Emerging Treatments and T e c h n o l o g i e s ORIGINAL AR T I C L E

Pa tte r ns of Quantitative Sensation Testing of Hypoesthesia and Hy p e r a l g e s i a Are Predictive of Diabetic Polyneurop a t h y A study of three cohorts

PETER J. DYCK, MD TIMOTHY S. LARSON, MD C O N C L U S I O N S — 1) D e c reased vibra- P. JAMES B. DYCK, MD PETER C. O’BRIEN, PHD to r y sensation (hypoesthesia) appears to be JORGE A. VELOSA, MD TH E NERV E GRO W T H FAC T O R STU D Y GRO U P characteristic of mild DPN, whereas pan- modality hypoesthesia is characteristic of se v e r e DPN. 2) A shift of vibratory and cold detection thresholds (and also of attributes of n e rve conduction and a measure of auto- OB J E C T I V E — To test quantitative sensation testing (QST) patterns of hypoesthesia and nomic dysfunction) from low-norm a l hyperalgesia as indicators of diabetic polyneuropathy (DPN) and its severity. (2.5th–50th percentiles) to high-norm a l (50th–97.5th percentiles) appears to prec e d e RESEARCH DESIGN AND METHODS — We used Computer-Assisted Sensory Exam- ov e r t expression of DPN and to thereby pro- ination IV; characterized the QST results of the foot of each patient in three diabetic cohorts vide evidence of subclinical abnorma l i t y . 3) ( 1,500 patients) as hyperesthetic ( 2.5th percentile), low-normal (2.5th–50th perce n t i l e s ) , Heat stimulus–induced hyperesthesia (low hi g h - n o r mal (50th–97.5th percentiles), or hypoesthetic ( 97.5th percentile); and tested associ- th r esholds) occurs especially in mild DPN, ations with symptoms, impairments, and test abnorma l i t i e s . and, because it correlates with DPN symp- toms and impairments, it must be attributed RE S U LT S — Overall neuropathic impairment was most severe in the pancrea s - r enal transplant to hyperalgesia rather than to supersensitivity. and nerve growth factor cohorts, but it was much less severe in the population-based Rochester Th e re f o r e, hypoalgesia or hyperalgesia may be Diabetic Neuropathy Study (RDNS) cohort. The frequency distribution of sensory abnorma l i t i e s an indicator of early DPN. mi r ror ed this diffe r ence. When the QST spectra of diabetic cohorts were compared with those of the control subject cohort for vibration and cooling sensations, the only abnormality observed was Diabetes Care23 :5 1 0 –517, 2000 hypoesthesia, which was expressed as an increased number of subjects with values at or above the 97.5th percentile or by an increased percentage of cases with high-normal values. Symptoms and uantitative sensation testing (QST) im p a i r ments of DPN were significantly more frequent in the subjects with values at or above the has been used to assess the effects of 97.5th percentile than in the subjects whose values were between the 50th and 97.5th perce n t i l e s . For heat pain (HP) sensation thresholds (intermediate pain severity [HP:5], pain thres h o l d Qmaturation, aging, anatomical loca- [HP:0.5], and pain-stimulus response slope [HP:5–0.5]), an increased frequency of both hypoal- tion, anthropometric characteristics, and gesia and hyperalgesia was observed (especially in the RDNS cohort). Steeper pain-stimulus disease on thresholds of diffe r ent modalities response slopes were significantly associated with sensory symptoms, including severity of pain. of cutaneous sensation (1,2). These tech- niques have been used to also study the physiological characteristics of sensory receptors and