Measuring Diabetic Neuropathy - Diabetes Care - American Diabetes ...

Measuring Diabetic Neuropathy Assessment and Comparison of Clinical Examination and Quantitative Sensory Testing

The need for a standardized and valid means of assessing diabetic neuropathy has been increasingly recognized. To identify potential components of such an assessment, interobserver variation (neurologist and internist) of a standard neurologic examination and the comparability of this examination with vibratory and thermal sensitivity testing was studied. The study population comprised the first 100 participants in a neuropathy substudy of 25- to 34-yr-old subjects with insulin-dependent diabetes mellitus taking part in a cohort follow-up study. Symptoms of dysesthesias, paresthesias, and burning, aching, or stabbing pain revealed good interobserver agreement. Signs of neuropathy, more prevalent in the great toe than index finger, showed poor interobserver agreement for vibration, but fair interobserver agreement for touch and pinprick. Mean quantitative sensory thresholds differed significantly by clinical category of abnormal vibratory and pinprick sensations. Threshold testing showed twice the prevalence of abnormality compared with clinical examination. It is concluded that components of the clinical examination can be identified that, along with quantitative sensory-threshold testing, may provide a satisfactory core assessment for use both in epidemiologic studies and incorporation into more in-depth protocols required for clinical research and practice. The clinical relevance of the greater prevalence of abnormalities on threshold testing will be established by long-term follow-up. Diabetes Care 12:270-75, 1989

From the Department of Epidemiology, University of Pittsburgh, the Department of Neurology, University of Pittsburgh School of Medicine, the Department of Endocrinology, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania; and the Department of Medicine, The Johns Hopkins Hospital, Baltimore, Maryland. Address correspondence and reprint requests to Raelene E. Maser, Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA 15261.

270

Raelene E. Maser, MS Viggo K. Nielsen, MD, PhD Eric B. Bass, MD, MPH Qurashia Manjoo, MD Janice S. Dorman, PhD Sheryl F. Kelsey, PhD Dorothy J. Becker, MBBCh, FCP Trevor). Orchard, MBBCh, MMedSci

T

he study of diabetic neuropathy presents one of the most challenging exercises in the field of diabetes epidemiology. The main issues contributing to this include 7) the lack of a clear definition of diabetic neuropathy, 2) the absence of simple repeatable tests of neuropathy that are not dependent on either expensive technology or subjective clinical judgment, 3) the varied manifestations of neuropathy, ranging from distal symmetrical polyneuropathy, which is most common, to mononeuropathies and autonomic neuropathies, and 4) the separation of diabetes from other potential etiologies of neuropathy. One method used for the diagnosis of diabetic neuropathy has been the neurological examination, which usually includes tendon reflex activity, touch-pressure, vibratory, joint position, and pinprick sensations and an assessment for other causes of neuropathy. Abnormalities found during physical examination can be related to the symptoms and functional deficits of the diabetic individual. The judgment whether a diabetic individual has neuropathy may vary among physicians according to their level of training and experience. New methods of quantitative sensory testing (vibratory and thermal) have been developed that use stimuli with defined wave forms and forced-choice approaches to decrease response bias. The comparability between these methods and the standard neurological examination in identifying diabetic neuropathies, and the predictive value of these methods for subsequent significant clinical neuropathy needs to be assessed. In this study, we examined 7) the degree of consistency between two trained observers (neurologist and internist) in performing the same neurological examination in a well-defined cohort of individuals with insulin-dependent diabetes mellitus (IDDM) and 2) the

