Reliability of a new scale for essential tremor

Abstract

Objective

To determine the reliability of a new scale for the clinical assessment of essential tremor.

Background

The Essential Tremor Rating Assessment Scale contains nine performance items that rate action tremor in the head, face, voice, limbs and trunk 0–4 in half-point intervals. Head and limb tremor ratings are defined by specific amplitude ranges in centimeters.

Design/Methods

Videos of 44 patients and 6 controls were rated by 10 specialists on two occasions, 1–2 months apart. Inter- and intra-rater reliabilities were assessed with a two-way random effects intraclass correlation, using an absolute agreement definition.

Results

The inter- and intra-rater intraclass correlations for head and upper limb tremor ranged from 0.86 to 0.96, and the intraclass correlations for the total score were 0.94 and 0.96. The intraclass correlations for voice, face, trunk and leg were less robust.

Conclusions

This scale is an exceptionally reliable tool for the clinical assessment of essential tremor.

Introduction

Before 1993, there were no rating scales for essential tremor (ET) with documented reliability and validity. Many different ad hoc scales had been used in numerous small clinical trials of essential tremor, making the results of these trials difficult or impossible to compare.¹

The Fahn-Tolosa-Marín scale has been used extensively in clinical trials of ET, at times with modifications to enhance relevance to ET.² There is still no comprehensive item-by-item analysis of this scale, and inter-rater reliability has been fair to poor, especially for writing and drawing.^3–5 Furthermore, severe extremity tremor is defined as > 4 cm, which is much less than advanced ET.⁶

The tremor rating scales of Bain and coworkers assess tremor in the head, limbs, Archimedes spirals, and handwriting using 0–10 ratings.^{7, 8} Ratings are defined subjectively or by examples in a published manual, making good reliability difficult to achieve.^{7, 8}

The original tremor rating scale of Louis and coworkers⁹ was subsequently revised “to broaden the applicability of this scale to clinical trials”.¹⁰ This scale measures only upper extremity tremor and requires video training to achieve high inter-rater reliability.¹⁰

Recognizing the limitations of existing scales, the Tremor Research Group (TRG) met several times over a period of nine years to develop the TRG Essential Tremor Rating Assessment Scale (TETRAS) (Appendix 1). TETRAS has a 12-item activities of daily living (ADL) subscale that addresses many of the activities assessed in the ADL scales of Fahn,⁶ Louis,¹¹ Bain,⁷ and their coworkers, and TETRAS also has a 9-item performance subscale that quantifies tremor in the head, face, voice, limbs and trunk. TETRAS was developed with three mandates: 1) rapid clinical assessment of ET, 2) no equipment other than pen and paper, and 3) objective metric anchors for each 0–4 rating whenever possible, so as to reduce experiential rating bias and uncertainty. The performance subscale takes less than 10 minutes. Ratings of head tremor and limb tremor are defined by specific amplitude ranges in centimeters (Table 1), so raters must first estimate the maximum amplitude of tremor and then assign the corresponding rating. The performance subscale allows 0.5-point increments in scoring. The 0.5-point increments in upper limb tremor ratings are defined by specific ranges of tremor amplitude. For all other items, the 0.5 increments may be used when the 0–4 integer rating is uncertain.

Table.

Metric anchors for TETRAS performance measures of head and extremity tremor

Head tremor	Upper limb tremor	Lower limb tremor
0 = no tremor	0 = no tremor	0 = no tremor
1 = < 0.5 cm	1 = barely visible	1 = barely visible
2 = 0.5– < 2.5 cm	1.5 = < 1 cm	2 = <1
3 = 2.5–5 cm	2 = 1– < 3 cm	3 = 1 – 5
4 = > 5 cm	2.5 = 3– < 5 cm	4 = > 5 cm
	3 = 5– < 10 cm
	3.5 = 10– < 20 cm
	4 = > 20 cm

In a preliminary study, 10 members of TRG simultaneously rated 10 ET patients during the live administration of the TETRAS performance subscale. Excellent inter-rater reliabilities were found for head and upper limb tremor.¹² In another preliminary study, ten members of TRG simultaneously rated three patients with mild, moderate and severe ET and rated the videos of these exams one month later. The correlations between video scores and live exam scores were greater than 0.87 for all items except face (0.67) and voice (0.63) tremor.¹³ We now estimate the inter- and intra-rater reliabilities of the TETRAS performance subscale and the correlation of this subscale with the ADL subscale, using 50 videotaped exams.

