We sought to evaluate the utility of quantitative sensory testing (QST) and nerve conduction velocity (NCV) studies as measures of distal symmetric polyneuropathy (DSP). We studied 36 diabetic patients divided into four clinical categories of increasing severity. QST included thermal testing and vibration thresholds. NCV studies included median, peroneal, and sural nerves. Results of QST and NCV were compared among clinical groups using survival methodology. The log-rank statistic showed significant differences among the groups; the direction of the differences were consonant with clinical severity. For each diabetic patient, the result of each measurement was classified as normal or abnormal; more diabetic patients had abnormal NCV than either vibration tests or thermal tests. In conclusion, findings of QST and NCV are in keeping with clinical categorization of patients, QST and NCV are complementary tests, and the sural sensory study is the best single predictor of DSP. 0 1992 John Wiley & Sons, Inc. Key words: sensory testing diabetes mellitus polyneuropathy MUSCLE & NERVE 15:1334-1339 1992
SENSORY TESTING V E R S U S NERVE CONDUCTION VELOCITY I N DIABETIC POLYNEUROPATHY JANICE M.T. REDMOND, MD, MRCPI, FACP, MALACHI J. McKENNA, MD, MRCPI, FACP, MARCIA FEINGOLD, PhD, and B.K. AHMAD, MD
Diabetes meIIitus is probably the most cornnioii cause of peripheral neuropathy in developed naIt is a chronic disorder characterized by abnormalities in the metabolism of carbohydrate, fat, and protein that is often accompanied after some time by certain complications broadly classified into microvascular, macrovascular, and neuropathic. There is no standard electrodiagnostic approach to DSI',' although QS'l', by a variety of methods, is a d ~ o c a t e d . " ~ ~T~h"e' main objective of this study was to evaluate QST and NCV as measures of DSP by comparison with studies in healthy individuals. T h e spectrum of diabetic pa-
From the Neurophysiology Laboratory, Department of Neurology (Drs. Redmond and Ahmad); Department of Internal Medicine (Dr. McKenna), and Division of Biostatistics. Research Epidemiology and Computing (Dr. Feingold), Henry Ford Hospital, Detroit, Michigan Presented on September 28, 1991. at the 38th Annual Meeting of the American Association of Electrodiagnostic Medicine in Vancouver, for which Dr Redmond received runner-up place in the young investigator award. Acknowledgments: Dr. Redmond received support from the Small Projects Funding Committee and Dr. McKenna received support from the Robert and June Gurwin Diabetes Research Fund, Henry Ford Hospital Address reprint requests to Dr. Janice M.T Redmond. Department of Neurology, Adelaide Hospital, Peter Street, Dublin 8. Ireland. Accepted for publication April 5, 1992 CCC 0148-639X/92/121334-06 0 1992 John Wiley & Sons, Inc
1334
Sensory Testing in Diabetic Neuropathy
tients in this study ranged from those without symptoms of'1)SP to those with severe disease. MATERIALS AND METHODS
T h e study design was approved by the Project Research arid Hurnan Rights Committee at Henry Ford Hospital. All volunteers signed a consent form. Patients were enrollcd from a clinical practice with a view to selecting a sample with a broad spectr-urn of DSl'. Patients with both type I arid type I1 diabetes rnellitus were chosen; the sole selection criterion was at least 5 years duration of diabetes rnellitus in those with typc 1 disease. T h e candidates were separated into four groups based on clinical findings alone that were as f-ollows:
Patient Selection.
Group 1: Asymptomatic (n = 9). Group 2: Sensory symptoms only (n = 12) (i.e., numbness, tingling, pain, hypersensitivity to touch). Group 3 : Sensory symptoms and clinical autonomic nerve system involvement (n = 10) (i.e., postural drop exceeding 10 mmHg; o r symptom complex of gastroparesis/enteropathy; or impotence without evidence of peripheral vascular disease).
MUSCLE & NERVE
December 1992
Group 4: Patients with foot ulcers and/or Charcot joints (n = 5). Ninety-one healthy individuals, ranging in age from 20 to 66 years, served as controls and volunteered to the same protocol.