DIABETES CARE, VOL. 12, NO. 4, APRIL 1989

R.E. MASER AND ASSOCIATES

comparability of quantitative sensory testing with the neurological examination. MATERIALS AND METHODS Subjects. Participants in the study were local residents from the Children's Hospital of Pittsburgh (CHP) IDDM onset registry from the years 1950-1979 and were a well-defined cohort of IDDM patients diagnosed or seen within 1 yr of diagnosis at CHP (1). All participants in the follow-up study (epidemiology of diabetes complications; EDC) were examined at baseline for diabetic neuropathy by one of three trained internists (E.B.B., T.J.O., orQ.M.). All 25- to 34-yr-old participants were invited to take part in the neuropathy special study. The first 100 participants of the neuropathy special study form the basis of this report. There were 46 women and 54 men, mean ± SD age 28.7 ± 2.7 yrand mean duration of diabetes 19.8 ± 5.1 yr. The neuropathy special study, consisting of vibratory and thermal-threshold testing, was performed a few months after the baseline EDC examination (mean time between visits 4.4 mo). In addition, 52 of the participants who were able to participate in the neuropathy special study on a weekday had a second neurological examination performed by a certified neurologist (V.K.N.) blinded to the results of any of the other testing. With the exception of 8 participants, the sensory testing and second neurological examination were performed on the same day. Clinical evaluation for neuropathy. The clinical evaluation was based on that currently used for the Diabetes Control and Complications Trial (2). A standard clinical history was taken with regard to /) any conditions other than diabetes that could cause neuropathy, 2) exposure to known neurotoxins, and 3) a family history of neuromuscular disorders. Participants were questioned about sensory and autonomic symptoms. Positive responses were recorded, e.g., numbness, dysesthesias and paresthesias, hypersensitivity to touch, and burning, aching, or stabbing pain in the hands and/or feet. Symptoms of autonomic neuropathy included orthostatic dizziness, early satiety, excessive diarrhea, severe constipation, genitourinary symptoms, sudomotor symptoms, and reduced awareness of hypoglycemia. They were recorded as positive if present for at least 30 days. A standard neurological examination tested the sensation to light touch (cotton wool), pain (pinprick), vibration (128-Hz tuning fork), and position on both sides of the body, but only the results of the dominant side were used for analyses. Sensation of the index finger and great toe were graded as normal, reduced, or absent. Tendon reflexes were graded with the Neurological Disability Score (normal, present only with reinforcement, or absent) and also with this scale (unobtainable 0, present with reinforcement ±, normal + , and brisknormal + + ).


Sensory threshold. Sensory thresholds were measured on the palmar aspect of the distal phalanx of the index finger and plantar aspect of the great toe on the dominant side of the body. Vibratory sensory thresholds were measured with the Vibratron II (Sensortek, Clifton, NJ). The procedure used was the two-alternative forced-choice method, where the individual was required to indicate which of two posts was vibrating. Only one post was set to vibrate according to a random procedure determined by the examiner. If the response was correct, the voltage was reduced by 10% for each subsequent trial until the subject made the first error. Then, the voltage setting was repeated twice. If the stimulus was correctly identified on two of three trials, the voltage was further reduced by 10%. If errors were made on two of three trials, the voltage was raised 10%. Testing was completed when five errors had been made. The vibration threshold was determined by identifying the five errors and the five lowest correct scores, eliminating the highest and lowest, and calculating the mean for the remaining eight scores. The threshold for temperature discrimination was measured with the Thermal Sensitivity Tester NTE-2 (Sensortek) via the two-alternative forced-choice method. The tester had two plates, and during each trial the individual decided which plate was colder. The procedure used was the same as for Vibratron II (3). Statistical methodology. Agreement between two observers (E.B.B. and V.K.N.) was analyzed, as was agreement between different methods of assessment. Categorical data was expressed as percentage agreement; i.e., the number of instances for which the two observers' observations agreed divided by the total number of comparisons. In addition to percentage agreement, the K statistic, a standardized measure, was also used as a means of correcting for observer agreement that could occur by chance. When complete agreement occurs, K = + 1 ; when values are >0, it indicates agreement greater than expected by chance; and when values are 10%. To compare sensory thresholds by clinical categories, analysis of variance was used. The best fit to a normal distribution was obtained by logarithmic transformation (log)0) of the vibratory and thermal sensory-threshold results. To avoid negative logs, one was added to each threshold value before transformation.

