Test-retest reliability, validity, and minimum detectable change of visual analog, numerical rating, and verbal rating scales for measurement of osteoarthritic knee pain.

OBJECTIVE: Several scales are commonly used for assessing pain intensity. Among them, the numerical rating scale (NRS), visual analog scale (VAS), and verbal rating scale (VRS) are often used in clinical practice. However, no study has performed psychometric analyses of their reliability and validity in the measurement of osteoarthritic (OA) pain. Therefore, the present study examined the test-retest reliability, validity, and minimum detectable change (MDC) of the VAS, NRS, and VRS for the measurement of OA knee pain. In addition, the correlations of VAS, NRS, and VRS with demographic variables were evaluated.
METHODS: The study included 121 subjects (65 women, 56 men; aged 40-80 years) with OA of the knee. Test-retest reliability of the VAS, NRS, and VRS was assessed during two consecutive visits in a 24 h interval. The validity was tested using Pearson's correlation coefficients between the baseline scores of VAS, NRS, and VRS and the demographic variables (age, body mass index [BMI], sex, and OA grade). The standard error of measurement (SEM) and the MDC were calculated to assess statistically meaningful changes.
RESULTS: The intraclass correlation coefficients of the VAS, NRS, and VRS were 0.97, 0.95, and 0.93, respectively. VAS, NRS, and VRS were significantly related to demographic variables (age, BMI, sex, and OA grade). The SEM of VAS, NRS, and VRS was 0.03, 0.48, and 0.21, respectively. The MDC of VAS, NRS, and VRS was 0.08, 1.33, and 0.58, respectively.
CONCLUSION: All the three scales had excellent test-retest reliability. However, the VAS was the most reliable, with the smallest errors in the measurement of OA knee pain.

Introduction

Knee pain is the major symptom of OA among the elderly, affecting both men and women.1 Assessment of pain intensity is one of the primary outcomes used to determine the progression of OA.2–6 Several scales are commonly used for the assessment of pain intensity.7,8 Among them, the NRS, VAS, and VRS are often used in clinical practice.7 These pain-rating scales have shown good validity and reliability for assessing pain intensity; however, none has shown superiority over the others,7,8 as various aspects, such as the response categories, patient preference, application methods, and correction for missing information, make each of them unique.9

Because it is easy to understand and administer, the NRS is preferred over the VAS by the elderly population.8,10 For instance, the administration of VAS requires a patient to perceive his or her pain level mathematically, which could be particularly difficult11,12 for the elderly. In addition, the VAS has a higher reported failure rate of 7–16% when compared to the NRS and VRS.13–15 Although the NRS and VRS are well correlated with and sensitive for pain assessment, the NRS shows a higher reliability, specifically in elderly and less educated patients, and is useful for the assessment of chronic pain.16–18 Furthermore, a previous study showed higher responsiveness of the NRS compared to the VRS for the assessment of chronic pain.19

A previous study indicated low intra-scale agreement between the VAS and VRS, and recommended against their interchangeable use for musculoskeletal pain.20 Another study reported that the three scales are sensitive for the assessment of chronic OA pain, with no differences among them.21 Similarly, Bellamy et al22 reported that all scales for pain measurement are able to detect statistically significant and clinically important improvements in OA pain following a pharmacological intervention. However, these studies did not perform psychometric analyses of their reliability and validity for the assessment of OA pain. Therefore, the present study examined the test–retest reliability and MDC of the VAS, NRS, and VRS for the measurement of OA knee pain.

Methods

Ethics approval

Ethical approval was obtained from the IRB, Rehabilitation Research Chair, King Saud University. In accordance with the IRB guidelines of the institution, written informed consent was obtained from each individual who agreed to participate.

Participants

The participants (n = 121) included 65 women and 56 men aged 40–80 years with OA of the knee. The severity of OA of the knee was assessed using the K–L scale.23 Patients with any history of neurological disease, neuropathic pain, inflammatory joint disease, diabetes, or inability to cooperate were excluded.

Pain measures

The VAS is a reliable, valid, responsive, and frequently used pain outcome measure.7 It consists of a bidirectional 10 cm straight line with two labels, that is, “no pain” and “worst possible pain”, located at either end of the line. Patients are instructed to draw a vertical mark on the line indicating their pain level.7 The NRS is an 11-point scale comprising a number from 0 through 10; 0 indicates “no pain”, and 10 indicates the “worst imaginable pain”. Patients are instructed to choose a single number from the scale that best indicates their level of pain.7 The VRS is a valid scale consisting of a list of descriptors used to represent various levels of pain, including none, mild, moderate, and severe.8,24 Patients are instructed to select the one descriptor that best indicates their pain level.