Methods

All studies were performed with the signed written informed consent of the patients and controls, approved by the institutional review board of each institution. Patients with ET and controls with no history of tremor were recruited from the authors’ clinics. The patients were diagnosed using Tremor Investigation Group criteria.¹⁴

Nine of the authors, all movement disorder specialists, videotaped TETRAS exams of one control and at least four patients. Each specialist was asked to video patients with mild, moderate and severe ET so that the videos were evenly distributed over these levels of severity. The TETRAS ADL and performance subscales were performed during each videotaping. Fifty videos (44 patients and 6 controls) were compiled in random order to a set of DVDs and mailed to the same specialists and one other. Each specialist rated all 50 videos. The same videos in different order were rated by the same specialists one to two months after the first rating.

Due to omissions in some of the videos, some of the test items could not be scored for every video. Four of the ten video raters had no experience or training in TETRAS, and nearly all video omissions came from these four raters. Nevertheless, all raters scored 31 to 46 videos for each item, except the standing item for which only 19 were scored.

Inter- and intra-rater reliability of the performance subscale were assessed with two-way random effects intraclass correlations (ICC), using an absolute agreement definition.

Results

The patients (mean 67, range 35–80) and controls (mean 50; range 27–82) had comparable ages, and 27 of 50 participants were men. The average duration of tremor in the 44 patients was 30 years (range 6–72). The distribution of total TETRAS performance scores for the 50 participants was fairly uniform (Appendix 2).

The inter- and intra-rater ICCs were greater than 0.85 for all items except face tremor, voice tremor, lower limb tremor and trunk (standing) tremor (see Appendices 3 and 4 for details). The six experienced and four inexperienced raters did not differ statistically (repeated measures ANOVA) in their inter- and intra-rater reliabilities for any of the test items, but the biggest differences were for the face, lower extremity and trunk (standing) items.

The raters performed live TETRAS assessments during the videotaping of the 50 TETRAS exams. The Pearson correlation between the total ADL scores and the total performance scores was 0.887 (p < 0.0001) (Figure).

Figure.

Total ADL and performance scores for the 50 patients and controls

Cronbach alpha for the live exams performed during the video tapings was 0.951, and it was 0.968 after removal of the face, lower limb and trunk items. Statistically identical results were obtained when Cronbach alpha was computed using the video ratings of each rater. The item-to-total score correlations ranged from 0.88 to 0.95 (mean 0.91) for all upper limb items except right upper limb postural tremor with the limb extended forward (0.75). Item-to-total correlations for the head (0.69), face (0.45), voice (0.68), lower limb (0.60) and trunk (0.46) were lower.

Discussion

Our use of performance ratings defined in terms of specific amplitude ranges (cm) resulted in exceptional inter- and intra-rater reliabilities, even for raters without prior experience or training with this scale. However, all raters were experienced movement disorder specialists, and it remains to be determined if raters with less expertise perform as well. Comparable inter-rater reliabilities were found in our earlier preliminary study in which ten TRG members, all developers of TETRAS, simultaneously rated ten ET patients during the live administration of the TETRAS performance subscale.¹²

The TETRAS ADL and performance subscales have obvious content validity for ET, and there is also evidence of strong construct validity. The TETRAS ADL and performance scores are highly correlated, and the TETRAS performance items of upper extremity function correlate strongly with transducer measures of upper limb tremor.^15–17

TETRAS was designed specifically for ET and has not been tested in children or other patient populations. Its metric anchors are too large for small children. TETRAS does not include an assessment of rest tremor because rest tremor is usually not present in ET.¹⁸ Furthermore, distinguishing rest tremor in ET from postural tremor in the presence of incomplete relaxation is difficult and is best avoided when the principal goal is assessment of ET severity rather than the diagnosis or complete characterization of ET in a particular patient.

TETRAS is heavily weighted by upper extremity tremor and is arguably not ideal when the principal interest is tremor elsewhere. The least reliable items of the TETRAS performance subscale were face tremor, lower limb tremor and trunk tremor while standing. This has been the experience with other tremor scales.⁷ These items could be deleted with little loss of content validity since the current clinical definition of ET focuses on upper extremity tremor and head tremor.¹⁴ Exclusion of the face, lower limb and trunk (standing) items would reduce the maximum total performance score from 64 to 52.

Finally, most of us thought that the anchor “barely perceptible” for grade 1 upper and lower extremity tremor could be improved by adding “< 0.5 cm” or “a few mm”. Most also thought that lower extremity tremor was difficult to distinguish from other irregularities of movement and posture. Some raters performed heel-knee-shin testing with the patient seated, and others performed this test with the patient supine, which seems preferable for the specific assessment of tremor. Training would improve consistency and quality of administering the lower extremity exam, but it is unclear whether training can reduce the uncertainty in rating tremor.

In conclusion, TETRAS was designed specifically for the clinical measurement of ET severity, requiring no instruments other than a pen and paper. The scale is brief, valid and highly reliable, and we believe it is ideal for ET clinical trials.