Elsevier

The Spine Journal

Volume 8, Issue 6, November–December 2008, Pages 968-974
The Spine Journal

Clinical Study
Minimum clinically important difference in lumbar spine surgery patients: a choice of methods using the Oswestry Disability Index, Medical Outcomes Study questionnaire Short Form 36, and Pain Scales

https://doi.org/10.1016/j.spinee.2007.11.006Get rights and content

Abstract

Background context

The impact of lumbar spinal surgery is commonly evaluated with three patient-reported outcome measures: Oswestry Disability Index (ODI), the physical component summary (PCS) of the Short Form of the Medical Outcomes Study (SF-36), and pain scales. A minimum clinically important difference (MCID) is a threshold used to measure the effect of clinical treatments. Variable threshold values have been proposed as MCID for those instruments despite a lack of agreement on the optimal MCID calculation method.

Purpose

This study has three purposes. First, to illustrate the range of values obtained by common anchor-based and distribution-based methods to calculate MCID. Second, to determine a statistically sound and clinically meaningful MCID for ODI, PCS, back pain scale, and leg pain scale in lumbar spine surgery patients. Third, to compare the discriminative ability of two anchors: a global health assessment and a rating of satisfaction with the results of the surgery.

Study design

This study is a review of prospectively collected patient-reported outcomes data.

Patient sample

A total of 454 patients from a large database of surgeries performed by the Lumbar Spine Study Group with a 1-year follow-up on either ODI or PCS were included in the study.

Outcome measures

Preoperative and 1-year postoperative scores for ODI, PCS, back pain scale, leg pain scale, health transition item (HTI) of the SF-36, and Satisfaction with Results scales.

Methods

ODI, SF-36, and pain scales were administered before and 1 year after spinal surgery. Several candidate MCID calculation methods were applied to the data and the resulting values were compared. The HTI of the SF-36 was used as the anchor and compared with a second anchor (Satisfaction with Results scale).

Results

Potential MCID calculations yielded a range of values: fivefold for ODI, PCS, and leg pain, 10-fold for back pain. Threshold values obtained with the two anchors were very similar.

Conclusions

The minimum detectable change (MDC) appears as a statistically and clinically appropriate MCID value. MCID values in this sample were 12.8 points for ODI, 4.9 points for PCS, 1.2 points for back pain, and 1.6 points for leg pain.

Introduction

Spinal surgery studies rely on patient's answers to health-related quality-of-life questionnaires (HRQOL) to assess treatment effect. Three commonly used HRQOLs are pain scales, the Oswestry Disability Index (ODI) [1], [2], [3], and the physical component summary (PCS) of the Short Form of the Medical Outcomes Study (SF-36) [4]. A difficulty in using HRQOL measures is that their numerical scores lack immediate meaning or clinical importance [5]. The concept of minimum clinically important difference (MCID) has been proposed as a critical threshold to measure treatment effectiveness. Treatment effect reaching MCID would impart significance to study results and might justify clinicians' decision to treat. The original conceptualization of MCID included an implicit ratio of benefits to costs and risks of treatment [6] but research in the field of MCID has mainly focused on the benefit side of this ratio and most commonly considers MCID to be “the smallest change that is important to patients” [7].

MCID values are instrument dependent. Efforts have been made to determine MCID for a variety of instruments such as the Roland-Morris Back Pain Questionnaire[7], [8], the Chronic Heart Failure Questionnaire [6], [9], the Chronic Respiratory Disease Questionnaire [6], [10], the SF-36 [9], [10], Visual Analog pain scales [11], and ODI [11]. Several methods have been used to calculate MCID; those methods are usually classified into anchor-based methods (comparing HRQOL scores to another measurement) and distribution-based methods (built on the variability of the HRQOL scores). A major limitation in the pursuit of MCID stems from the fact that the different calculation methods produce different MCID values [5], [12]. There is, so far, no agreed upon method to calculate an optimal MCID and no definite MCID has been established for the three commonly used HRQOL scales in spinal surgery patients: pain scales, ODI, and SF-36 (PCS).

