Longitudinal assessment of diabetic polyneuropathy using a composite score in the Rochester Diabetic Neuropathy Study cohort

There is some information on the frequency and risk factors for diabetic neuropathies,^1-16 but there is little longitudinal data on change in severity of diabetic polyneuropathy(DP) over time, especially from population-based diabetic cohorts. This lack exists because investigators did not measure quantitative symptoms, neurologic deficit, tests of nerve dysfunction, and health and work outcomes longitudinally in population-based cohorts of diabetic patients and control subjects. In addition, they did not use composite measures of severity of DP, using reference values from large nondiabetic populations. There is some longitudinal data of the change in frequency of DP based on dichotomous measures of DP, for example, normal or decreased vibratory perception at the lateral malleolus or great toe, decreased ankle reflexes,^1,2 or abnormal attributes of nerve conduction.³ Knowing the frequency of polyneuropathy is of value, but it provides limited information about severity, course, or outcome. To measure severity, we introduced a staging approach^14,17 that takes into account occurrence and severity of symptoms and of neurologic impairment. The approach provided an improved estimate of the overall severity of DP as compared with a simple tally of the percentage of patients with DP.^13,15 Because symptoms are not constant but tend to come and go, for purposes of following course, it is useful to have an overall measurement of severity of polyneuropathy excluding symptoms. Such measurements might be thought of as overall (or composite) measurements of neuropathic impairment. Development of this composite score and availability of longitudinal data from the Rochester Diabetic Neuropathy Study (RDNS) permitted assessing change in DP over time and estimating power, which are the subjects of this report.

Methods. The RDNS cohort. The RDNS was initiated approximately 12 years ago to estimate the magnitude of the health problems from diabetic neuropathies and from other micro- and macrovascular complications and also to determine risk factors and complications in a population-based cohort.¹³ Because there is a high degree of ascertainment of diabetes mellitus (DM) in Rochester, MN, and because comorbidity of diabetic patients participating in the study was not significantly different from those who did not participate in the study,¹⁵ the RDNS cohort was thought to be representative of northern midwestern U.S. diabetic persons predominantly of Northern European extraction. Participating patients were prospectively evaluated at regular intervals for nerve, eye, and kidney complications using standard and validated assessments for which reference values were available based on the same assessments of a large cohort of healthy subjects randomly drawn from the same population.¹⁶ Risk factors were prospectively evaluated every 3 months or less often. The RDNS cohort included patients of all adult ages and both diabetic disorders, insulin-dependent DM (IDDM), and non-insulin-dependent DM (NIDDM).

Of 64,573 inhabitants of Rochester, Minnesota, on January 1, 1986, 870(1.3%) had clinically recognized DM (by National Diabetes Data Group criteria) of whom 380 were enrolled in the RDNS. Of these, 102 (26.8%) had IDDM and 278 (73.2%) had NIDDM. Overall, approximately two-thirds of diabetic patients had objective evidence for some variety of neuropathy but only about 13% had symptoms from DP.

Longitudinal RDNS cohort. From the prevalent cohort, we included for longitudinal study only those diabetic patients who had been serially evaluated for at least 2 years, had the neuropathic evaluations and tests needed for this study at each longitudinal evaluation, and were without confounding neurologic disease.

Assessment of symptoms and impairments. At yearly intervals, neuropathic symptoms were assessed by the Neuropathy Symptoms Score(NSS),¹⁸ the Neuropathy Symptom Profile(NSP),¹⁹ and later by the Neuropathy Symptoms and Change Score (NSC). NSS is a standard tally of neuropathic symptoms abstracted from a neurologic history that inquires about neuropathic symptoms. NSP is a standard true and false questionnaire completed by patients and read by optical reader and computer that derives scales of neuropathy, weakness, sensation, and autonomic dysfunction by comparing patient responses to responses obtained from a reference population, the Rochester Diabetic Neuropathy Study of Healthy Subjects cohort. The NSC is a standard true and false questionnaire of neuropathic symptoms based on an interview of patients by a neurologist. There was insufficient longitudinal data for analysis of NSC.