Clinical and Laboratory Data. The following clinical data was accrued on each subject: height, weight, body mass index (weight/height2)duration of diabetes mellitus, record of therapy for diabetes mellitus, presence of microvascular (retinopathy and nephropathy) and macrovascular (ischemic heart disease, peripheral vascular disease, stroke) complications, blood pressure (lying and standing), smoking history, and alcohol intake. The f-01lowing laboratory studies were obtained: glycosylated hemoglobin A,, (HbA,,), electrolyte profile, serum creatinine, creatinine clearance, 24-hour urine for total protein excretion (if albustix-positive) or for urinary albumin excretion (if albustixnegative). HbA,, was measured by high-performance liquid chromatography, urinary albumin by radioimmunoassay,’ and the other tests by routine laboratory procedures.
Studies. Thermal sensation was measured using a Somedic type I thermotest stimulator according to a modification of the Marstock method as previously d e ~ c r i b e d . ~The . ’ ~ following measurements were determined bilaterally at four different sites (i.e., cheek, thenar eminence, lateral calf, and sole of the foot): warm and cold thresholds, warm-cold limens, warm and cold sensitivities, and cold-pain and warm-pain thresholds. Vibration measurements were recorded using a Somedic type 111 vibrameter at three sites bilaterally: distal part of the second metacarpal bone, midportion of proximal tibia, and dorsal aspect of the first metatarsal head.”.“ T h e average of right and left sides was computed giving a total of 31 sensory tests per patient. Standard nerve conduction studies were performed on the dominant side, including two motor nerves (median and peroneal) and two sensory nerves (median and sural) using a DISA electromyography machine with bipolar stimulating and surface recording electrodes. Skin temperature was recorded at the thenar eminence and sole of the foot; studies were performed if it was above 32°C. Warming was carried out if required. T h e neurophysiological evaluation took about 1.5 hours. Neurophysiological
Sensory Testing in Diabetic Neuropathy
Descriptive statistics were obtained for all variables. Pearson correlation coefficients were obtained to evaluate pairwise relationships among dependent variables. The four neurological groups were compared with respect to covariates. T h e Mantel-Haenszel chi-square statistic was calculated in order to test for association between each of the categorical variables (gender, type of diabetes, alcohol consumption, and smoking history) and neuropathy groupings. To compare the four groups on the continuous covariates (age, height, weight, body mass index, duration of diabetes mellitus, HbA,,) a one-way analysis of variance was employed where appropriate, and a nonparametric median test otherwise. The 35 dependent variables (all conduction and sensory measurements) were analyzed. For all variables, some responses reached the limit of the measuring capabilities of the various techniques. Therefore, the data were handled by survival analysis methodology, which accommodates censored variables. For each of the four neuropathy groups the product-limit estimated survival distribution functions were plotted for all the dependent variables. -1he survival analysis log-rank statistic was calculated. Using data from the 91 normal subjects, onesided normal ranges were defined for each of the dependent variables. One-sided normal ranges were chosen on the basis that the direction of an abnormality is axiomatic, i.e., an abnormal cold sensitivity would be low and an abnormal warm sensitivity would be high. The 5th percentile point was chosen as the cut-off point for variables when a low value would be abnormal, and the 95th percentile point was chosen as the cut-off point for variables when a high value would be abnormal. If the 5th or 95th percentile cut-off points occurred beyond the “censoring” value, then no cut-off could be defined for that variable. Cut-off values could not be defined for all cold-pain thresholds and for all warm-pain thresholds except at the face leaving 28 dependent variables out of a total of 35 for this analysis. For each patient in the diabetic data set, their measurement on each of the 28 variables (conduction, vibratory, thermal) was classified as normal or abnormal. For each patient we computed the proportion of abnormal conduction tests (out of 4) and the proportion of abnormal sensory tests (out of 24). T o test whether conduction tests were better at classifying diabetics as abnormal than were sensory tests, a two-sided paired Student’s t-test was applied to the arcsines of the proportions. First, the paired Student’s Statistical Analysis.
MUSCLE & NERVE
December 1992
1335
t-test was done considering all 36 patients together, and second, it was done separately for each of the neuropathy groups. RESULTS
O u t of the total number of diabetic patients ( n = 36), 53% were men and 47% were women; 22% were black and 78% were white; 2 1 % were smokers, 79% were nonsmokers; alcohol were riot regularly consumed in 62741,consumed monthly in 18%, and weekly in 21%. T h e median age was 48 years (range 19 t o 77). As regards type of diabetes, 33% had type I and 67% had type 11, the latter consisted of those being treated with insulin (50%), oral hypoglycemic agents (8%), anti diet only (8%).Fifth-three percent were receiving therapy for hypertension. Postural hypotension was found in 28%. Impotence was a complaint in 68%)of' men, induding 90%)of' men in group 3 (a criterion for entry into this group) arid all men in group 4. Uackground retinopathy was noted in 25% and macular edema/proliferative retinopathy in 3 1%. Forty-two percent had normal albumin excretion, 42% had microalburriiriuria (albumin excretion rate = 13.1 to 250 Fginiin) and 15% had macroproteinuria (>250 pg/min). A history of' ischemic heart disease was evident in 14%, and o f peripheral vascular disease in 8%; all of' these cases were in groups 3 and 4.
