Abstract
Introduction
The association between disk degeneration (DD) on MRI and low back pain (LBP) remains controversial and different studies report conflicting findings. There are many possible reasons for the conflicting findings including the setting or patient population, study design, analysis used and, probably most importantly, how both DD and back pain status are measured.
A key issue for almost all studies in this area is how to define DD and rate an individual as having it or not. Most studies use a scale such as that reported by Pfirrmann et al1 to quantify DD at each lumbar level; however, there is no agreed threshold (e.g., 3/5 or 4/5) before someone is considered to have clinically relevant DD. It is also not known if the association between DD and LBP is significantly stronger when DD is measured on a continuous scale (e.g., 1 to 5), than when dichotomized at any single threshold. A related issue is whether DD for an individual should be based on the single worst level or on a summary score from multiple levels. For example, is a person with minimal DD at all levels apart from one level with severe DD more or less likely to have LBP than a person who has three levels with moderate DD? Both the worst single level and summary score approach to rating DD are used in the literature and it is possible that one approach is more strongly related to the presence of LBP.
Therefore, the aims of this secondary exploratory analysis were as follows: (1) assess different thresholds of the Pfirrmann et al1 scale to determine if an optimal threshold exists for differentiating people with and without LBP, (2) compare findings obtained from different thresholds to a continuous score from the Pfirrmann scale, and (3) compare findings based on an individual's worst DD score at any level to summary scores from multiple levels.
Materials and Methods
We compared DD between 30 cases with current acute LBP and 30 pain-free controls. Cases were patients presenting for care with likely discogenic LBP (demonstrated centralization with repeated movement testing), of moderate intensity and with minimal past history of back pain. Controls were matched for age, gender, and past history of back pain. Cases and controls underwent MRI scanning which was read by two blinded assessors for DD at each lumbar level (1–5) according to the criteria described by Pfirrmann et al.1 For the purpose of this secondary analysis, we used the findings from only one assessor, a radiologist specializing in spinal conditions.
To investigate if an optimal threshold of disk degeneration exists for differentiating those with and without LBP, we calculated area under the curve (AUC) and 95% CI for different levels of DD (≥2, ≥3, ≥4, ≥5) based on the worst spinal level for each individual. We compared this to AUC when using a continuous measure (1–5) for the worst level of DD in each individual. To investigate if a summary DD score was better than the worst level DD score at differentiating those with and without LBP, we calculated AUC for a range of summary scores including, addition of DD score for all levels, addition of DD score for lowest 2 or 3 spinal levels, and the number of spinal levels with ≥3 or ≥4 DD score.
Results
Table 1 presents the AUC for the continuous DD score and for different thresholds of DD (both using worst single level) as well as sensitivity and specificity for each threshold. A threshold of ≥4 (worst single levels) was most strongly associated with the presence of back pain (AUC = 0.77). The continuous DD score had a similar AUC to the threshold of ≥4. AUC was similar to DD rated according to the single worst level (AUC = 0.79, 95% CI, 0.67–0.90) and the DD summary scores suggest that neither option is superior with regard to being able to differentiate those with and without LBP. AUC for DD summary scores was 0.80 (0.68 to 0.91) for addition of all five levels, 0.79 (0.68 to 0.91) for addition of lower three spinal levels, 0.75 (0.62 to 0.87) for addition of lower two spinal levels, 0.77 (0.65 to 0.89) for the number of spinal levels with DD score ≥3, and 0.78 (0.66–0.90) for number of spinal levels with DD score of ≥4.
| Criteria | AUC (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) |
|---|---|---|---|
| DD score ≥2 | 0.58 (0.44–0.73) | 0.97 (0.83–1.00) | 0.20 (0.08–0.39) |
| DD score ≥3 | 0.65 (0.49–0.78) | 0.90 (0.73–0.98) | 0.37 (0.20–.56) |
| DD score ≥4 | 0.77 (0.64–0.89) | 0.73 (0.54–.88) | 0.80 (0.61–.92) |
| DD score ≥5 | 0.53 (0.39–0.68) | 0.07 (0.00–.22) | 1.0 |
| DD continuous score (1–5) | 0.79 (0.67–0.90) | N/A | N/A |
Conclusion
This study found that the threshold used to classify a person as having DD strongly influences the relationship between DD and LBP. Low thresholds for DD produced high sensitivity and low specificity whereas high thresholds for DD had low sensitivity and high specificity. When using the Pfirrmann et al scale, a threshold of ≥4 appears to be most strongly associated with the presence of LBP. Using a continuous measure of DD (1 to 5) produced very similar AUC to the dichotomized threshold of ≥4 and did not seem to be significantly more informative in this study. Rating individuals DD on their worst single spinal level or on summary scores from multiple levels produced very similar associations with the presence of LBP so neither approach can be recommended to be superior based on this study.
Yes
None declared
Pfirrmann CW, Metzdorf A, Zanetti M, Hodler J, Boos N. Magnetic resonance classification of lumbar intervertebral disc degeneration. Spine 2001;26:1873–1878