The Neuropathy Impairment Score (NIS) was obtained yearly. A standard group of muscles were evaluated for weakness: 1, 25% weak; 2, 50% weak; 3, 75% weak; 3.25, movement against gravity; 3.5, movement with gravity eliminated; 3.75, muscle flicker without movement; and 4, paralyzed. A standard group of muscle stretch reflexes were graded as normal, 0; decreased, 1; or absent, 2. Touch-pressure, vibration, joint position and motion, and pinprick were graded on index finger and great toe as normal, 0; decreased, 1; or absent, 2. All evaluations were corrected for age, gender, and physical fitness. For this study, neuropathic impairment due to another disorder other than DP was not scored. For evaluating the NIS for the lower limbs (NIS(LL)), only neurologic abnormalities of the lower limb were tallied.

Quantitative Sensory Testing (QST). At yearly intervals, vibration detection threshold (VDT) was evaluated using CASE IV and the 4, 2, and 1 stepping algorithm. This CASE system uses standard and calibrated 125-Hz vibratory mechanical oscillations in 25 graded steps of stimulus magnitude from 0.1 to 576.6 µm. The objective was to determine the smallest mechanical displacement that could be detected. During testing of VDT, sound at 125 Hz was given continuously by head phones to ensure that the vibration stimulus could not be heard. Written and spoken instructions and a practice session were used before the test was given. The technician should be convinced that the patient understood and was cooperating, that there was concordance between the practice and test results, and that positive responses were not given beyond a certain percentage of trials when null stimuli were presented. Patients should not be drowsy or under the influence of sedatives, pain medications, or other mind-altering drugs.

Nerve conduction (NC). Because NC provides sensitive, reproducible, and quantitative methods of assessing dysfunction of peripheral nerves, evaluated bi-yearly were motor nerve conduction velocity (MNCV), compound muscle action potential (CMAP) amplitude, and motor nerve distal latency (MNDL) of motor fibers of ulnar (U), peroneal (P), and tibial (T) nerves and of sensory nerve action potential (SNAP) amplitude, sensory nerve distal latency (SNDL), and conduction velocity (SNCV) of U and sural (S) nerves. F-waves were also assessed for some nerves. Five-millimeter-diameter tin disks were used in all studies for percutaneous stimulation and surface recording. The limbs of all patients were warmed throughout the study with heat lamps, and the temperature of the limbs was monitored lamps, and the temperature of the limbs was monitored continuously to maintain a surface temperature >32 °C (≥30.5 °C in the foot). All recordings were made on TECA TE 42 (White Plains, NY) or Nicolet Viking (Madison, WI) machines. Responses were recorded on light-sensitive paper. Amplitudes and latencies were measured and conduction velocities calculated separately by two individuals.

Heart beat response to deep breathing. The variability of the heart beat to deep breathing (HB DB) was determined.

Analyses. The different endpoint results of patients were expressed in original units, natural log (In) values, or a percentile value specific for nerve, attribute of NC, modality of sensation, age, gender, and applicable physical characteristics. Results were also expressed as a normal deviate value (the normal deviate corresponding to the observed percentile). Generally, abnormality of individual endpoints were set at ≥99th or≤1st percentile (whichever applied) or at the 97.5th percentile for composite scores (e.g., NIS(LL)+7 tests), described below.

We assessed change with time for individual endpoints and for composite scores for all diabetic patients and for patients with and without DP. To assess which neurologic sign, test result, or composite score was best for assessing longitudinal change in severity of DP, we compared the frequency of abnormality among different endpoints at the first evaluation; calculated the percent of patients whose polyneuropathy status (present or absent) was unchanged, better, or worse at the 2-year evaluation compared with baseline-a large difference between the percent of patients who worsened from those who improved, reflecting favorably on the endpoint; and assessed monotonicity and magnitude of worsening over time. In the third approach, we were concerned with evaluating the sensitivity of an endpoint to detect change in individual patients over time. It was assumed that the observation of a monotone change(consistently improving or worsening over time) was real and reflected favorably on the test. Conversely, changes that were not monotone (e.g., better-worse-better) most likely represented noise and reflected unfavorably on the test. Spearman's rank correlation coefficient was used as a measure of this sensitivity. For each patient, the time points from earliest to latest and the endpoint measurement from worst to best were ranked, and the correlation coefficient between these ranked values were computed. The mean rank indicated the performance of the measurement in a group of patients. A positive correlation indicated improvement over time; a negative coefficient indicated worsening. Because this measurement was unit-free, it provided a convenient method for comparing various endpoints in their ability to reflect monotone change. In addition to monotonicity, we were also interested in knowing which endpoint provided the greatest magnitude of change. To compare the change in magnitude between different endpoint results, we expressed results as a normal deviate as estimated from the healthy subject RDNS cohort.^20,21

The derivation of the NIS(LL)+7 test score is provided intable 1.

