Clinically significant changes in pain along the Pain Intensity Numerical Rating Scale in patients with chronic low back pain

Low back pain (LBP) is the most common cause of chronic pain. Numerous clinical scales are available for evaluating pain, but their objective criteria in the management of LBP patients remain unclear. This study aimed to determine an objective cutoff value for a change in the Pain Intensity Numerical Rating Scale (ΔPI-NRS) three months after LBP treatment. Its utility was compared with changes in six commonly used clinical scales in LBP patients: Pain Disability Assessment Scale (PDAS), Pain Self-Efficacy Questionnaire (PSEC), Pain Catastrophizing Scale (PCS), Athens Insomnia Scale (AIS), EuroQoL 5 Dimension (EQ5D), and Locomo 25. We included 161 LBP patients treated in two representative pain management centers. Patients were partitioned into two groups based on patient’s global impression of change (PGIC) three months after treatment: satisfied (PGIC = 1, 2) and unsatisfied (3–7). Multivariate logistic regression analysis was performed to explore relevant scales in distinguishing the two groups. We found ΔPI-NRS to be most closely associated with PGIC status regardless of pre-treatment pain intensity, followed by ΔEQ5D, ΔPDAS, ΔPSEC, and ΔPCS. The ΔPI-NRS cutoff value for distinguishing the PGIC status was determined by ROC analysis to be 1.3–1.8 depending on pre-treatment PI-NRS, which was rounded up to ΔPI-NRS = 2 for general use. Spearman’s correlation coefficient revealed close relationships between ΔPI-NRS and the six other clinical scales. Therefore, we determined cutoff values of these scales in distinguishing the status of ΔPI-NRS≥2 vs. ΔPI-NRS<2 to be as follows: ΔPDAS, 6.71; ΔPSEC, 6.48; ΔPCS, 6.48; ΔAIS, 1.91; ΔEQ5D, 0.08; and ΔLocomo 25, 9.31. These can be used as definitive indicator of therapeutic outcome in the management of chronic LBP patients.

Introduction study was approved by the Institutional Review Boards of Yamaguchi University (H28-1351) and in Aichi Medical School (No.12-067). All participants provided written informed consent.

Patient care
Therapeutic regimens for CLBP patients were provided based on the Pain Management Program (PMP). PMPs are rehabilitation-based multidisciplinary programs for people with chronic pain [22,23]. They involve a group of clinicians, nurses, physiotherapists, and psychologists led by pain medicine specialists who collaborate in assisting patients to bring their pain under control. The patients attended PMPs after several therapies had met with limited success. The team of experienced health care professionals at each hospital worked closely with the patients, and, in general, the PMPs were tailored to the individual patient's clinical needs. The team collaborated in the rehabilitation by providing exercise plans, pharmacotherapy, psychotherapy, cognitive behavioral therapy, patient education, nerve blocks guided by ultrasound or X-ray, and radiofrequency nerve ablation to promote the relief of pain in the rehabilitation program [21].  Table 1. Inclusion and exclusion criteria for enrollment of patients with chronic LBP.

Inclusion criteria Exclusion criteria
Age 20-85 years Inability to understand and read the Japanese language, drug abuse, dementia, or other reasons to suspect poor adherence to follow-up

Scales of multifaceted clinical assessment
For multifaceted clinical assessment of the patients' conditions, we used the following seven well-established scales: Pain Disability Assessment Scale (PDAS): A scale for measuring lifestyle disabilities of chronic pain patients. Higher scores (on a scale of 0 to 60 points) indicate greater degrees of lifestyle disability [15].
The Hospital Anxiety and Depression Scale (HADS): A self-reported instrument used to evaluate depression and anxiety in clinical research. The HADS has advantages over other such assessments in that it is efficient in assessing both anxiety and depression. It is composed of 14 questionnaire items and was originally developed for a general medical rather than psychiatric field. Higher scores (0 to 21 points both for anxiety and depression) indicate greater degrees of anxiety and depression [21].
Pain Self-Efficacy Questionnaire (PSEQ): PSEQ is a 10-item questionnaire developed to assess the confidence of patients with ongoing pain in performing daily activities while in pain. The PSEQ is applicable to any type of persisting pain. It covers a range of functions, including household chores, socializing, work, and coping with pain without medication. Scores can range from 0 to 60, and higher scores indicate greater degrees of performing activities while in pain [11].
Athens Insomnia Scale (AIS): This scale assesses the severity of insomnia using diagnostic criteria set forth by the International Classification of Diseases (ICD-10). The eight-item questionnaire evaluates sleep onset, night and early-morning waking, sleep time, sleep quality, frequency and duration of complaints, distress caused by the experience of insomnia, and interference with daily functioning [13].
Locomo 25: This was developed as a screening tool for locomotive syndrome by a Japanese orthopedic surgeon group in 2008. It consists of 25 questions aimed at musculoskeletal disorders such as walking disability, difficulty in daily living, or suffering pain within the body. Scores can range from 0 to 100, and higher scores indicate a greater degrees of performing activities while in pain [10].