Descriptive Statistics.
T h e correlation matrix of dependent variables comprised 1225 (35 x 35) values of which 595 are distinct. For ease of presentation these results are reduced to a 9 x 9 matrix, by computing means of correlations (Table 1). For similar categories (e.g., cold sensitivity vs. cold sensitivity), the average of 21 correlations among six different pairs of sites (facekhenar,
Pairwise Correlations.
face/calf, f'ace/foot, thenadcalf, thenar/foot, and calf/foot) was computed. For dissimilar categories, the average of 46 correlations among 1 0 different pairs of sites was computed. This data reduction was reasonable, because the correlations comprising each average were similar. Correlations were highest between sirriilar tests with the exception of warm-pain and cold-pain thresholds. Average correlations were determined for thermal tests at each site, excluding cold-pain and warm-pain thresholds. Vibration correlated poorly with thermal tests in the face ( r = 0.10), but better with thermal tests in both upper and lower extremities ( r = 0.33 and 0.38, respectively). T h e mean corr-elation between conduction (both sensory and motor) and thermal tests was best when examined at similar sites ( r = 0.31 at upper extremity, r = 0.52 at lower- extremity). The association between each of the categorical covariates arid the neuropathy group were significant for sex and alcohol intake, but not for smoking habits o r type of diabetes mellitus. A greater proportion of men was found in groups 3 and 4. Alcohol consumption was most cornnion in groups 3 and 4. Kegarcling continuous covariates, significant differences between the clinical groups were noted for height ( P = 0.03) arid weight ( P = O . O O l ) , but not for body mass index ( P = 0.06), HbA,, ( P = 0.10), age ( P = 0.14), and duration of diabetes mellitus ( P = 0.68). Mean values for weight, height, and body mass index increased in a consistent manner from groups 1 through 4. Mean HbA value was lowest in group 1. Pearson correlation coefficients between covariates and neuropathy groups w e r e as follows: height ( r = 0.46), weight ( r = 0.58), body mass index ( r =
Comparisons Between Neuropathy Groups.
,
SENSORY TESTING V E R S U S NERVE CONDUCTION VELOCITY I N DIABETIC POLYNEUROPATHY JANICE M.T. REDMOND, MD, MRCPI, FACP, MALACHI J. McKENNA, MD, MRCPI, FACP, MARCIA FEINGOLD, PhD, and B.K. AHMAD, MD
Diabetes meIIitus is probably the most cornnioii cause of peripheral neuropathy in developed naIt is a chronic disorder characterized by abnormalities in the metabolism of carbohydrate, fat, and protein that is often accompanied after some time by certain complications broadly classified into microvascular, macrovascular, and neuropathic. There is no standard electrodiagnostic approach to DSI',' although QS'l', by a variety of methods, is a d ~ o c a t e d . " ~ ~T~h"e' main objective of this study was to evaluate QST and NCV as measures of DSP by comparison with studies in healthy individuals. T h e spectrum of diabetic pa-
From the Neurophysiology Laboratory, Department of Neurology (Drs. Redmond and Ahmad); Department of Internal Medicine (Dr. McKenna), and Division of Biostatistics. Research Epidemiology and Computing (Dr. Feingold), Henry Ford Hospital, Detroit, Michigan Presented on September 28, 1991. at the 38th Annual Meeting of the American Association of Electrodiagnostic Medicine in Vancouver, for which Dr Redmond received runner-up place in the young investigator award. Acknowledgments: Dr. Redmond received support from the Small Projects Funding Committee and Dr. McKenna received support from the Robert and June Gurwin Diabetes Research Fund, Henry Ford Hospital Address reprint requests to Dr. Janice M.T Redmond. Department of Neurology, Adelaide Hospital, Peter Street, Dublin 8. Ireland. Accepted for publication April 5, 1992 CCC 0148-639X/92/121334-06 0 1992 John Wiley & Sons, Inc
1334
Sensory Testing in Diabetic Neuropathy
tients in this study ranged from those without symptoms of'1)SP to those with severe disease. MATERIALS AND METHODS
T h e study design was approved by the Project Research arid Hurnan Rights Committee at Henry Ford Hospital. All volunteers signed a consent form. Patients were enrollcd from a clinical practice with a view to selecting a sample with a broad spectr-urn of DSl'. Patients with both type I arid type I1 diabetes rnellitus were chosen; the sole selection criterion was at least 5 years duration of diabetes rnellitus in those with typc 1 disease. T h e candidates were separated into four groups based on clinical findings alone that were as f-ollows:
Patient Selection.