Procedure

Brief instructions about the procedure were given to the patients before they completed the scales. Demographic information such as sex, age, weight, height, BMI, and the severity of knee OA was recorded. Patients were asked to rate their baseline level of pain on the VAS,25,26 NRS,25,26 and VRS25 in a random order. Patients were then requested to complete a second round of pain assessment using the VAS, NRS, and VRS after 24 h to determine test–retest reliability. Two independent examiners were involved in both sessions.

Statistical analysis

SPSS (Statistical Package for the Social Sciences) for Windows (version 22; IBM Inc., Armonk, NY, USA) was used for data analysis. Using the Shapiro–Wilk test, it was determined that the data were not normally distributed (P < 0.05). The test–retest reliability of VAS, NRS, and VRS was assessed using ICC2,1. The Bland–Altman plot method was used to assess the agreement between two readings (Figures 1 and 2). The validity was tested using Pearson’s correlation coefficients between the baseline scores of VAS, NRS, and VRS and the demographic variables (age, BMI, sex, and OA grade). The SEM and MDC were calculated based on the results of reliability analyses.27–29 For every test, the level of significance was set at P < 0.05 with a 95% confidence interval.

Figure 1

Bland–Altman plot: intraindividual differences (n = 121) between the visual analog scale (VAS) on test and retest, plotted against the average of the two scores.

Note: The central line represents the mean difference, and the dashed lines display the 95% limit of agreement.

Figure 2

Bland–Altman plot: intraindividual differences (n = 121) between the numerical rating scale (NRS) on test and retest, plotted against the average of the two scores.

Note: The central line represents the mean difference, and the dashed lines display the 95% limit of agreement.

Results

Table 1 presents the characteristics of the participants. The average age was 52.9 years. There were 65 women and 56 men. Most of the participants had a K–L grade of 2–3 (75.3%). Baseline VAS, NRS, and VRS scores are presented in Table 2.

Table 1

Participants’ characteristics

Characteristics	All participants
Gender, n (%)
Male	56 (46.3)
Female	65 (53.7)
Age (years)
Mean (SD)	52.9 (12.5)
Range	40–80
Height (m)
Mean (SD)	1.7 (0.06)
Range	1.5–1.9
Weight (kg)
Mean (SD)	86.4 (13.9)
Range	56–123
BMI (kg/m2)
Mean (SD)	30.01 (4.2)
Range	21.4–41.4
K–L scale score, n (%)
Grade 1	19 (15.7)
Grade 2	55 (45.5)
Grade 3	36 (29.8)
Grade 4	11 (9.1)

Abbreviations: BMI, body mass index; K–L, Kellgren and Lawrence.

Table 2

Descriptive statistics of baseline scores

Variables	All participants
VAS (0–10 cm)
Mean (SD)	5.8 (2.2)
Range	0.5–10
NRS (0–10)
Mean (SD)	5.9 (2.2)
Range	1–10
VRS n [%]
Mild	30 (24.8)
Moderate	41 (33.9)
Severe	50 (41.3)

Abbreviations: VAS, visual analog scale; NRS, numerical rating scale; VRS, verbal rating scale.

Test–retest reliability, SEM, and MDC

The reliability data are presented in Table 3. The ICC of the VAS, NRS, and VRS was 0.97, 0.95, and 0.93, respectively. The Bland–Altman plot showed reasonable agreement between the test–retest scores of the VAS and NRS (Figures 1 and 2, respectively). The SEM of VAS, NRS, and VRS was 0.03, 0.48, and 0.21, respectively. The MDC of VAS, NRS, and VRS was 0.08, 1.33, and 0.58, respectively (Table 3).

Table 3

Reliability data of VAS, NRS, and VRS

	ICC (95% CI)	SEM	MDC
VAS	0.97 (0.96–0.98)	0.03	0.08
NRS	0.95 (0.93–0.96)	0.48	1.33
VRS	0.93 (0.90–0.95)	0.21	0.58

Abbreviations: VAS, visual analog scale; NRS, numerical rating scale; VRS, verbal rating scale; ICC, intraclass correlation coefficient; CI, confidence interval; SEM, standard error of measurement; MDC, minimal detectable change.