The choice of a calculation method and, hence, of an MCID value has important implications for the evaluation of treatment. Obviously, a lower MCID value would put any given treatment in a more favorable light, because a greater proportion of patients would be able to reach a lower threshold. On the other hand, an artificially high MCID would not appropriately credit the value of the intervention. No study so far has examined the range of possible MCID values in lumbar spine surgery patients that can be obtained by different calculation methods.

This study has three purposes. First, to illustrate the variability of values obtained by the common anchor-based and distribution-based methods to calculate MCID [13]. We purposely adhered to calculation methods as found in the MCID literature without attempt to revise those methods. Second, to determine an MCID value for each HRQOL that is both statistically sound and clinically meaningful for patients undergoing spinal surgery. Although it is possible that MCID values may differ by treatment and pathology, we aimed to describe MCID values for a general lumbar spine surgery population, encompassing a broad range of diagnostic entities. Third, to validate the discriminative ability of a global health assessment as anchor by comparing it with another anchor, specifically, a scale based on the patient satisfaction with the outcomes of their surgery.

Section snippets

Patient sample

Physicians from the Lumbar Spine Study Group collaborate in prospectively collecting outcomes data on their patients. The Lumbar Spine Study Group database contained 948 patients; of the 948 patients, 460 were retained for this study because they had answered the “satisfaction with results” questions. These patients were contributed by 11 different surgeons at three medical centers. The following numbers of patients had both baseline and 1-year scores: 457 for ODI, 460 for PCS, 427 for back

Baseline, 1-year, and change scores

Table 1 reports baseline and 1-year scores for ODI, PCS, and pain scales. The change scores broken down by the patient answers to the HTI and the Satisfaction scales are reported in Table 2. Analysis of variance and Bonferroni post hoc tests showed that the average change scores are different between HTI and Satisfaction answers except for the following: the ODI score of the “Somewhat Worse” is not different from the “About the same” and the “Much Worse.” The ODI score of the “Dissatisfied” is

Choice of MCID value

In this study, we evaluated a range of statistical parameters with potential relevance in identifying an optimal MCID value for each of the HRQOL measures studied. It is clear that the different calculation methods yielded widely different threshold values. These ranged from 2.92 to 15.36 for ODI, 1.26 to 5.95 for PCS, 0.28 to 2.88 for back pain, and 0.53 to 2.85 for leg pain. These constitute five- to 10-fold variations. The smallest value consistently resulted from the effect size

Conclusions

MCID is a quantitative assessment of clinical relevance for a given magnitude of HRQOL score change. HRQOL scores above MCID would be indicative of a potentially important change. A given treatment would then be justified when a large proportion of patients report changes above MCID. This study determined an MCID value for the four common HRQOL measures used for the surgical treatment of lumbar spine patients. Amongst a variety of threshold values, the MDC was selected as an appropriate MCID

Acknowledgement

The authors thank the members of the Lumbar Spine Study Group for their support.

References (29)

  • J.E. Ware

    SF-36 health survey update

    Spine

    (2000)
  • P.W. Stratford et al.

    Sensitivity to change of the Roland-Morris Back Pain Questionnaire: Part 1

    Phys Ther

    (1998)
  • D.L. Riddle et al.

    Sensitivity to change of the Roland-Morris Back Pain Questionnaire: Part 2

    Phys Ther

    (1998)
  • K.W. Wyrwich et al.

    Linking clinical relevance and statistical significance in evaluating intra-individual changes in health-related quality of life

    Med Care

    (1999)
  • Cited by (880)

    View all citing articles on Scopus

    FDA device/drug status: not applicable.

    The authors acknowledge a financial relationship (research support from Medtronic [AGC, BRS] and consultants [BRS, TCS, and SDG] for Medtronic Sofamor Danek) that may indirectly relate to the subject of this research.

    View full text