RESULTS Interobserver variation. Conditions other than diabetes that could cause neuropathy, symptoms of numbness,

271

MEASURING DIABETIC NEUROPATHY

TABLE 1 Interobserver agreement for clinical evidence of diabetic neuropathy in IDDM subjects Prevalence (%)

History Other concurrent disease processes Sensory symptoms Numbness Dysesthesias and paresthesias Pain Autonomic symptoms Orthostatic dizziness Nausea Reduced awareness of hypoglycemia Sensation (great toe)* Touch Vibratory Pinprick Reflexes Ankle jerkt Ankle jerk* Ankle jerk§ Clinical neuropathy Neuropathy|| NeuropathyH

1

N

M

Agreement (%)

K

26

6

16

80

0.31

35

24

29

69

0.26

28 14

29 18

29 16

82 88

0.57 0.55

6 12

14 10

10 11

88 90

0.35 0.49

18

20

19

90

0.68

14 29 31

12 12 15

13 20 23

87 75 81

0.39 0.26 0.48

49 50 58

18 17 37

34 34 47

61 67 75

0.38 0.35 0.52

37 32

25 27

31 30

81 56

0.56 0.33

n = 52. I, internist; N, neurologist; M, percentage mean prevalence of abnormality judged across two observers. *2-Category scale: normal and abnormal. t4-Category scale: 0, ± , + , and + + . *2-Category scale: present and absent. §2-Category scale: normal and abnormal (present with reinforcement or absent combined). ||2-Category scale: no neuropathy and definite neuropathy. 113-Category scale: no neuropathy, possible neuropathy, and definite neuropathy.

and postural hypotension revealed poor interobserver agreement, whereas symptoms of dysesthesias and paresthesias, pain in the hands and/or feet, and autonomic symptoms of nausea and reduced awareness of hypoglycemia showed good agreement (Table 1). Large and small nerve fiber involvement was assessed during the neurological examination by testing touch, vibration, and pinprick sensations of the index finger and great toe (prevalence of abnormality with regard to the index finger was extremely low, therefore only results of the great toe will be included). Agreement improved marginally for all sensations when the number of categories was reduced (e.g., pinprick K = .41 for three categories, K = .48 for two categories). Interobserver agreement for ankle-jerk activity showed improvement of the K-value and percent agreement when evaluation of the Neurological Disability Score was reduced to two categories, where the abnormal category

272

comprised either absent reflex or activity that had to be elicited with reinforcement (Table 1). Definite neuropathy was defined as the presence of two or more of the following: symptoms, sensorimotor signs, and/or absent reflexes. Possible neuropathy was defined as the presence of only one of the above. When none of the above was present, the individual was classified as having no neuropathy. Combining the categories of possible neuropathy with no neuropathy improved agreement between the two observers (Table 1). In general, it appeared that the internist recorded more abnormalities taking a history or on examination than the neurologist with whom he was initially standardized (Table 1). Sensory threshold testing. The qualitative portion of the neurological examination, performed by the internists, for touch and vibratory sensations of the index finger and great toe was compared to the mean vibratory thresholds, whereas pinprick sensation was compared with the mean thermal thresholds (Table 2). It was observed that the vibratory and thermal thresholds increased as sensation by physician testing decreased for all parameters tested for the great toe. This observation for the index finger was only seen with regard to the vibratory sensation. The mean levels for participants in the three categories (normal, reduced, or absent) were statistically different from each other with regard to the threshold determinations for touch (P < .05), vibratory (P < .0001), and pinprick (P < .01) evaluations for the great toe but not for the index finger. Figures 1 and 2 display the distribution of sensory thresholds by clinical category. Vibratory (P < .0001) and thermal (P < .001) thresholds were also significantly different from each other among the three participant groups defined by clinical evidence of neuropathy (Table 3). With the use of normal ranges given by the manufacturer of the Vibratron II and Thermal Sensitivity Tester NTE-2 as cutoff points, comparisons were made between vibratory and thermal thresholds for the great toe and results of the neurological examination. Touch, vibratory, and position sensations combined were compared with the vibratory thresholds and gave an agreement of 61% (K = .30; Table 4). Pinprick sensation compared with thermal thresholds resulted in agreement of 60% (K = .19; Table 5). In most cases, disagreement was due to a normal grade by the neurological exam but an elevated threshold detected (elevated vibratory threshold 36% and elevated thermal threshold 32%). Symptoms of numbness, dysesthesias and paresthesias, hypersensitivity to touch, and pain combined were compared with vibratory and thermal thresholds. Both percentage agreement and K values were poor (vibratory threshold 46%, K = .09, and thermal threshold 54%, K = .07), again due to the number of individuals who had an abnormal threshold but no symptoms (elevated vibratory threshold 49% and elevated thermal threshold 35%). Variability of sensory threshold testing was evaluated



