Diagnostic Accuracy of Qualitative vs. Quantitative Tuning Forks: Outcome Measure for Neuropathy

Abstract

Objective

To assess the diagnostic accuracy of the Rydel-Seiffer vs. the qualitative 128 Hz tuning fork for detecting axonal neuropathy.

Methods

100 consecutive patients seen at the Neurology Outpatient Electromyography (EMG) clinic at a major academic center were recruited and consented for this study. Study personnel who were blinded to results of nerve conduction studies, collected data on vibratory perception with both tuning forks at bilateral (a) great toe, and (b) distal inter-phalangeal joint on the second digit. Published normative data was used to determine abnormal scores for the Rydel-Seiffer tuning fork and the qualitative tuning fork; axonal neuropathy was determined based on our EMG laboratory standards.

Results

A total of 296 nerves from 100 patients were tested with tuning forks and compared to sensory nerve action potential obtained through EMG. The sensitivity and specificity of the Rydel-Seiffer tuning fork was 29% and 86% respectively, and the sensitivity and specificity of the conventional 128 Hz tuning fork was 21% and 88% respectively. The extended McNemar’s test showed no significant difference in sensitivity or specificity of the two tuning forks (χ²=0.83).

Conclusions

There is no difference in diagnostic accuracy between the Rydel-Seiffer and conventional tuning fork for detecting axonal neuropathies; however, the Rydel-Seiffer is easier to use, and superior for longitudinal assessments.

Introduction

Sensation testing in standard neurological examination is usually qualitative and subjective.^1,2 The 128 Hz tuning fork is the most commonly clinically used tool to test for vibratory sensation; however, the graduated quantitative 64 Hz Rydel-Seiffer tuning fork, validated as a proxy-measure of axonal neuropathy,^3,4 is gaining popularity as an outcome measure in clinical trial settings.^5,6 To our knowledge, no study has assessed the diagnostic accuracy of the qualitative 128 Hz tuning fork vs. the quantitative Rydel-Seiffer tuning fork as a predictor of axonal neuropathy. The purpose of this study was to calculate the diagnostic accuracy of the Rydel-Seiffer tuning fork vs. the qualitative 128 Hz tuning fork in the detection of electrodiagnostically confirmed axonal peripheral neuropathy.

Methods and Materials

Study design and participants

This was a prospective, cross-sectional study. Approval was obtained from our Institutional Review Board committee to perform the study. Informed consent was obtained from 100 consecutive patients referred to the Neurology Outpatient Electromyography (EMG) clinic at a major academic center for this study. All willing participants were included in this study; there were no exclusion criteria. Demographic and clinical data were not individually collected: with this study design, the individual patient acted as both the test and the control for the two tuning forks; the age, gender and diagnosis would have a similar impact on both tuning forks, and therefore wash out any confounding effect.

Standard protocol approvals, registrations, and patient consents

After providing written informed consent, patients were examined prior to the nerve conduction studies (Figure 1). The study personnel performing the physical exam were blinded to any medical history of the patient, and also were blinded to the results of the nerve conduction studies on the patient. The board-certified neuromuscular physicians interpreting the nerve conduction data were blinded to the results of the tuning fork comparative examination.

Figure 1. Study flow diagram of patient assessment with tuning forks and nerve conduction studies.

Legends: n = number of patients; dp = data points; NCS = nerve conduction studies; NL= normal; ABN = abnormal

Standardized assessments

Following training under board-certified neuromuscular physicians, the qualitative and quantitative tuning forks were tested in a consecutive fashion by neuromuscular fellows and medical students. The tuning forks were struck maximally and applied perpendicularly to the (a) dorsal head of the first metatarsal joint, and (b) distal inter-phalangeal joint on the second digit (index finger). For the Rydel-Seiffer tuning fork, values ≤ 5 was categorized as abnormal at the distal interphalangeal joint on the second digit⁷; values ≤ 4 was categorized as abnormal at the first metatarsal joint.³ For the qualitative 128 Hz tuning fork, absent sensation was categorized as abnormal, however, no uniform criteria has been described for decreased vibratory sensation. Some have suggested placing the examiner’s finger on the opposite surface of the joint being tested from the tuning fork and noting if the vibration persists after the patient no longer senses it,² or assessing the patient’s threshold against the examiner’s, by applying the tuning fork to their finger.¹ These approaches compare proximal sensation to distal, which would overestimate abnormal values. We recorded sensation as abnormal if vibration was perceived on the corresponding joint of the examiner (for example, comparing the metatarsal joint vibration sensation of the patient to the same joint on the examiner), for a count of five seconds after patient reported cessation of sensation, thus increasing the specificity of the finding.