Validity

A good-to-excellent correlation was found between the VAS and NRS (r = 0.941), VAS and VRS (r = 0.878), and NRS and VRS scores (r = 0.925). Age, BMI, and OA grades were significantly correlated with all the three scales (Table 4). The sex of the participants did not correlate with any scale (Table 4).

Table 4

Correlation of VAS, NRS, and VRS with demographic variables

	VAS	NRS	VRS
Age	0.262*	0.224*	0.261*
BMI	0.379**	0.359**	0.399**
Gender	0.071	0.056	0.048
K–L grade	0.844**	0.817**	0.754**
NRS	0.941**	–
VRS	0.878**	0.925**	–

Notes:

Correlation, P < 0.05.

Correlation, P < 0.001.

Abbreviations: VAS, visual analog scale; NRS, numerical rating scale; VRS, verbal rating scale; BMI, body mass index; K–L, Kellgren and Lawrence.

Discussion

The present study assessed the relative and absolute reliability, validity, and MDC of the three pain scales (VAS, NRS, and VRS) for the measurement of OA knee pain. The relative reliability, measured by the ICC, was excellent for all the three scales. The VAS was the most reliable (ICC = 0.97), and the VRS was the least reliable (ICC = 0.93). To the best of our knowledge, no prior study has assessed the reliability of VAS, NRS, and VRS for the measurement of OA knee pain; thus, a direct comparison of the present findings with those from other studies could not be made. One study reported excellent reliability for VAS and NRS (0.97 and 0.99, respectively) for the measurement of musculoskeletal pain.16 Ferraz et al30 reported that the three scales had excellent reliability when used in literate and illiterate groups of patients with rheumatoid arthritis. However, in the former study, the sample comprised a group of less educated orthopedic patients. In contrast, the present study did not consider the level of education of the patients with OA of the knee.

A few previous studies compared different pain-rating scales in patients with OA of the knee.21,22,31 Bashir et al21 reported good sensitivity of VAS, NRS, and VRS for assessment of pain in chronic OA; however, there were no significant differences. In addition, Bellamy et al22 reported that all measures of pain were able to detect statistically significant and clinically important improvements in OA pain following a pharmacological intervention. However, these studies did not report the validity and reliability of these three scales in patients with OA of the knee.

Although ICC provides a good measure of reproducibility, it was found to be sensitive to heterogeneity in the sample studied.32 On the contrary, SEM provides a more reliable and practical indicator, and helps calculate the MDC.32 In the current study, among the three scales, the VAS had minimal measurement error (SEM = 0.03). The MDC of the NRS (MDC = 1.33 points) was higher when compared to that of the VAS and VRS. In a previous study, Childs et al33 reported a slightly higher MDC value (MDC = 2 points) for NRS in patients with low-back pain. However, to our knowledge, no study has investigated the SEM and MDC for VAS, NRS, and VRS in patients with OA of the knee.

A good-to-excellent correlation was observed for VAS, NRS, and VRS, supporting the consistency of pain measurements among the three scales in patients with OA of the knee. These findings indicate that all the three scales are valid in assessing pain levels in patients with OA of the knee. Previous studies indicated a high correlation between VAS and NRS in patients with rheumatoid arthritis and chronic pain.29,34 In addition, the current study indicated a good correlation between VAS and VRS (r = 0.878). Similarly, Bolognese et al35 and Averbuch and Katzper36 reported a good correlation between the VAS and categorical pain scales in OA of the knee. Moreover, in the current study, VAS, NRS, and VRS correlated well with the radiographic severity of OA of the knee as measured using the K–L grading scale. This finding supports the predictive ability of these three scales to measure the radiographic severity of OA of the knee, but this needs to be confirmed by a future study.

The current study has some potential limitations that need to be considered while interpreting these findings. The study results were limited only to patients with pain due to OA of the knee; therefore, the application of these scales to other causes of musculoskeletal pain requires some caution. In the current study, the participants had a history of pain for more than 6 months, indicating chronic pain. Thus, additional research is required to assess the use of these three scales in populations with different stages of pain. Furthermore, in the current study, patients’ education level was not assessed. It is assumed that the educational status of the patients might affect pain perception; therefore, the educational background should be assessed for the effect on these scales.

Conclusion

All the three scales showed excellent test–retest reliability. A good-to-excellent correlation was found among the three scales. The VAS, NRS, and VRS were significantly correlated with the demographic variables. However, the VAS was considered the most stable, with the smallest error in measurement and MDC values for OA knee pain. Long-term follow-up studies are required to determine the responsiveness of these three scales in patients with OA of the knee.