TABLE 2 Comparison of neurological exam with vibratory and thermal sensory thresholds (log transformed) in IDDM subjects Graded sensation on neurological examination

Vibratory finger Touch Vibratory Thermal finger Pinprick Vibratory toe Touch* Vibratoryt Thermal toe Pinprick*

Absent

Reduced

Normal Mean ± SD

n

Mean ± SD

n

0.34 ± 0.09 (0.15-0.59) 0.34 ± 0.09 (0.15-0.59)

89

0.32 ± 0.11 (0.18-0.58) 0.40 ± 0.10 (0.30-0.58)

11

0.28 ±0.14 (0.08-0.90)

89

0.32 ± 0.16 (0.11-0.53)

0.65 ± 0.20 (0.30-1.09) 0.58 ± 0.16 (0.30-0.90)

89

0.76 ± 0.24 (0.47-1.14) 0.85 ±0.16 (0.47-1.09)

0.56 ± 0.40 (0.06-1.27)

75

93

70

0.70 ± 0.39 (0.23-1.27)

Mean ± SD

n

10

0.11

1

9

1.01 (0.94-1.08) 1.11 (1.09-1.14)

2

1.06 ± 0.38 (0.33-1.30)

8

7

28

17

2

Ranges in parentheses, n = 100. *P < .05 for difference among means by analysis of variance (ANOVA). iP < .0001 for difference among means by ANOVA. tP < .01 for difference among means by ANOVA.

for five nondiabetic subjects (aged 23-30 yr) on three separate occasions (1-145 days between testing procedures) and revealed the following mean coefficients of variation (C.V.) of Iog10-transformed variables (vibratory: finger 20%, great toe 16%; thermal: finger 34%, great toe 24%).

Variability of sensory threshold testing was also evaluated for five diabetic subjects (aged 24-40 yr) on three separate occasions (1-142 days between testing procedures) and revealed the following mean C.V. of log,0transformed variables (vibratory: finger 8%, great toe 8%; thermal: finger 29%, great toe 26%).

1* 0"

A

VIBRATORY SENSORY THRESHOLDS (LOG TRANSFORMED)

FIG. 1. Percentage frequency of 100 diabetic subjects with normal {open bars), reduced (hatched bars), or absent (solid bars) vibratory sensation of great toe on clinical examination by vibratory sensory thresholds.


THERMAL SENSORY THRESHOLDS (LOG TRANSFORMED)

FIG. 2. Percentage frequency of 100 diabetic subjects with normal (open bars), reduced (hatched bars), or absent (solid bars) thermal sensation of great toe on clinical examination by thermal sensory thresholds.