such dysfunctions as stimu- Fr om the Peripheral Neuropathy Research Center (P.J.D., P.J.B.D.), the Nephrology/Dialysis Transplant Center lus-induced hyperalgesia and pain (3–8). (J . A . V ., T.S.L.), and the Department of Health Sciences Research (P.C.O.), Section of Biostatistics, the Mayo Clinic These approaches were used to show also and Mayo Foundation, Rochester, Minnesota. Ad d r ess correspondence and reprint requests to Peter J. Dyck, MD, 200 First St., S.W., Rochester, MN 55905. that vibratory sensation persisted for a E-mail: [email protected]. longer time after application of a tourni q u e t Received for publication 14 September 1999 and accepted in revised form 4 January 2000. in diabetic subjects than in control subjects P.J.D. and P.C.O. receive royalties from the sale of CASE IV systems, which they donate in part (P.C.O.) or (3). This resistance to ischemic conduction in whole (P.J.D.) to charitable proj e c t s . Ab b re v i a t i o n s : CASE IV, Computer-Assisted Sensory Examination IV; CDT, cooling detection thres h o l d ; block is now usually attributed to the grea t e r DPN, diabetic polyneuropathy; HBDB, heart beat response to deep breathing; HP, heat pain; NC, nerve con- availability of energy substrate in diabetic duction; NIS, Neu r opathy Imp a i r ment Sco r e; NIS(LL), NIS for impairment of lower limb; NIS(LL) 4, NIS(LL) ne r ves (9). Increa s i n g l y , QST approa c h e s su b s c o r e with four nerve tests; NIS(S), NIS for sensation; NIS(W), NIS for weakness; NSC, Neu r opathy Sym p - ar e being used in epidemiological studies toms and Change; NSC[A], NSC subscore for autonomic dysfunction; NSC[LF], NSC subscore for large - d i a m - and controlled clinical trials (10–12), and eter fiber sensation; NSC[P], NSC subscore for pain threshold; NSC[SF], NSC subscore for small-diameter fiber sensation; NSC[W], NSC subscore for weakness; PRT, pancrea s - r enal transplant; QST, quantitative sensation they have been introduced into medical testing; RDNS, Rochester Diabetic Neuropathy Study; rhN G F , recombinant human nerve growth factor; VDT, practice (13). In parti c u l a r , QST has been vi b r a t o r y detection thres h o l d . used to detect and characterize sensation The names of the investigators of the NGF Study Group will be listed in a scientific article that is being pre- alteration in various human polyneu- pa r ed for publication. Schwartz S, et al.: A phase III multicenter double-blind placebo-controlled study of the effi - cacy and safety of recombinant human nerve growth factor in subjects with diabetic polyneurop a t h y . ropathies, including diabetic polyneurop a - A table elsewhere in this issue shows conventional and Système International (SI) units and conversion thy (DPN) (14–16), and this informa t i o n factors for many substances. has recently been reviewed (17).

510 DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 Dyck and Associates

between participating and nonparti c i p a t - ing patients. The cohort was thought to be rep r esentative of north e r n U.S. cities of p redominantly Caucasian extraction. Ap p r oximately one-half of the patients in this population had DPN, but such cases we r e usually mild, as described below. The rhNGF cohort was rec r uited for trials of rhNGF therapy (Genentech, San Francisco, CA) as treatment for DPN. Patients 65 years old with stable dia- betes, DPN (asymptomatic or sympto- matic), and an abnormal ( 9 0 t h pe