Electrophysiologic data

Within minutes of the physical examination, nerve conduction studies were obtained and interpreted for each patient by board-certified neuromuscular physicians; axonal neuropathy was determined based on our standardized EMG laboratory standards: amplitudes < 6μV, < 25 μV and < 10μV were categorized as reduced amplitudes at the sural nerve, median nerve and ulnar nerve respectively, when all other technical limitations had been accounted for. Since disability in neuropathy is reflected by axonal loss as opposed to conduction slowing, the decision to analyze only the amplitudes was made early on in the study design. All electrodiagnostic nerve assessments were limited to the referring doctor’s request; no additional conductions were performed for the sole purpose of this present study. This study was reported in accordance with the Standards for the Reporting of Diagnostic Accuracy Studies guidelines.

Statistical Analysis

For the analysis, right or left sural nerve conduction data was correlated to the measurements of the two tuning forks at the corresponding metatarsal joint, and right or left median and ulnar nerve conduction data was correlated to the measurements at the corresponding distal interphalangeal joint. True Negative (TN), True Positive (TP), False Negative (FN) and False Positives (FP) were calculated for the quantitative Rydel-Seiffer tuning fork and the qualitative 128 Hz tuning fork, as follows: TN was defined as normal tuning fork reading with normal sensory nerve action potentials (SNAPs); TP was defined as abnormal tuning fork reading with abnormal SNAPs; FN was defined as normal tuning fork reading with abnormal SNAPs; and FP was abnormal tuning fork reading with normal SNAPs.

Each of the median, ulnar and sural nerves was categorized separately, and data points for left and right of the same nerve also counted separately within the same category. For example, if a patient had a TP reading by the Rydel – Seiffer tuning fork on both left and right sural nerve, this patient contributed two data points.

For both tuning forks, the following measures of diagnostic accuracy were then calculated: Sensitivity = TP/(TP + FN), Specificity = TN/(TN + FP), Positive Predictive Value = TP/(TP + FP), and Negative Predictive Value = TN/(TN+FN). Finally, to compare the sensitivity and specificity of the two tuning forks to each other, an extended McNemar’s test⁸ was applied to determine whether there was a significant difference between the two tools.

Results

100 consecutive patients referred to the Neurology Outpatient Electromyography (EMG) clinic at a major academic center were consented for this study between April and August of 2012; 296 data points were collected from these individuals for the analysis below. No adverse events occurred during the data collection phase of this study. Using the normative threshold stated in the methods section, we categorized results obtained by the tuning forks and the SNAPs as either normal or abnormal. We then correlated the data from both tuning forks with the nerve conduction data to calculate the True Negative (TN), True Positive (TP), False Negative (FN) and False Positives (FP) (Table 1).

Table 1.

Diagnostic accuracy of the Rydel-Seiffer and Qualitative 128 Hz Tuning Fork

Rydel-Seiffer Tuning Fork
	Sural SNAPs (n)	Ulnar SNAPs (n)	Median SNAPs (n)	Total
True negative	47	78	41	166
True Positive	12	7	11	30
False Negative	13	10	51	74
False positive	4	15	7	26
TOTAL	76	110	110	296

128 Hz Tuning fork
	Sural SNAPs (n)	Ulnar SNAPs (n)	Median SNAPs (n)	Total
True negative	44	82	43	169
True Positive	8	5	9	22
False Negative	17	12	53	82
False positive	7	11	5	23
TOTAL	76	110	110	296

Legend: SNAP = sensory nerve action potential

Table 2 shows the calculated specificity, sensitivity, positive predictive value and negative predictive value of each tuning fork for axonal loss in the three corresponding nerves. Overall, both tuning forks showed better specificity than sensitivity. While not reaching significance, the quantitative Rydel-Seiffer had slightly higher overall sensitivity of 0.29; 95% CI [0.25, 0.33]; compared to the qualitative 128 Hz tuning fork’s sensitivity of 0.21; 95% CI [0.17, 0.25]. The 128 Hz tuning fork had slightly higher overall specificity of 0.88; 95% CI [0.84, 0.92] compared to the Rydel-Seiffer tuning fork’s specificity of 0.86; 95% CI [0.82, 0.90]. Predictive value calculations slightly favored the Rydel-Seiffer tuning fork: the positive predictive value of the Rydel-Seiffer tuning fork was slightly higher (54% vs. 49%), and the negative predictive value of the Rydel-Seiffer tuning fork was also slightly higher (69% vs. 67%).