273

MEASURING DIABETIC NEUROPATHY

TABLE 3 Comparison of clinical evidence of neuropathy and vibratory and thermal sensory thresholds (log transformed) in IDDM subjects Clinical evidence of neuropathy

Vibratory toe* Thermal toet

No

Possible

Definite

0.56 ± 0.16(36) 0.44 ± 0.35 (36)

0.64 ± 0.19(30) 0.64 ± 0.42 (30)

0.80 ± 0.20 (34) 0.81 ± 0.41 (34)

n = 100; ranges and n are in parentheses. Neuropathy defined as at least 2 of 3 criteria: 7) symptoms consistent with neuropathy, 2) abnormal sensory or motor signs, and 3) diminished or absent tendon reflexes. *P < .0001 for difference among means by analysis of variance (ANOVA). fP < .001 for difference among means by ANOVA.

his study evaluated interobserver variation in a neurological examination to assess diabetic neuropathy. In addition, vibratory and thermal thresholds were compared with the clinical evaluation of signs and symptoms. With regard to interobserver agreement, although the K-values were good for the assessment of some somatic and autonomic signs and/or symptoms, most K-values were regarded as poor. Clinical reliability was also evaluated by determining intraobserver agreement in the assessment of diabetic neuropathy. Intraindividual reproducibility was examined with the same neurological exam for six diabetic subjects (aged 12-38 yr, duration of diabetes 10-18 yr), not part of the EDC population, by one of the internists (E.B.B.) on two separate occasions (~2 wk apart). Agreement ranged from 83 to 100% for signs and symptoms of neuropathy suggesting fair reproducibility. Although the techniques of the clinical examination are effective in evaluating gross sensory response and are moderately reproducible, they are not quantitative and are subjective (5,6). For example, the intensity of the stimulus provided by the examiner and the perception of that stimulus by the patient can only be graded roughly, which may explain some of the variability in inter- and intraobserver variation (7). Logarithmic transformation (log10) of the threshold values obtained for sensory testing enabled examination

of differences between categories based on normally distributed data. Vibratory-threshold testing showed good separation between clinical examination categories, whereas thermal thresholds showed considerable overlap between categories, although a significant difference was present. Thus, these modalities may have a useful place as objective supplements to clinical examination; however, they do have a large intraindividual variation. Our studies (of diabetic and nondiabetic subjects) indicated intraindividual C.V.s for vibratory and thermal thresholds in the 8-34% range. Other studies of diabetic subjects have also observed high coefficients for intraindividual variation (8,9). Considering the high C.V.s with these methods on repeat testing, individuals with borderline abnormal sensory thresholds could be determined to be normal and vice versa. In this IDDM population, they also yield a more sizable group of abnormal findings, based on established normal ranges (69% vibratory vs. 36% exam, 49% thermal vs. 25% exam). The significance of the twofold increase in abnormality rates needs to be carefully determined in longterm follow-up studies of representative well-defined populations, with particular focus on symptomatic disease. This investigation provides baseline values on which to assess the predictive value of quantitative sensory testing with subsequent follow-up of this cohort. The issues of observer variation, reproducibility of sensory testing, and the comparability of these tests with the clinical exam are critical to the development of a standard methodology for the measurement and grading

TABLE 4 Comparison of signs of great toe with vibratory sensory thresholds in IDDM subjects

TABLE 5 Comparison of signs of great toe with thermal sensory thresholds in IDDM subjects

DISCUSSION

T

Thermal threshold*

Vibratory threshold* Signst

Normal

Abnormal

Total

Signst

Normal

Abnormal

Total

Normal Abnormal Total

28 3 31

36 33 69

64 36 100

Normal Abnormal Total

43 8 51

32 17 49

75 25 100

n = 100. Agreement is 61%; K = .30. *Vibratory great toe (normal ^2.20, absent >2.20). tSigns: touch, vibration, and position combined.