r centile) cold detection or HP:5 thres h - olds (albeit, one of the two thresholds had to be obtained) were rec r uited. This cohort consisted of 1,019 diabetic patients from 84 medical centers. The diabetic PRT cohort consisted of 78 patients with type 1 diabetes who were 40-years-old, had end-stage diabetic Figure 1—Frequency distribution of severity of DPN as measured by NIS(LL)4 tests for the three ne p h r opathy (and other diabetic complica- diabetic cohorts and the control cohort. The PRT and rhNGF cohorts tend to have more severe- tions),neu and had undergone pancreas and ropathy than the RDNS cohort. The QST abnormalities rmirored the severity of neuropathy shown inrenal transplantation. Patients with symp- the figure. NS, normal control subjects. tomatic coron a r y and peripheral vascular disease were excluded. Test results were expressed as per- We introduced interactive prog r a m m e d po l y n e u r opathy (as measured by the per- centiles and standard deviates, based on a co m p u t e r- c o n t r olled systems to evaluate cent having defined levels of abnorma l i t y ) study of 330 randomly selected and neu- touch pres s u r e, vibration, cooling, and heat and spectra of QST abnormalities permi t - rologically examined subjects fro m pain (HP) thresholds by use of defined, ted an assessment of how well QST res u l t s R o c h e s t e r, Minnesota, who showed no st a n d a r dized, and validated stimuli and reflect overall neuropathy severity. Com- signs of diabetes, neurological disease, or pre p r ogrammed algorithms of testing and parisons of the frequency of impairme n t s diseases predisposing to neuropathy (con- finding thresholds (18,19). We estimated or symptoms between those groups with tr ol subject cohort) (1,2) re f e rence values that were specific for and without hypoesthesia (decreased or mo d a l i t y , site, age, sex, and other anthro- absent sensitivity) or hypere s t h e s i a CASE IV system pometric characteristics so that the res u l t s (i n c r eased sensitivity) allowed inferen c e s The CASE IV system was used to study the could be expressed as a percentile value or to be made about the value of QST in co n t r ol subjects, the RDNS cohort, and the as a standard deviate ( see RE S E A R C HD E S I G N detecting and characterizing DPN. PR T cohort. The system was used to obtain A N DM E T H O D S) value (1,2). The use of such vi b r a t o r y detection thresholds (VDTs), cool- specific percentile values permitted a valid RESEARCH DESIGN AND ing detection thresholds (CDTs), and HP comparison of results for an individual at ME T H O D S th r esholds (intermediate pain [HP:5], pain di ff e r ent ages, because age and anthrop o - t h reshold [HP:0.5], and pain-stimulus metric characteristics are considered when Cohorts studied response slope [HP:5–0.5]). The system pe r centile values are calculated. By using We evaluated QST results and neurop a t h i c consists of a personal computer, control l i n g ac r oss-test comparisons, we found that sen- im p a i r ments, symptoms, and test res u l t s el e c t r onics, transducers, a cueing device, a sation loss was correlated with clinical, (at first evaluation) in three diabetic cohorts : ke y b o a r d, and ancillary devices, as prev i - ne r ve conduction (NC), and morphometric the Rochester Diabetic Neuropathy Study ously described (18). The stimulus wave- me a s u r es of biopsied nerve (15,20). Vib r a - (RDNS) cohort, the