Group 1: Asymptomatic (n = 9). Group 2: Sensory symptoms only (n = 12) (i.e., numbness, tingling, pain, hypersensitivity to touch). Group 3 : Sensory symptoms and clinical autonomic nerve system involvement (n = 10) (i.e., postural drop exceeding 10 mmHg; o r symptom complex of gastroparesis/enteropathy; or impotence without evidence of peripheral vascular disease).
MUSCLE & NERVE
December 1992
Group 4: Patients with foot ulcers and/or Charcot joints (n = 5). Ninety-one healthy individuals, ranging in age from 20 to 66 years, served as controls and volunteered to the same protocol.
Clinical and Laboratory Data. The following clinical data was accrued on each subject: height, weight, body mass index (weight/height2)duration of diabetes mellitus, record of therapy for diabetes mellitus, presence of microvascular (retinopathy and nephropathy) and macrovascular (ischemic heart disease, peripheral vascular disease, stroke) complications, blood pressure (lying and standing), smoking history, and alcohol intake. The f-01lowing laboratory studies were obtained: glycosylated hemoglobin A,, (HbA,,), electrolyte profile, serum creatinine, creatinine clearance, 24-hour urine for total protein excretion (if albustix-positive) or for urinary albumin excretion (if albustixnegative). HbA,, was measured by high-performance liquid chromatography, urinary albumin by radioimmunoassay,’ and the other tests by routine laboratory procedures.
Studies. Thermal sensation was measured using a Somedic type I thermotest stimulator according to a modification of the Marstock method as previously d e ~ c r i b e d . ~The . ’ ~ following measurements were determined bilaterally at four different sites (i.e., cheek, thenar eminence, lateral calf, and sole of the foot): warm and cold thresholds, warm-cold limens, warm and cold sensitivities, and cold-pain and warm-pain thresholds. Vibration measurements were recorded using a Somedic type 111 vibrameter at three sites bilaterally: distal part of the second metacarpal bone, midportion of proximal tibia, and dorsal aspect of the first metatarsal head.”.“ T h e average of right and left sides was computed giving a total of 31 sensory tests per patient. Standard nerve conduction studies were performed on the dominant side, including two motor nerves (median and peroneal) and two sensory nerves (median and sural) using a DISA electromyography machine with bipolar stimulating and surface recording electrodes. Skin temperature was recorded at the thenar eminence and sole of the foot; studies were performed if it was above 32°C. Warming was carried out if required. T h e neurophysiological evaluation took about 1.5 hours. Neurophysiological
Sensory Testing in Diabetic Neuropathy
Descriptive statistics were obtained for all variables. Pearson correlation coefficients were obtained to evaluate pairwise relationships among dependent variables. The four neurological groups were compared with respect to covariates. T h e Mantel-Haenszel chi-square statistic was calculated in order to test for association between each of the categorical variables (gender, type of diabetes, alcohol consumption, and smoking history) and neuropathy groupings. To compare the four groups on the continuous covariates (age, height, weight, body mass index, duration of diabetes mellitus, HbA,,) a one-way analysis of variance was employed where appropriate, and a nonparametric median test otherwise. The 35 dependent variables (all conduction and sensory measurements) were analyzed. For all variables, some responses reached the limit of the measuring capabilities of the various techniques. Therefore, the data were handled by survival analysis methodology, which accommodates censored variables. For each of the four neuropathy groups the product-limit estimated survival distribution functions were plotted for all the dependent variables. -1he survival analysis log-rank statistic was calculated. Using data from the 91 normal subjects, onesided normal ranges were defined for each of the dependent variables. One-sided normal ranges were chosen on the basis that the direction of an abnormality is axiomatic, i.e., an abnormal cold sensitivity would be low and an abnormal warm sensitivity would be high. The 5th percentile point was chosen as the cut-off point for variables when a low value would be abnormal, and the 95th percentile point was chosen as the cut-off point for variables when a high value would be abnormal. If the 5th or 95th percentile cut-off points occurred beyond the “censoring” value, then no cut-off could be defined for that variable. Cut-off values could not be defined for all cold-pain thresholds and for all warm-pain thresholds except at the face leaving 28 dependent variables out of a total of 35 for this analysis. For each patient in the diabetic data set, their measurement on each of the 28 variables (conduction, vibratory, thermal) was classified as normal or abnormal. For each patient we computed the proportion of abnormal conduction tests (out of 4) and the proportion of abnormal sensory tests (out of 24). T o test whether conduction tests were better at classifying diabetics as abnormal than were sensory tests, a two-sided paired Student’s t-test was applied to the arcsines of the proportions. First, the paired Student’s Statistical Analysis.
