SE and SP of the examination for pain of neural origin 109. This has not been specifically evaluated.

The validity of the FNST has not been well studied 117.

Stankovic, et al 118 found those patients with the complaint of LBP and/or leg pain whose imaging revealed a herniated disc were more likely to have distal pain in the lower extremity on the performance of the Slump test, although the difference was not statistically significant (p < 0.017). No values with regard to SE, SP and PPV and NPV were calculated.

Nice, et al 119 used 12 examiners to assess 50 patients with LBP for trigger points, using the standard criteria of the presence of a "taut band" and localized "nodule", the presence of a "twitch response" and the reproduction of familiar pain. They found IER to be poor (k = 0.29 to 0.38). Njoo and Van der Does 120 also found poor IER when considering all of the standard criteria of TrP presence. However, when considering only tenderness to palpation, particularly when combined with the identification of concordant pain on the part of the patient, IER increased greatly (k > 0.5). Hsieh, et al 121 used 1 "expert" DC with many years of experience with TrP palpation, 2 DC's with 15 years of practice experience but not with extensive experience with TrP palpation, and several chiropractic and psychiatry residents. They provided all clinicians with 3 2-hour lectures and 3 2-hour hands-on sessions as training in TrP palpation, and compared the agreement between the expert and the others for the presence of taut band, local twitch response and referred pain. They found generally poor IER, concluding that even with experienced clinicians, short term training in TrP palpation is not enough to provide IER.

It would appear that if the examiner places greatest emphasis on tenderness to palpation and reproduction of concordant pain, and less emphasis on the presence of a taut band and a twitch response, the IER of muscle palpation signs will be enhanced. Also, Simons has pointed out 122 that those studies using untrained and/or inexperienced examiners have generally found poor IER, whereas those using trained and experienced examiners have generally found favorable IER in TrP examination, indicating the importance of examiners having appropriate training and experience with muscle palpation signs.

As with the cervical spine, the validity of myofascial signs in the lumbar spine is unknown due to the absence of a Gold Standard for the identification of myofascial pain.