recombinant human fo r m, the null stimulus, the 4-, 2-, and to r y sensation of the toe, using the Com- ne r ve growth factor (rhNGF) cohort, and 1-stepping with null stimuli (for VDTs and pu t e r -Assisted Sensation Examination IV the pancrea s - r enal transplant (PRT) cohort. CD T s), and the nonrepeating ascending (CASE IV) and the 4-, 2-, and 1-stepping The RDNS cohort consisted of all of the with null stimuli (for HP tests) algorithms with null stimuli algorithm, perf o rm e d consenting diabetic patients from the city of have been previously described and vali- favorably when compared with other nerve Ro c h e s t e r , Minnesota, who did not have a dated (19). Patient instructions by use of tests in assessing the changes over time in confounding neurological disease and who cueing cards were standard. A commerci a l the severity of DPN (19). we r e participating in a cross-sectional and model, the Computer-Assisted Sensation In the present study, we compare QST longitudinal study of diabetic complications Evaluator IV (WR Medical Electronics, Still- results in three diabetic cohorts with quite (21). For patients 70-years-old, comor- wa t e r , MN), was used in the rhNGF trial. d i ff e rent severities of polyneuro p a t h y. bidity (as evaluated from coded discharge The design of the WR system was modeled Comparisons of the spectra of severity of diagnoses) was not significantly diffe re n t after our system and uses the same stimuli,

DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 511 Hypoesthesia and hyperalgesia in diabetic polyneuropathy algorithms of testing, and finding thres h o l d and ref e r ence values. Calibration of the ver- tical displacement was calibrated for both systems with the vibrating tactator res t i n g on a silicone tube (11/16” OD, 1/2” ID, sh o r e A 50) by use of laser triangulation (Keyence, Osaka, Japan). The thermo d e was calibrated for absolute temperature by using a high-purity platinum re s i s t o r (Goodrich, Minneapolis, MN) interposed between the thermode and a Styrof o a m block at controlling temperatures of 10 and 40°C for periods of 15 min at each temper- at u r e. Rapid fluxes (e.g., the 4°C/s rise from baseline and ret u r n to baseline) were mon- it o r ed with bead thermi s t o r , thermo c o u p l e , or infrared thermo m e t e r . Variation over the su r face of the thermoelectric unit was mon- it o r ed with cholesterol ester mixtures. The vi b r a t o r y tactator rested on the skin with a constant baseline load. The baseline tem- pe r a t u r e for the CDT was 30°C, and the baseline temperature for the HP thres h o l d was 34°C. For the VDT, CDT, and the HP th r eshold, 25 exponentially increasing steps of stimulus magnitude are standard. QST results were monitored in a central rea d i n g la b o r a t o r y to ensure that 1) the correct site, algorithm, and conditions of testing were used; 2) practice and algorithm thres h o l d s we r e within two steps of each other (except for HP, for which a practice sequence was not used); 3) the stimulus response pattern did not show inattention or drowsiness; 4) s u fficient turn - a rounds to estimate th r eshold had been attained; 5) the order of testing by modality was correct; and 6) suf- ficient trials had been used to find thres h - old. If the test did not pass quality assurance criteria, then it was rep e a t e d . For VDTs or CDTs, the quantity esti- mated was the stimulus step that was