Results. Clinical characteristics of the RDNS. Intable 2, we provide characterizing information about all the diabetic patients (n = 195) and those with polyneuropathy (n = 42) by the criteria of NIS(LL)+7 tests (97.5th percentile). The group studied encompassed a broad range of ages (19 to 88 years), men and women, and both IDDM and NIDDM. The degree of glycemic control and other metabolic derangements varied widely among diabetic patients and among patients with DP.

Frequency of neuropathic abnormality among individual neurologic examinations tests or composite scores. The percent of diabetic patients who were abnormal by various individual neurologic symptoms or signs varied widely (table 3). The highest frequency of clinical abnormality occurred from evaluation of ankle reflexes (no. 29 of NIS, 16.4%) followed by clinical vibration of the toes (no. 36 of NIS, 13.3%). If several neurologic symptoms or signs were combined, higher percentages of abnormality were found (e.g., NSS ≥ 1 = 10.3%, NIS [≥2)] = 27.7%, and NIS(LL)[≥2] = 27.2%).

Considering individual attributes of nerve conduction, VDT, and HB DB, the percentage of patients with abnormality varied widely, but generally the frequency of abnormality was higher for attributes of NC than for individual clinical abnormalities. If abnormality of any attribute of NC was the basis for the diagnosis of DP, even higher frequencies were found.

In table 3, we also show the frequency of DP from use of various composite scores. Based on symptoms elicited by a neurologist (NSS≥ 1), only ∼10% had DP. The patient-answered questionnaire (NSP) produced a higher frequency of abnormality than NSS. Inspection of NSP profiles, however, indicated that not infrequently non-neuropathic symptoms were included with neuropathic symptoms. Based on neurologic impairment (NIS≥ 2), 27.7% had DP; using (NIS(LL) ≥2) 27.2% had DP. Intable 3, we also show the frequency of DP using our previously published minimal criteria for DP criterion 1: ≥2 abnormalities from among the five evaluations, symptoms (NSS ≥ 1); neuropathy impairment (NIS ≥ 2 points); NC abnormality (≥1 abnormal [≥99th or ≤1st percentile, whichever applies] attributed in ≥2 nerves); abnormality (≥99th percentile) of vibration (VDT), cooling (CDT), or heat-pain thresholds (HP:5) or abnormality of quantitative autonomic tests, e.g., abnormality (≤1st percentile) of HB DB²² or criterion 2,¹⁷ ≥2 abnormalities from the five evaluations listed in 1, but ≥1 must be an abnormality of NC or of HB DB. The first criterion will be designated ≥2 abn 5 evals; the second criterion ≥2 abn 5 evals; ≥1 must be NC or HB DB. The frequency of DP using criterion 1 or 2 is similar to what was found from using the 97.5th percentile of the composite score NIS(LL)+7 tests; the composite score, found here, is among the best available.

Percent of RDNS patients in whom polyneuropathy status (present or absent) was unchanged, better, or worse at 2 years compared with baseline using individual or grouped criteria for polyneuropathy. Intable 4, we provide information about change in abnormality of clinical symptoms, neurologic findings, tests, or composite scores indicative of DP. Abnormality of these endpoints are compared at 2 years and at baseline. This information may be used to indicate the course of neuropathy and to indicate the value of an endpoint to recognize change. We assume for these analyses that in a diabetic cohort, there should be worsening of neurologic function. Endpoints that reflect this worsening are useful, and ones that do not show change or improvement are not useful.

As shown in table 4, some individual symptoms(decreased feeling, paresthesia, and pain) as elicited by a neurologist showed worsening, but the composite NSS score showed slight improvement. The patient-answered questionnaire (NSP) showed improvement.

For neuropathic impairment, as had been found for symptoms, not all measures were equal in showing a greater frequency of patients who worsened, as compared with those who improved, among the various endpoints. The best individual measure of worsening was abnormality of ankle reflexes, followed by composite NIS and NIS(LL) and by pinprick of the toe. The frequency of clinical vibration abnormality of the toes actually was lower at 2 years than at baseline (improvement), which is assumed to be an incorrect result because diabetic patients worsen with time, and this result is not what was found using quantitative sensory testing.

QST of VDT as measured by CASE IV and various attributes of NC showed worsening more frequently than improvement. Various composite scores, and especially NIS(LL)+7 tests, showed a much higher frequency of worsening than improvement. Stage of DP also had worsened more frequently than improved(1.1%).