Measurements
Before and 3 months after the therapy, a PI-NRS score reported on a scale of 0−10 was obtained as an indicator of the average level of pain over the past 7 days [22]. For multi-faceted assessments of the chronic LBP patients, the PDAS, HADS, PCS, PSEQ, and AIS were administered twice, once before and once after the therapy, together with PI-NRS. We also administered the Locomo 25 and EQ-5D to assess physical functions (Fig 1).

Patient's global impression of change
To detect clinically relevant changes in the PGIC, the concept of the transition method was used [17][18][19]. The transition questionnaire investigates current pain intensity compared to that before treatment. The PGIC was administered at the time of the 3-month follow-up, and patients were asked to score the change on the following scale: (1) much improved, (2) improved, (3) slightly improved, (4) no change, (5) slightly worse, (6) worse, or (7) much worse. [5,14,28] 1.6. Statistical analyses for assessment of clinical utility of ΔPI-NRS and changes in other clinical scales 1.6.1. MCID and cutoff value for ΔPI-NRS. The post-treatment change in PI-NRS, or ΔPI-NRS, was calculated as (PI-NRS follow-up − PI-NRS baseline). We sought to predict the MCID and cutoff value for ΔPI-NRS using the anchor-based approach by setting PGIC as the anchor [23]. Therefore, we partitioned LBP patients into two groups based on post-treatment PGIC score: Group 1: satisfied (PGIC = 1 or 2) vs. Group 2: not satisfied (PGIC = 3-7) as was done in previous studies [19,[24][25]. The utility of ΔPI-NRS in distinguishing the two groups were evaluated from satisfied (PGIC = 1 or 2) vs. not satisfied (PGIC = 3-7).
The effect ratio of observed between-group differences was defined as the difference in averages of scale changes (ΔS) or ΔPI-NRS for Group 1 and Group 2, DS G1 ; DS G2 , divided by the average within-group variations of ΔS, or SD (ΔS), as shown in the formula: Based on Cohen's criteria [26], the MCID for ΔPI-NRS was defined as half of the average inherent variability of ΔPI-NRS or 0.5×SD (ΔPI-NRS) [26][27][28].
Therefore, if DPI � NRS G1 À DPI � NRS G2 > 0:5 � SDðDPI � NRSÞ, we regarded ΔPI-NRS as having clinical utility and proceeded to determine its cutoff value at which the sensitivity of detecting satisfied cases is equal to the specificity of detecting non-satisfied cases. The magnitude of the clinical utility of ΔPI-NRS was expressed as an area under the curve (AUC) of a receiver operating characteristic (ROC) curve.

Results
Patient characteristics and profiles of 12 clinical scales before and three months after treatment are respectively presented in Tables 2 and 3, separated by mild, moderate, and severe pain groups. The post-treatment levels of patient's satisfaction by PGIC score are shown in Table 4 in relation to ΔPI-NRS.