274

n = 100. Agreement is 60%; K = .19. *Thermal great toe (normal 2.23). tSigns: pinprick.



of neuropathy. The need for a standard core of measurements, exemplified by the need for reliable tests on which to base clinical decisions with regard to treatment, particularly in light of new drug therapies, was recently discussed at a conference on diabetic neuropathy sponsored by the American Diabetes Association and the American Academy of Neurology, where the importance of standardization was emphasized. Our findings lead us to the following recommendations for clinical and epidemiological purposes in the context of the need for appropriate core measurements as noted above. These recommendations are based on the recognition that the assessments of diabetic neuropathy should be fairly reproducible, have reasonably low interobserver variation, and be internally consistent (e.g., show association with an alternative assessment modality). In addition, the complication that they are designed to detect should be prevalent for epidemiologic studies. Dysesthesias and paresthesias and burning, aching, or stabbing pain have good interobserver agreement and, therefore, are recommended components, whereas numbness shows more variation between observers and, thereby, is of less value. Symptoms and signs in the fingers were rare and, therefore, are probably less useful for epidemiologic and research purposes. Signs of neuropathy were more prevalent in the great toe, with touch sensation revealing fair interobserver agreement and vibratory sensation revealing poor interobserver agreement. There was, however, a high association between abnormal touch and vibratory sensations on clinical examination and an elevated vibratory threshold. Vibratory threshold testing, especially in diabetic subjects, appears very reproducible. Consequently, clinical touch and vibratory-threshold measurements may be useful measures when used together. Pinprick sensation shows better interobserver agreement but relates to thermal-threshold testing less with more overlap between the clinical categories than was seen for touch and vibration. In addition, thermal-threshold testing appears less reproducible than vibratory thresholds. Therefore, the best way to determine small fiber defects is less clear and thermal-threshold testing needs further assessment.


Finally, although both modalities of quantitative-sensory testing show useful promise and relate to clinical disease, the true significance of the higher prevalence of abnormality detected by these modalities must await long-term follow-up studies.

ACKNOWLEDGMENTS

This study was supported by a grant from National Institutes of Health R-01-DK-34818-03.

REFERENCES 1. Wagener DK, Sacks JM, LaPorte RE, MacGregor JM: The Pittsburgh study of insulin-dependent diabetes mellitus: risk for diabetes among relatives of IDDM. Diabetes 31:13644, 1982 2. The DCCT Research Group: Manual of Operations for the Diabetes Control and Complications Trial. Washington, DC, U.S. Dept. of Commerce, 1987 3. Arezzo JC, Schaumburg HH, Laudadio C: Thermal sensitivity tester: device for quantitative assessment of thermal sense in diabetic neuropathy. Diabetes 35:590-92, 1986 4. Fleiss JL: Statistical Methods for Rates and Proportions. 2nd ed. New York, Wiley, 1981 5. Katims JJ, Naviasky EH, Rendell MS, Ng LKY, Bleecker ML: Constant current sine wave transcutaneous nerve stimulation for the evaluation of peripheral neuropathy. Arch Phys Med Rehabil 68:210-13, 1987 6. Bleecker ML: Quantifying sensory loss in peripheral neuropathies. Neurobehav Toxicol Teratol 7:305-308, 1985 7. Katims JJ, Naviasky EH, Ng LKY, Rendell M, Bleecker ML: New screening device for assessment of peripheral neuropathy. I Occup Med 28:1219-21, 1986 8. Levy DM, Abraham RR, Abraham RM: Small- and largefiber involvement in early diabetic neuropathy: a study with the medial plantar response and sensory thresholds. Diabetes Care 10:441-47, 1987 9. Abraham RR, Abraham RM, Simpson R, Levy DM: Variability of neurological measurements in diabetic neuropathy (Abstract). In Int Symp Polyol Pathway and Its Role in Diabetic Complications: Current Topics in Experimental and Clinical Aspects of Aldose Reductase Inhibitor, Kashikojima, Japan, 28-30 October 1986, p. 34-35

275