felt 50% of the time. For HP thresholds, we esti- mated threshold (HP:0.5), an interme d i a t e Fi g u r e 2—Each patient’s QST threshold and each of the three modalities tested were expressed as a per - severity of HP (HP:5), and the pain-stimu- centile specific for test, site, age, sex, and anthropometric characteristics. The frequency distributions by cat - lus response slope (HP:5–0.5). QST res u l t s egories of percentile abnormality are shown for each of the studied cohorts. The characteristics, associations, ar e expressed as noticeable diffe r ence step of and inferences that were derived from these data are described in text. NS, normal control subjects. stimulus magnitude (1 to 25), the corre- sponding physical measurement (µm of dis- placement or the change from 0°C), and the co r responding percentile or standard devi- [NSC]), attributes of NC, and the heart- ex p r essed as a composite score (points). ate to make appropriate corrections for beat response to deep breathing (HBDB) The examiner graded abnormality of each mo d a l i t y , anatomical site, age, sex, and we r e compared with the QST results of item of the NIS specific for age, sex, phys- applicable physical characteristics. each patient. In NIS muscle weakness, ical fitness, and other anthro p o m e t r i c reflex diminution or loss and sensation characteristics. NIS subscores for impair- Other neuropathy tests diminution or loss were summated based ment of lower limb (NIS[LL]), weakness S t a n d a rdized measurements of neuro- on a standard evaluation of a pred e t e r - (NIS[W]), and sensation (NIS[S]) were pathic impairment (Neuropathy Imp a i r - mined group of muscles, muscle stret c h derived. We also used a composite score ment Score [NIS]), neuropathic symptoms reflexes, and sensation modalities of index that included the results of the NIS(LL) (Ne u ropathy Symptoms and Change fingers and great toes with abnorma l i t y and four nerve tests (NIS[LL] 4), which

512 DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 Dyck and Associates

Table 1—Average composite severity of DPN of patients in the RDNS,h NrG F , and PRT cohorts severity of DPN was re p resented by the points of the NIS(LL) 4 tests (20). VD T CD T 50th to 97 . 5 t h 97 . 5 t h 50th to 97 . 5 t h 97 . 5 t h Analysis n Me a n n Me a n P n Me a n n Me a n P To compare diffe r ent test results among di ff e r ent modalities of sensation and attrib- RD N S utes of NC, we used percentile values NIS(LL) 4 tests 12 6 3. 3 6 36 12 . 4 6 0. 0 0 1 13 3 4. 4 4 16 17 . 0 9 0. 0 0 1 derived by study of the control cohort (2), NI S ( L L ) 12 6 1. 4 9 36 8. 2 1 0. 0 0 1 13 3 2. 2 6 17 11 . 5 3 0. 0 0 1 as previously described (1). Thus, it was Lo w e r -limb NIS(W) 12 6 0. 1 5 36 2. 1 0 0. 0 0 1 13 3 0. 2 2 17 3. 7 6 0. 0 0 1 possible to express each of the test results as Lo w e r -limb NIS(S) 12 6 0. 4 8 36 3. 0 8 0. 0 0 1 13 3 0. 8 9 17 4. 0 0 0. 0 0 1 a specific percentile value that considered rhN G F test, site, age, sex, and applicable anthro- NIS(LL) 4 tests 46 6 15 . 7 3 51 4 20 . 0 6 0. 0 0 1 69 1 16 . 8 8 31 9 19 . 8 5 0. 0 0 1 pometric characteristics. A standard deviate NI S ( L L ) 46 7 12 . 0 1 51 6 14 . 9 9 0. 0 0 1 69 4 12 . 8 8 31 9 14 . 6 2 0. 0 0 1 value relates each percentile value to a nor- Lo w e r -limb NIS(W) 46 7 1. 5 2 51 6 2. 0 1 0. 0 0 8 69 4 1. 6 4 31 9 2. 0 0 0. 0 2 6 mal distribution. For example, for the 50th Lo w e r -limb NIS(S) 46 7 7. 