Determination of cutoff values and MCID for post-treatment ΔPI-NRS
LBP patients were partitioned into two groups in reference to post-treatment PGIC: satisfied (PGIC = 1-2) and non-satisfied (PGIC = 3-7) groups. The utility of the clinical scales in distinguishing the status of satisfaction by PGIC was explored by the use of multivariate logistic regression analysis in three ways according to the pre-treatment level of pain by PI-NRS: groups of patients with moderate, severe, and moderate + severe pain ( Table 5). The final list of clinical scales that were found significant in predicting PGIC satisfaction status as determined by the stepwise selection method are listed in each table.
For the analysis targeting the moderate and moderate + severe pain groups, ΔPI-NRS was the leading predictor of satisfaction status with P values of 0.00064 and 0.0001, respectively. Other significant predictors among the clinical scales were ΔEQ5D-Q2, ΔHADS-anxiety, and ΔAIS in the analysis of the moderate pain group and ΔEQ5D-Q2, ΔHADS-anxiety, ΔHADSdepr, and ΔAIS in the analysis of the moderate + severe pain group. In the analysis targeting the severe pretreatment pain group, ΔPI-NRS, ΔEQ5D-Q1, and ΔLocomo 25 showed nearly equal contribution in predicting a satisfactory status.
Post-treatment changes (Δ) in 8 major clinical scales were compared between the two groups with and without satisfaction by PGIC. The degree of separation of the two groups by each clinical scale was evaluated by ROC analyses (Fig 2). The AUC was highest by ΔPI-NRS (AUC = 0.770), followed by ΔHADS-dep (AUC = 0.733), ΔEQ5D (AUC = 0.686), and ΔLocomo 25 (AUC = 0.648).
The optimal cutoff value was estimated as the ΔPI-NRS value at which sensitivity and specificity are equal. For the moderate, severe, and moderate + severe pain groups, the cutoff values for ΔPI-NRS were 1.3, 1.8, and 1.5, respectively (Fig 3). Because the PI-NRS takes an integer value between 0 to 10, these cutoff values can be rounded up to 2 for practical use regardless of the pre-treatment pain severity: i.e., when ΔPI-NRS �2, it is appropriate to consider that the patients felt much improved or improved (PGIC of 1 or 2). However, the MCID based on the inherent variability of ΔPI-NRS was 0.553 for the moderate pain group, 0.596 for the severe pain group, and 0.512 for the moderate + severe pain group (Fig 3). It is notable that the cutoff values shown above all exceeded the respective MCID values. As a whole, those MCID values for ΔPI-NRS can be rounded up to 1 for practical use regardless of the pretreatment level of pain.

Clinical utilities of post-treatment changes in the clinical scales.
We applied the same method as was done for ΔPI-NRS to calculate MCIDs and cutoff values of ΔPDAS, ΔPSEC, ΔPCS, ΔAIS, ΔEQ5D, and ΔLocomo 25. The anchor for this analysis was the status of ΔPI-NRS partitioned at its cutoff value: ΔPI-NRS�2 vs. ΔPI-NRS<2.
The MCID and cutoff value in predicting ΔPI-NRS�2 status were 4.67, 6.71 for ΔPDAS (score range: 0−60); 6.48, 6.48 for ΔPSEC (0−60); 5.05, 6.71 (0−52) for ΔPCS; 1.9, 1.64 for ΔAIS (0−24); 0.08, 0.053 for ΔEQ5D (0−1.0); and 7.5, 9.31 for ΔLocomo 25 (0−100) (Fig 4). These cutoff values exceeded the respective MCID except for ΔAIS and ΔEQ5D, indicating that a change in their score above the cutoff value can be interpreted as clinically relevant. The imbalance of the MCID and cutoff value observed for ΔAIS and ΔEQ5D were caused by unequal scatter of values between the two groups resulting in unreliable cutoff values. Therefore, their cutoff values were raised to the level of respective MCIDs to avoid false-positive interpretation of their changes.