Comparison of the monotonicity and magnitude of worsening among individual tests or composite scores. In table 5, we show the average worsening per year expressed as normal deviate values or points (equivalent to NIS points) for the neuropathic endpoints or composite scores. We do not show endpoints that remained unchanged or improved. Because the endpoint values have been corrected for age, gender, and other physical variables, the change is attributable to DM. The average change is small but in most cases highly significant. Upper limb attributes of NC sometimes showed as large (or even larger) changes than lower limb endpoints. All the endpoints demonstrated a high degree of monotonicity.

Using the composite score NIS(LL)+7 tests, a change of 0.34 points per year was found for all diabetic patients-a greater magnitude than from use of single measures. The NIS(LL)+7 tests provide a high mean r (= 0.337)(p < 0.0001)).

In table 6, we present similar data to that shown intable 5 but for patients from the longitudinal RDNS with DP. A somewhat smaller number of endpoints showed significant worsening. The magnitude of the NIS(LL)+7 tests is even greater (0.85 points) than in the entire cohort (0.34 points).

Sensitivity and specificity of various criteria for DP. Because the score NIS(LL)+7 tests provided a composite measure of the neuropathic impairments of lower limbs (including muscle weakness, reflex loss, sensory loss, abnormality of NC, and decreased HB DB) which impairments occur in DP and provided a monotone worsening with magnitude, it was chosen as the best criterion for the diagnosis of DP. In table 7, we show the sensitivity and specificity of other criteria for DP as compared with NIS(LL)+7 test as the gold standard. Our previously published minimal criteria had sensitivities and specificities of approximately 85%. Decrease or loss of the ankle reflexes alone identified approximately 60% of DP and had a high specificity (∼91%). Clinical vibration alone identified only∼17% of patients with DP but had a very high specificity (∼96%). The clinical neurologic examination alone (NIS(LL)) identified almost 70% of patients and had a reasonable specificity (∼87%). The best nerve conduction criteria was no more than two nerves with NC abnormality, with a sensitivity of 81.0% and specificity of 91.2%.

Estimating power to perform a controlled clinical trial based on 2-year follow-up data of RDNS diabetic patients ≤65 years old with polyneuropathy (≤97.5th NIS(LL)+7). To estimate sample size requirements for a clinical trial using NIS(LL)+7 as the primary endpoint, we consider patients in the RDNS cohort with polyneuropathy and age <65 years. Because data were collected for NIS(LL)+7 at 2-year intervals, sample size calculations are provided below for clinical trials including follow-up at 2- or 4-year intervals.

For the 2-year calculations, 31 subjects had all the necessary measurements, obtained using exactly the same procedures at both baseline and 2 years. The mean and SD of the changes in NIS(LL)+7 were 1.08 and 3.57(omitting an extreme outlier), respectively.

A treatment effect of two NIS points is considered to be a clinically meaningful effect. To detect an effect of this magnitude with a 2-year study and an SD of 3.57, one would require 68 patients in each treatment arm to have a power of 0.90, using a two-sided test at the 0.05 level. If the effect of treatment is to eliminate progression of neuropathy without effecting improvement, these data indicate that a study of approximately 3.7 years is needed.

Using data on subjects with data at baseline and 4 years, the mean and SD of the changes in 29 subjects (no outliers) were 3.28 and 4.74, respectively. Based on these results, we estimate that a 4-year study would require 45 patients per arm to achieve power of 0.90 to detect a treatment effect that eliminates progression of disease and that a clinically meaningful effect can be expected to be observed at approximately 2.4 years.

Based on the calculations provided above, we would conservatively project that a clinical trial involving these types of patients should extend for a period of at least 3 years and include approximately 70 to 100 patients per arm to have a high probability of demonstrating a clinically meaningful effect.

Discussion. The present studies focus primarily on three questions concerning DP. First, what are the best individual endpoints or composite scores for diagnosing DP, for assessing its overall severity, and for assessing worsening with time? Second, what change in overall severity of DP occurs with time in a representative diabetic cohort with and without DP? Third, what number of patients (and for how long) are needed to perform a controlled clinical trial?