4 4 51 6 8. 8 6 0. 0 0 1 69 4 8. 0 2 31 9 8. 3 4 0. 3 1 0 p e rcentile, the corresponding standard PR T deviate is 0, because 50% of the values in a NIS(LL) 4 tests 12 9. 9 2 43 22 . 2 8 0. 0 0 1 11 13 . 4 5 41 22 . 7 3 0. 0 1 2 no r mal distribution fall below 0. The stan- NI S ( L L ) 12 4. 5 0 43 13 . 6 6 0. 0 0 1 11 6. 7 3 41 14 . 4 3 0. 0 1 2 da r d deviate corresponding to the 95th Lo w e r -limb NIS(W) 12 1. 6 7 43 5. 4 5 0. 0 3 4 11 3. 1 8 41 5. 7 7 0. 1 9 0 pe r centile is 1.96, because 95% of the val- Lo w e r -limb NIS(S) 12 0. 1 7 43 4. 0 2 0. 0 0 1 11 0. 9 1 41 4. 1 5 0. 0 0 7 ues in a normal distribution fall below 1.96. The mean neuropathic impairme n t s HP : 5 HP : 5 – 0 . 5 (as measured by NIS, NIS[LL], NIS[LL] 50 t h to 97 . 5 t h 97 . 5 t h 50 t h to 97 . 5 t h 97 . 5 t h 4, NIS[LL][W], or NIS[LL][S]) was com- n Me a n n Me a n P n Me a n n Me a n P pa r ed among patients with diffe r ent cate- gories of sensation abnormality by use of RD N S two-sided Wil c o x o n ’ s rank-sum tests. NIS(LL) 4 tests 61 5. 6 1 16 14 . 3 1 0. 0 0 4 10 2 5. 5 2 8 4. 5 0 0. 6 5 9 NI S ( L L ) 61 2. 8 0 16 10 . 5 6 0. 0 0 2 10 2 3. 0 5 8 2. 5 0 0. 4 1 5 RE S U LT S Lo w e r -limb NIS(W) 61 0. 4 3 16 3. 6 9 0. 0 2 7 10 2 0. 6 3 8 0 0. 5 8 2 Lo w e r -limb NIS(S) 61 0. 9 2 16 4. 2 5 0. 0 0 1 10 2 1. 1 3 8 1. 0 0 0. 4 8 7 The severity of DPN among the rhN G F cohorts studied NIS(LL) 4 tests 37 8 16 . 1 3 38 9 20 . 2 1 0. 0 0 1 57 6 17 . 5 5 19 0 19 . 3 0 0. 0 4 1 Th e r e was a striking diffe r ence in the sever- NI S ( L L ) 37 9 12 . 0 9 39 0 15 . 4 0 0. 0 0 1 57 8 13 . 3 0 19 1 14 . 6 2 0. 0 9 4 ity spectra of DPN, as measured by NIS(LL) Lo w e r -limb NIS(W) 37 9 1. 1 8 39 0 2. 3 4 0. 0 0 1 57 8 1. 6 5 19 1 2. 4 4 0. 4 3 2 4 tests, among the three disease cohorts Lo w e r -limb NIS(S) 37 9 7. 5 4 39 0 9. 1 4 0. 0 0 1 57 8 8. 1 3 19 1 8. 3 4 0. 5 4 1 studied, as compared with the contro l PR T co h o r t (Fig. 1). In the RDNS cohort, only NIS(LL) 4 tests 15 21 . 0 0 11 25 . 2 7 0. 3 1 1 23 22 . 3 5 0 — — 35% of the subjects had DPN and only a few NI S ( L L ) 15 12 . 7 3 11 16 . 5 9 0. 2 7 4 26 14 . 9 7 0 — — patients had NIS(LL) 4 test values 20 Lo w e r -limb NIS(W) 15 5. 3 3 11 5. 6 8 0. 5 0 7 26 5. 8 6 0 — — points, which indicates severe neurop a t h y . Lo w e r -limb NIS(S) 15 2. 6 7 11 7. 0 0 0. 0 0 9 26 4. 8 1 0 — — By contrast, virtually all of the patients in the Data were attained by use of Wil c o x o n ’ s rank-sum test. The data for HP:0.5 (not shown) were similar to the rhNGF and PRT cohorts had DPN, and data for HP:5. 40% of each cohort had NIS(LL) 4 test values 20 points. DPN in the PRT cohort we r e conducted by proc e d u r es prev i o u s l y cases, only symptoms occurring in the was somewhat more severe than that in the outlined (20). The four tests consisted of lower limbs were used. rhNGF cohort. the peroneal nerve motor nerve conduc- tion velocity, the compound muscle action QST and DPN severity spectra The severity spectrum of QST potential, the motor nerve distal latency, Each patient’s QST results, by modality of abnormalities and the tibial distal latency tests. The NSC sensation, were categorized as being at or The severity spectrum of QST results for sc o r e is obtained through a standardi z e d below the 2.5th percentile (hypere s- VD T s, CDTs, and HP thresholds of the physician-completed questionnaire that is thetic), between the 2.5th and 50th per- foot are shown in Fig. 2 