Discussion
It is most important to set the objective goal in clinical after CLBP treatment because it is difficult that we make the patients with completely no pain. The patients with CLBP commonly have not only lumber dysfunction but also physical disability and psychosocial issues [1]. The quantification of the outcome of CLBP treatment using some simple clinical scale is crucical in the clinical management [12,15]. To reveal the cutoff points and the MCID for ΔPI-NRS based on PGIC in CLBP are the most simple and acceptable for the medical staff when we quantify the treatment goal [17][18][19][20][21][22][23][24][25].
We comprehensibly evaluated the physical disability and psychosocial conditions in patients with CLBP using self-administered multifaceted measures because these patients have multidimensional musculoskeletal, social, mental and cognitive disorders, and other issues [11-12, 21, 29, 30]. However, we cannot tell how much of an improvement in each score can be regarded as clinically meaningful [12]. The lack of objective cutoff values of improvement in these clinical scales restricts proper interpretation of the scores when conducting research into treatment outcomes in CLBP [12,31].
At our pain management centers, we evaluate the intensity of pain in patients with CLBP using the PI-NRS, PDAS, HADS, PCS, PSEQ, EQ-5D, AIS, and Locomo 25, which cover multifaceted issues [9,15,21,[32][33]. If these complex clinical scales can be interpreted in a unified way, evaluation of a patient becomes less difficult and more objective, and current efforts of the medical staff will be eased. In addition, the availability of objective cutoff values for these clinical scales for use as treatment goals can widely expand the number of clinical facilities capable of managing pain in CLBP patients.  When both clinicians and investigators calculate cutoff values for these measures, we think the patient's perspective on the meaning of changes following treatment will become core outcome measures [28,[34][35][36]. A commonly used method to determine thresholds for patientperceived meaningful change is to compare changes in pain scores with patients' global ratings of the magnitude of change [28,[37][38]. We thus focused on ΔPI-NRS as the most important assessment anchoring treatment because the threshold of change in PI-NRS, i.e., ΔPI-NRS, following treatment is generally considered to be quite useful for judging the effect of therapy, and the threshold is the universal indicator of change in individual patients [3-4, 16, 39]. In the present study, we sought to determine the cutoff points and the MCID for ΔPI-NRS based on PGIC in CLBP [18]. We revealed that the MCID of ΔPI-NRS was 1 regardless of the level of pre-treatment pain in CLBP patients, and this led to the determination of a treatment goal of 2 for ΔPI-NRS as the cutoff value in CLBP.
Several papers have reported cutoff values for chronic musculoskeletal pain [17,[39][40]. Salaffi et al. reported the MCID in chronic musculoskeletal pain intensity measured on a NRS [40]. A change in the NRS score of −2.0 and a percent change score of −33.0% were best associated with the concept of "much better" improvement. Although the database used in their report did not contain CLBP patients, they reported the important threshold score of 2 for NRS change, which was the same as our cutoff value of 2.
In the measurement of pain relief in patients with trigeminal neuralgia, Sandhu et al. reported that the MCIDs for the 3 domains of the Brief Pain Inventory-Facial were 57% and 28% improvement in pain intensity for the worst and average pain, respectively, 75% improvement in interference with general activities of daily living, and 62% improvement in interference with facial activities of daily living [39]. Another paper defined the MCID for grade I degenerative lumbar spondylolisthesis following lumbar surgery [41]. The MCID values were 1.6 points for NRS-back pain, 1.7 points for NRS-leg pain, 14.3 points for the Oswestry Disability Index, and 0.2 points for the EQ-5D. These previous papers reported similar scores for cutoff values and MCIDs even if the target disease was different from that in the present study. We thus think a ΔPI-NRS of 2 is the key threshold score in evaluating treatment outcome of CLBP patients.
As a next step, the cutoff values and the MCID were also determined for ΔPDAS, ΔPSEC, ΔPCS, ΔAIS, ΔEQ5D, and ΔLocomo 25 based on ΔPI-NRS (ΔPI-NRS�2 or ΔPI-NRS<2) in the present study. To our best knowledge, this is the first report to reveal a meaningful threshold value for the treatment goal and the cutoff value for treatment effect for each of these clinical scales. Only limited cutoff values were reported previously, and no reports showed the meaningful threshold change for treatment. Yamashiro et al. reported that the cutoff value for the PDAS was 10. A score of over 41 points in the PSEQ means the patient has high self-efficacy, whereas that below 20 points indicates low self-efficacy [15]. Sullivan et al. reported that the cutoff value in the PCS was 30 [14]. Four to five points in the AIS suggests a sleep disorder, and over 6 points suggest the high possibility of a sleep disorder. Over 16 in the Locomo 25 indicates locomotive syndrome [10]. We established objective and meaningful threshold scores for ΔPDAS of 6.71, ΔPSEC of 6.48, ΔPCS of 6.71, ΔAIS of 1.9, ΔEQ5D of 0.08, and ΔLocomo 25 of 9.31 in the treatment of CLBP patients.
From another point of view, when medical staff observe a ΔPI-NRS of �2 in patients after treatment, we can interpret this to indicate that the CLBP patients have gotten well with satisfaction and improvement in the multifaceted scores measuring life disabilities, patient confidence with ongoing pain, pain catastrophizing, sleep disorder, QOL, and musculoskeletal disorders.
In summary, as clinically significant threshold and treatment target in CLBP treatment, we revealed a ΔPI-NRS of �2 and the threshold scores for ΔPDAS of 6.71, ΔPSEC of 6.48, ΔPCS of 6.71, ΔAIS of 1.9, ΔEQ5D of 0.08, and ΔLocomo 25 of 9.31. We propose those definitive target scores as directly correlating with PGIC for use all of the medical staff in the management of CLBP patients.
A limitation of this study is that the score cannot be generalized to all CLBP patients because the data were taken only from a short 3-month follow-up period and from CLBP patients treated in only two pain management centers. Second, it is inevitable to have some degree of inconsistencies between MCID score and cutoff values in some clinical scales like AIS and EQ5D, although we derived MCID in the same way as in the previous reports possibly due to insufficient number of cases enrolled. In the present study we adopted either the MCID or the cutoff value with higher score as indicative of more clinical utility.
In conclusion, we revealed a new indicator in the evaluation of CLBP treatment. A ΔPI-NRS value of 2 is the key score in CLBP treatment.