Table 2.

Sensitivity and Specificity of Rydel-Seiffer vs. Qualitative 128 Hz Tuning Fork

Rydel-Seiffer Tuning Fork
	Sural SNAPs (%)	Ulnar SNAPs (%)	Median SNAPs (%)	Overall (%)
Sensitivity	48	41	18	29
Specificity	92	84	85	86
Positive predictive value	75	32	61	54
Negative predictive value	78	89	45	69

128 Hz Tuning fork
	Sural SNAPs (%)	Ulnar SNAPs (%)	Median SNAPs (%)	Overall (%)
Sensitivity	32	29	15	21
Specificity	86	88	90	88
Positive predictive value	53	31	64	49
Negative predictive value	72	87	45	67

Legend: SNAP = sensory nerve action potential

To assess whether the sensitivity and specificity of the two tuning forks were significantly different from each other, a 4 × 4 table was constructed to contrast the errors made by both tuning forks (see Table 3). 18 patients with neuropathy were falsely identified as normal by the 128 Hz tuning fork, while 13 patients with neuropathy were falsely identified as normal by the Rydel-Seiffer tuning fork. On the other hand, 22 patients without neuropathy were incorrectly identified as having neuropathy by the Rydel-Seiffer tuning fork, and 21 patients without neuropathy were incorrectly identified as having neuropathy by the 128 Hz tuning fork. An extended McNemar’s test, i.e. test for correlated proportions was applied and the following results were obtained: χ² = (13 – 18)²/(13+18) + (21 – 22)²/(21+22) = 0.83, which is much lower than the critical value for 5% significance level for two degrees of freedom (χ² = 5.99); i.e. there was no significant difference in diagnostic utility between the graduated 64 Hz Rydel-Seiffer and the qualitative 128 Hz tuning fork.

Table 3.

McNemar’s Test of Rydel-Seiffer vs. Qualitative 128 Hz Tuning Fork

		True Neuropathy Status	Rydel – Seiffer Tuning Fork
			True positive	False positive	False negative	True negative
			Positive	Negative	Positive	Negative
128 Hz Tuning Fork	True positive	Positive	–		13
	False positive	Negative		–		21
	False negative	Positive	18		–
	True negative	Negative		22		–

Discussion

The use of specific tools from pins, tuning forks, reflex hammers and ophthalmoscopes in routine neurological examination is largely based on personal preference of the physician. The graduated 64 Hz Rydel-Seiffer tuning fork has become a popular choice in testing vibratory sensation at the bedside or in the outpatient setting, and has been increasingly used to monitor nerve function in clinical trials for treatment of polyneuropathy. The Rydel-Seiffer tuning fork has the advantage of being a portable and relatively inexpensive instrument compared to electrodiagnostic studies or other quantifiable measures of sensory function. Compared to the qualitative 128 Hz tuning fork, the Rydel-Seiffer costs more, however, it is far easier to use in those patients with reduced as opposed to absent perception- the numeric scale that has been validated against nerve biopsy data makes it easy to determine abnormal values, while the lack of standardized approaches in the literature of determining reduced sensation with the qualitative fork led us to develop our own, erring on the side of increased specificity.

In this study we compared the diagnostic accuracy of the quantitative Rydel-Seiffer tuning fork with the qualitative 128 Hz tuning fork for predicting axonal neuropathy, as reflected by reduced sensory nerve action potentials. We found that both instruments have high specificity (i.e. capable of predicting the lack of neuropathy), but both have limited sensitivity. In a trial setting, where it is important to exclude those without the disease, both are equally valuable. We also found there is no significant difference in the diagnostic accuracy of both tools, as determined by our extended McNemar’s test. Previous studies have reached different conclusions about the relative superiority of the quantitative Rydel-Seiffer tuning fork⁵, this may be due to differences in study methodology-our study limited the definition of neuropathy to electrodiagnostic criteria only and excluded deep tendon reflexes and other similar, and perhaps equally valuable, components of the neurologic examination. However, our statistical evaluation of the comparative accuracy of both tunings forks was more rigorous than has been previously described.

Our study does not negate the value of a quantitative tuning fork, especially given its documented reliability, ease of use as described earlier, and portability. From our perspective, the advantage of using the Rydel-Seiffer fork as a quantitative outcome measure in a clinical trial is its potential to show responsiveness, i.e. sensitivity to change over time, compared to the qualitative 128 Hz tuning fork. In those settings where diagnostic accuracy is favored over responsiveness, either tool will serve the investigator and/or clinician well.