for the control used to judge the number, severity, and centiles (low-normal), between the 50th subjects and the three disease cohort s . change of symptoms. Subscores on the and 97.5th percentiles (high normal), and Results for HP:0.5 are not shown. VDTs severity of the NSC were also obtained to at or above the 97.5th percentile (hypoes- and CDTs were shifted from left ( 50 t h me a s u r e weakness (NSC[W]), sensation thetic). This categorization was decided p e rcentile) to right ( 50th perc e n t i l e ) , (NSC[S]), autonomic dysfunction on a priori without assessing other limits. resulting in more patients in the (NSC[A]), large-diameter fiber sensation Small percentiles represent small-magni- 50th–97.5th percentile category than in (NSC[LF], small-diameter fiber sensation tude stimuli, and large percentiles repre- the 2.5th–50th percentile category, with (NSC[SF]), and pain (NSC[P]). In all sent large-magnitude stimuli. O v e r a l l an increase in patients with unequivocal

DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 513 Hypoesthesia and hyperalgesia in diabetic polyneuropathy hypersensitivity or insensitivity ( 97 . 5 t h Table 2—Average NSC scores of severity of patients in the RDNS,h N r G F , and PRT cohorts pe r centile). In the RDNS, the VDTs were much more frequently abnorm a l VD T CD T ( 97.5th percentile) than they were for 50 t h to 97 . 5 t h 97 . 5 t h 50 t h to 97 . 5 t h 97 . 5 t h the CDTs or HP:5 thresholds. By compar- n Me a n n Me a n P n Me a n n Me a n P ison, for the PRT cohort, the frequency of VDT and CDT abnormality ( 97.5th per- RD N S NSC severity 12 6 0. 4 8 36 1. 5 6 0. 0 6 8 13 3 0. 5 1 17 1. 7 1 0. 1 1 6 centile) were approximately equal for the NS C ( W ) 12 6 0. 0 2 36 0. 2 8 0. 0 1 0 13 3 0. 0 7 17 0. 2 4 0. 3 8 4 rhNGF cohort. The pattern was approx i - NS C ( S ) 12 6 0. 2 2 36 0. 9 4 0. 0 0 1 13 3 0. 3 2 17 0. 8 8 0. 0 9 1 mately intermediate between the RDNS NS C ( A ) 12 6 0. 2 3 36 0. 3 3 0. 8 4 7 13 3 0. 1 3 17 0. 5 9 0. 0 7 0 and PRT cohorts. For the HP thres h o l d , NS C ( L F ) 12 6 0. 1 3 36 0. 5 0 0. 0 0 3 13 3 0. 1 7 17 0. 4 1 0. 1 5 9 the pattern of abnormality was strikingly NS C ( S F ) 12 6 0. 0 5 36 0. 2 8 0. 0 9 4 13 3 0. 0 6 17 0. 4 7 0. 0 7 9 di ff e r ent than that described for the VDT NS C ( P ) 12 6 0. 0 5 36 0. 1 7 0. 0 9 8 13 3 0. 0 9 17 0 0. 3 7 8 and CDT (Fig. 2). No t a b l y , in the RDNS rhN G F co h o r t, and to a lesser degree in the PRT NSC severity 46 7 8. 8 1 51 6 10 . 5 3 0. 0 0 1 69 4 9. 2 4 31 9 10 . 5 6 0. 0 0 1 co h o r t, abnormality (i.e., an increased fre- NS C ( W ) 46 7 0. 2 0 51 6 0. 2 5 0. 9 9 0 69 4 0. 2 2 31 9 0. 2 7 0. 8 9 4 quency of subjects with hypoalgesia and NS C ( S ) 46 7 6. 6 4 51 6 8. 4 1 0. 0 0 1 69 4 7. 1 1 31 9 8. 3 9 0. 0 0 1 hyperalgesia) at both ends of the spectrum NS C ( A ) 46 7 1. 9 6 51 6 1. 8 7 0. 2 5 0 69 4 1. 9 1 31 9 1. 9 0 0. 8 0 2 was seen. In the RDNS cohort, there were NS C ( L F ) 46 7 1. 6 0 51 6 2. 1 6 0. 0 0 1 69 4 1. 8 2 31 9 2. 0 1 0. 2 1 0 mo r e patients who had HP:5 in the low- NS C ( S F ) 46 7 1. 2 1 51 6 2. 0 0 0. 0 0 1 69 4 1. 4 3 31 9 2. 0 1 0. 0 0 1 no r mal than in the high-normal range. By NS C ( P ) 46 7 3. 8 2 51 6 4. 2 2 0. 1 8 2 69 4 3. 8 4 31 9 4. 3 4 0. 0 7 0 contrast, in the other two disease cohorts , PR T t h e re were more patients whose HP:5 NSC severity 12 1. 6 7 45 4. 0 9 0. 0 8 3 12 0. 5 8 44 3. 8 2 0. 0 4 2 NS C ( W ) 12 0. 3 3 45 1. 0 4 0. 