In previous descriptive studies, epidemiologic surveys, and intervention trials, investigators do not assess severity of DP, only its presence or absence. Usually, therefore, they declare DP to be present or absent based on clinical findings (e.g., decreased or absent ankle reflexes or decreased vibratory sensation of the feet, elevated VDT, abnormality of NC, or decreased variability of HB DB). Even when symptoms and neurologic examinations were comprehensively evaluated, the information was only used to decide whether polyneuropathy was present or not. There may have been several reasons why DP was not quantitated by a single continuous number. First, in medical practice the emphasis has mostly been on whether or not a medical disease was present-not on its severity. Second, the diagnostic limits of DP, especially its separation from proximal diabetic neuropathy, diabetic small fiber polyneuropathy with weight loss, and uremic polyneuropathy, were not usually made. Third, an approach to scale DP severity was not available.

The need to characterize DP more fully led to the introduction of a variety of tests and recommendations that these tests be used to characterize DP.^13,23 But with the introduction of more evaluations, lack of use of standard tests, different reference values, and use of different criteria for the diagnosis of DP, great variability in estimating DP ensued. To reduce this variability, we^{16,18,19,22,24-26} and others^27,28 have emphasized the use of standard examinations and test reference values based on study of a randomly selected large cohort and defined percentile abnormality and minimal criteria for DP. We proposed two minimal criteria for the diagnosis of polyneuropathy: two abnormal evaluations²² and two abnormal evaluations with one abnormality of NC or HB DB.¹⁵ Compared with results of nerve morphometry, the first minimal criteria seemed to perform reasonably.²² The present study suggests a better minimal criteria: the 97.5th percentile value of a composite score that includes lower limb neurologic abnormalities and seven tests. Other composite scores could be fashioned to emphasize large or small fiber dysfunction.

The present study, like our earlier studies, provides evidence that no one neuropathic symptom, finding, or test is consistently abnormal, so that it may serve as a reliable marker for DP. The use of one marker for DP is intuitively not the best approach to judge for the presence and severity of polyneuropathy. DP is the sum of motor, sensory, and autonomic symptoms, impairments, and outcomes related to dysfunction of peripheral nerves in DM. Simply identifying DP as present or absent by one or two markers is unsatisfactory for several reasons. First, individual markers are too variable a criteria for DP. Second, abnormality of different markers may have different implications for poor health and disease consequences. A slightly abnormal conduction velocity of one nerve has a different implication than severe loss of pain sensation in the foot. Third, as shown here, in cross-sectional diabetic cohorts, the sum total of all of the main dysfunctions provides a better measure of the characteristics and magnitude of DP.

The present studies provide insights about which individual symptoms, impairments, tests, or composite scores provide the most robust and consistent change in severity of DP with time. Among symptoms, loss of feeling, paresthesia, and pain reflected worsening, whereas the composite NSS score did not; it actually improved slightly. The patient-answered questionnaire (NSP) performed poorly, presumably because of the inclusion of symptoms unrelated to DP. In decreasing order, decrease or loss of ankle reflexes, the NIS or NIS(LL) composite scores, and pinprick sensation of the toe reflected worsening of DP, whereas clinical vibration of the toe did not; it actually showed a spurious improvement. Generally, tests of nerve function, particularly VDT and attributes of NC, worsened more frequently than improved. Worsening was also more frequent than improvement using our previously described minimal criteria for DP and using our staging recommendations. The composite score NIS(LL)+7 tests provided not only a broad assessment of weakness, reflex alterations, sensory loss, and test abnormalities but also a monotone worsening of DP with time. Because clinical and test abnormality are combined, it provides the largest point change with time.

Similar composite scores can be fashioned for tests alone (without use of NIS) for large fiber (e.g. NIS(LL)+6 [deletion of HB DB]), for small fiber function (e.g., heat-pain 5, cooling detection threshold, and HB DB), or an even greater number of tests (e.g., NIS(LL)+9 (NIS(LL)+7+CDT+HP5).

There is information on change in severity of DP with time based on longitudinal studies. Pirart^1,2 found that the frequency of polyneuropathy (as judged by decreased ankle reflexes and vibration sensation) among diabetic patients attending Brussels hospitals and diabetic clinics increased with duration of diabetes. He inferred that complications of nerve, kidney, and eye were related and due to chronic hyperglycemia. In a study of 71 type 2 diabetic patients for 5 years, Hillson et al.²⁹ found a slight deterioration of vibratory sensation of the feet, measured by bioesthesiometer. Factors that related to this decrease were initial sensory threshold, age, gender, mean fasting plasma glucose, and failure to lose weight.