MUSCLE & NERVE
December 1992
1335
t-test was done considering all 36 patients together, and second, it was done separately for each of the neuropathy groups. RESULTS
O u t of the total number of diabetic patients ( n = 36), 53% were men and 47% were women; 22% were black and 78% were white; 2 1 % were smokers, 79% were nonsmokers; alcohol were riot regularly consumed in 62741,consumed monthly in 18%, and weekly in 21%. T h e median age was 48 years (range 19 t o 77). As regards type of diabetes, 33% had type I and 67% had type 11, the latter consisted of those being treated with insulin (50%), oral hypoglycemic agents (8%), anti diet only (8%).Fifth-three percent were receiving therapy for hypertension. Postural hypotension was found in 28%. Impotence was a complaint in 68%)of' men, induding 90%)of' men in group 3 (a criterion for entry into this group) arid all men in group 4. Uackground retinopathy was noted in 25% and macular edema/proliferative retinopathy in 3 1%. Forty-two percent had normal albumin excretion, 42% had microalburriiriuria (albumin excretion rate = 13.1 to 250 Fginiin) and 15% had macroproteinuria (>250 pg/min). A history of' ischemic heart disease was evident in 14%, and o f peripheral vascular disease in 8%; all of' these cases were in groups 3 and 4.
Descriptive Statistics.
T h e correlation matrix of dependent variables comprised 1225 (35 x 35) values of which 595 are distinct. For ease of presentation these results are reduced to a 9 x 9 matrix, by computing means of correlations (Table 1). For similar categories (e.g., cold sensitivity vs. cold sensitivity), the average of 21 correlations among six different pairs of sites (facekhenar,
Pairwise Correlations.
face/calf, f'ace/foot, thenadcalf, thenar/foot, and calf/foot) was computed. For dissimilar categories, the average of 46 correlations among 1 0 different pairs of sites was computed. This data reduction was reasonable, because the correlations comprising each average were similar. Correlations were highest between sirriilar tests with the exception of warm-pain and cold-pain thresholds. Average correlations were determined for thermal tests at each site, excluding cold-pain and warm-pain thresholds. Vibration correlated poorly with thermal tests in the face ( r = 0.10), but better with thermal tests in both upper and lower extremities ( r = 0.33 and 0.38, respectively). T h e mean corr-elation between conduction (both sensory and motor) and thermal tests was best when examined at similar sites ( r = 0.31 at upper extremity, r = 0.52 at lower- extremity). The association between each of the categorical covariates arid the neuropathy group were significant for sex and alcohol intake, but not for smoking habits o r type of diabetes mellitus. A greater proportion of men was found in groups 3 and 4. Alcohol consumption was most cornnion in groups 3 and 4. Kegarcling continuous covariates, significant differences between the clinical groups were noted for height ( P = 0.03) arid weight ( P = O . O O l ) , but not for body mass index ( P = 0.06), HbA,, ( P = 0.10), age ( P = 0.14), and duration of diabetes mellitus ( P = 0.68). Mean values for weight, height, and body mass index increased in a consistent manner from groups 1 through 4. Mean HbA value was lowest in group 1. Pearson correlation coefficients between covariates and neuropathy groups w e r e as follows: height ( r = 0.46), weight ( r = 0.58), body mass index ( r =
Comparisons Between Neuropathy Groups.
,
