Zygapophysial joint blocks in chronic low back pain: a test of Revel's model as a screening test

It is estimated that 15 – 40% of chronic low back pain patients have pain arising from the lumbar zygapophysial joints (ZJ)[1, 2]. Previous studies have indicated that historical and physical examination findings cannot predict results from diagnostic ZJ blocks [1‐6]. A specific sequence of injections of local anesthetic under fluoroscopic guidance into the joint space or targeting the medial branches of the dorsal rami, are reference standards for diagnosis. For the ZJ to be considered the sole source of pain, there must be a total or near total pain ablation for a time consistent with the known properties of the anesthetic agent[7, 8]. In a recent study by Revel M et al (1998), the authors found that a cluster of seven items, (hereinafter called Revel's criteria) were shown to be of value in predicting a 75% reduction of pain following a single intra-articular anesthetic injection into the ZJs. The items in the cluster are: age over 65 years, pain well relieved by recumbency, no exacerbation of pain with: coughing and sneezing, forward flexion, extension, rising from flexion and the extension-rotation test[9]. These authors suggested two clinical strategies: Strategy 1 consists of five or more items being true. Strategy 2 is the same except that one of the true statements must be 'pain well relieved by recumbency'. Estimated sensitivities and specificities for strategies 1 and 2 were 100/92%, and 66/80% respectively. A subsequent study of 200 consecutive patients using double blocks failed to confirm these findings[6].

Because prognosis for acute low back is good, invasive and expensive and invasive diagnostic testing cannot be justified. However, persistent disabling pain needs more intensive investigation in order to determine appropriate management strategies. Back pain of ZJ origin may be treated using intra-articular steroid injection[7, 10]., or radio frequency neurotomy [11‐14]. However, the low prevalence of the condition dictates that a clinical selection process capable of identifying patients unlikely to respond to the initial screening diagnostic block is desirable. Patients with a low probability of a positive anesthetic response need not be subjected to the screening block and the tissue origin of pain should be sought elsewhere. Revel's criteria are currently the only documented clinical means by which response to an initial screening block may be predicted, but it remains a provisional finding only, until further research confirms the previous findings.

The current study objectives were to estimate the predictive value of the two clinical strategies of Revel et al (1998), using similar measurement parameters (75% reduction in pain VAS after ZJ block), and apply alternative analytic methods to explore any potential utility of the variables used.

Initial ZJ blocks were carried out on 151 chronic low back pain patients. Thirty-four patients were excluded from analysis as they received another intervention in the same procedure session and did not return for differentiating and confirmatory blocks. One case was excluded through incomplete data on Revel's criteria. Following ZJ block, 27 of 116 patients satisfied reference standard A. Data required for determination of Reference standard B were missing for five of the 116 cases. Eighteen of these 111 patients satisfied reference standard B. Table 1 contains demographic and other descriptive characteristics with comparisons between included and excluded patients. Included patients had longer time off work than excluded patients (mean 55 versus 99 weeks) but otherwise had similar characteristics.

In specifically evaluating Revel's criteria against reference standard A, logistic regression failed to achieve significance as a model (n = 108, p = 0.46). Two variables; "absence of pain with coughing and sneezing" and "no exacerbation of pain rising from flexion", showed a trend towards significance as predictors within the model (p = 0.07).

Using Revel's criteria within a logistic regression with reference standard B as the response variable, a strong trend towards significance was reached (n = 100, p = 0.06). One component variable (age over 65) individually reached significance within the model: (p = 0.004, odds ratio 16.1 with 95% confidence intervals of 2.4 and 107.8). In the whole sample 12.5% were aged over 65 years whereas of the 19 positive responders six were over 65 years (31.6%).

The same patients satisfied both strategies of Revel et al. Estimated sensitivity, specificity, predictive values and positive likelihood ratios for both strategies of Revel et al are presented in Table 2. With respect to reference standards A & B, sensitivities are low at 11 and 17%, specificities are high at 91 and 93% and likelihood ratios are 1.2 and 2.5 respectively.

Of the 27 patients satisfying reference standard A, 13 (48%) returned for confirmatory ZJ blocks. Three of these reported 75% or more reduction in pain following the confirmatory block. None of the three patients with confirmed ZJ pain satisfied Revel's criteria.

The current data produces results that are in stark contrast to those of Revel et al (1998) with sensitivity low and specificity high. However, 'no pain with cough and sneeze' and 'no exacerbation of pain rising from flexion' approached statistical significance in relation to a 75% reduction pain after ZJ block (reference standard A). These variables are in line with Revel's results. Age over 65 is associated with reference standard B (abolition of primary pain). Likelihood ratios for the criteria are lower in the current data also. Some of the differences in results may be explained by a number of factors:

1. Revel et al's study was a placebo controlled design whereas we did not routinely utilize a similar or equivalent control.

2. The patients in Revel et al's sample were older (mean 58 versus 43 years), and had a shorter duration of symptoms (mean 78 versus 160 weeks).

3. It is also possible that the high level of standardization when acquiring the criteria data in the current study, and the prospective methodology might have contributed to the differences in results.

Following anaesthetic blocks, patients frequently state that the pain prompting consultation is abolished, yet a post-procedure pain VAS still registers more than 0/100. In this study reference standard B was evaluated as an alternative to the usual standard, so that pain in areas above the lumbar spine or pain directly attributable to the needle insertion site do not result in inappropriate post procedure VAS scores. In the interests of developing more precise instruments documenting results of diagnostic blocks, we propose that in future studies, the VAS scale should specifically refer to the primary pain complaint for which the diagnostic block is undertaken.

Based on the current data, Revel's criteria are not suitable as a screening test to select patients suitable or unsuitable for an initial screening ZJ block. Such a screening device should have high sensitivity to ensure that most patients likely to respond are included in the initial diagnostic block. Our study suggests that, at best, Revel's criteria, might identify a small subset (11%) of patients likely to respond to a screening block (specificity 93%).

Neither strategy utilizing Revel's criteria is suitable as a clinical screening device for selection of chronic LBP patients for initial diagnostic ZJ blocks. In contrast to Revel's findings, the current data demonstrated low sensitivity and high specificity for these clinical criteria. The high specificity of the criteria reported in this study relates to a single uncontrolled screening block. Consequently these criteria can not considered diagnostic of painful lumbar ZJ. Only placebo controlled or double ZJ blocks are able to diagnose this source of low back pain.