Patients in the placebo arm of double-blind controlled clinical trials may provide information on change in measures of diabetic polyneuropathy with time, but results have generally not been consistent. Several reasons may explain this variability: patients are usually highly selected for a given study and different kinds of patients may be selected for different studies; clinical responses may have been biased toward improvement because of a patient's or a physician's expectation that the experimental drug would be efficacious (a placebo effect); or lack of use of standard tests and adequate reference values or composite scores. Florkowski et al.³⁰ randomly assigned 54 type 1 and 2 diabetic patients with symptoms of pain, numbness, or paresthesia into treatment with an aldose reductase inhibitor (ARI) or placebo for a period of 24 weeks. They found no statistically significant worsening of tibial nerve conduction velocity in the placebo arm of the trial. In another ARI study of symptomatic DP, investigators reported symptoms and nerve conduction velocities to have improved in the placebo-treated patients over a 6-month period, whereas vibration sensation at the wrist worsened. The improvement of pain and paresthesia by ∼0.6 points of a 4-point scale in the placebo group is hard to explain except perhaps as a placebo effect. In our series, the frequency of pain or paresthesia was higher with time, not lower. Intuitively, and in light of the present study, the improvement of 5.9± 1.6 m/sec (p < 0.001) in peroneal motor nerve conduction velocity in the placebo group is hard to explain and unlikely to be correct. In the Sorbinil Retinopathy Trial Research Group study,³¹ the frequency of clinically diagnosed polyneuropathy rose from 15% at 1 year to 30% at 4 years. NC worsened, for example, by 4.8 m/sec for median MNCV and by 0.3 m/sec for peroneal MNCV-the difference is hard to understand. Comparable values for CMAP were 1.0 and 1.1 mV. In another ARI trial,³² investigators found small degrees of worsening for cardiovascular autonomic test results.^32,33 Because the values in these latter three studies were not corrected for age and physical variables, one should not assume that all of the worsening was due to diabetes mellitus, but at least the direction of the change and its magnitude are similar to what we found.

Information about change in polyneuropathy over time also came from intervention trials of glycemic control. In a study from Oslo, Norway,³⁴ 45 IDDM patients, 18 to 42 years old, were assigned into three treatment groups using insulin (infusion by pump, multiple injections, or twice daily injections [conventional treatment]). The trial was for 4 years and thereafter the patient could choose their treatment for the next 4 years. These investigators compared nerve conduction values after 8 years and found a striking difference in worsening of neuropathic endpoints for the poor glycemic control group. To illustrate, whereas the tibial MNCV had decreased by 3.9 m/sec in patients whose average HbA_1C had been controlled below 10%, it had decreased by 6.8 m/sec for the group whose HbA_1C had averaged ≥10%. The effect appeared to be greater in lower than in upper limb nerves. Likewise, they found a significantly greater decrease of HB DB or HB response to standing in patients whose average HbA_1C was ≥10%. In the Diabetes Control and Complications Trial,^7,35 only IDDM patients without or with mild background retinopathy were included for longitudinal study. The percentage of patients who developed "evident polyneuropathy" and abnormalities of nerve conduction were significantly lower in the near euglycemic group than in the conventionally treated group. They did not estimate severity or stage of DP overall.

The present data provide an improved estimate of the change in severity of DP with time because it is based on evaluation of a population-based cohort that is thought to be representative of community diabetic patients mainly of northern European extraction; comprehensive and standard neuropathic assessment of symptoms, neuropathy impairments, nerve conductions, quantitative sensory tests, and quantitative autonomic tests were used to prospectively evaluate patients at yearly intervals; test abnormalities were estimated as a percentile considering test, site, age, and physical variables so that observed change with time should be due to diabetes mellitus and not to other anthropomorphic features; and composite scores of impairment were used.

Based on these studies, the change in DP with time is small but measurable, and it is greater in diabetic patients with polyneuropathy. The score should be very useful in epidemiologic studies for assessment of risk factors. It should also be an excellent composite score for use in controlled clinical trials.

The data we have developed provide an improved insight about the numbers of patients and duration of study needed for an intervention trial. Selecting patients with DP who are less than 65 years old, it should be possible to achieve statistical significance and a difference of ≥2 NIS points with 70 persons in each arm of a trial in a period of 3 years.

Acknowledgments

We gratefully acknowledge the help of the RDNS patients, Mrs. Kay Kratz, Miss Karen Lehman, Mrs. Jeannine Karnes, Mrs. Carol Overland, and F. John Service, MD.