4 0 7 12 0. 0 8 44 0. 7 5 0. 2 3 2 we r e in the high-normal than in the low- NS C ( S ) 12 1. 0 0 45 2. 5 3 0. 0 5 6 12 0. 2 5 44 2. 4 8 0. 0 3 2 no r mal range. NS C ( A ) 12 0. 3 3 45 0. 5 1 0. 6 8 0 12 0. 2 5 44 0. 5 9 0. 5 7 8 NS C ( L F ) 12 0. 1 7 45 0. 7 6 0. 2 1 6 12 0 44 0. 7 7 0. 0 5 9 The association of QST hypoesthesia NS C ( S F ) 12 0 45 0. 8 7 0. 0 3 0 12 0. 2 5 44 0. 8 9 0. 2 5 9 with NIS NS C ( P ) 12 0. 7 5 45 1. 2 7 0. 1 8 6 12 0. 0 8 44 1. 3 0 0. 1 1 5 As shown in Table 1, mean severity of DPN, as measured by the NIS(LL) 4 tests, the NIS(LL), the lower- l i m b NIS(W), or the lower-limb NIS(S) sub- HP : 5 HP : 5 – 0 . 5 sc o r es, was consistently greater in patients whose QST thresholds were at or above 50 t h to 97 . 5 t h 97 . 5 t h 50 t h to 97 . 5 t h 97 . 5 t h the 97.5th percentile than in those n Me a n n Me a n P n Me a n n Me a n P patients whose thresholds were between RD N S the 50th and 97.5th percentiles for VDT, NSC severity 61 0. 8 2 16 1. 8 1 0. 6 7 1 6 10 2 0. 7 5 8 0. 6 3 0. 6 8 0 CDT, and HP:5 in all three cohorts. Usu- NS C ( W ) 61 0. 1 1 16 0. 3 1 0. 5 7 4 3 10 2 0. 0 4 8 0 0. 8 0 6 ally, the differences were statistically sig- NS C ( S ) 61 0. 2 6 16 1. 2 5 0. 2 7 0 1 10 2 0. 4 3 8 0. 6 3 0. 9 5 2 nificant. With only a few exceptions, NS C ( A ) 61 0. 4 4 16 0. 2 5 0. 8 5 5 0 10 2 0. 2 8 8 0 0. 3 6 2 mean symptom scores were larger when NS C ( L F ) 61 0. 1 1 16 0. 6 3 0. 1 0 9 9 10 2 0. 2 5 8 0. 2 5 0. 9 0 7 QST results were at or above the 97.5th NS C ( S F ) 61 0. 1 3 16 0. 3 1 0. 1 5 0 9 10 2 0. 1 0 8 0 0. 6 3 9 percentile than when they were between NS C ( P ) 61 0. 0 2 16 0. 3 1 0. 0 4 5 1 10 2 0. 0 8 8 0. 3 8 0. 2 5 1 the 50th and 97.5th percentiles (Ta b l e rhN G F 2). In 20 of 63 comparisons, symptoms NSC severity 37 9 9. 1 2 39 0 10 . 7 4 0. 0 0 1 57 8 9. 4 7 19 1 10 . 3 8 0. 1 6 6 w e re significantly more severe in the NS C ( W ) 37 9 0. 1 6 39 0 0. 3 2 0. 0 0 5 57 8 0. 2 4 19 1 0. 2 7 0. 9 0 0 groups with thresholds at or above the NS C ( S ) 37 9 7. 0 1 39 0 8. 4 1 0. 0 0 1 57 8 7. 3 1 19 1 8. 3 1 0. 0 5 0 97.5th percentile vs. those groups with NS C ( A ) 37 9 1. 9 4 39 0 2. 0 1 0. 7 6 6 57 8 1. 9 1 19 1 1. 8 0 0. 6 1 0 NS C ( L F ) 37 9 1. 6 4 39 0 2. 2 2 0. 0 0 1 57 8 1. 8 6 19 1 1. 9 6 0. 2 8 3 thresholds between the 50th and 97.5th NS C ( S F ) 37 9 1. 3 0 39 0 2. 1 0 0. 0 0 1 57 8 1. 8 5 19 1 1. 8 5 0. 0 2 8 pe r centiles. A statistically significant asso- NS C ( P ) 37 9 4. 0 5 39 0 4. 0 7 0. 5 9 0 57 8 3. 8 4 19 1 4. 5 0 0. 1 0 0 ciation was observed most fre q u e n t l y PR T with the following symptoms: NSC(SF) (n NSC severity 16 5. 5 0 13 6. 3 1 0. 5 6 4 30 5. 0 7 3 11 . 3 3 0. 2 6 1 = 5), NSC severity (n = 5), NSC(LF) (n = NS C ( W ) 16 1. 2 5 13 1. 0 8 0. 8 4 4 30 0. 9 7 3 3. 3 3 0. 1 4 5 4), NSC(S) (n = 3), NSC(W) (n = 2), and NS C ( S ) 16 3. 2 5 13 4. 3 1 0. 1 8 8 30 3. 2 0 3 8. 0 0 0. 2 3 4 NSC(P) (n = 1). For autonomic symp- NS C ( A ) 16 1. 0 0 13 0. 9 2 0. 9 5 3 30 0. 9 0 3 0 0. 4 8 2 toms, there were no significant associa- NS C ( L F ) 16 1. 0 0 13 1. 1 5 0. 4 4 6 30 1. 0 3 3 3. 0 0 0. 1 3 5 tions. Of the 63 comparisons, none NS C ( S F ) 16 0. 6 9 13 2. 1 5 0. 0 0 3 30 1. 1 3 3 2. 6 7 0. 2 4 9 showed symptoms significantly greater in NS C ( P ) 16 1. 9 4 13 2. 5 4 0. 3 3 8 30 2. 1 7 3 11 . 3 3 0. 1 1 8 the high-normal group, as compared with Data were attained by use of Wil c o x o n ’ s rank-sum test. The data for HP:0.5 (not shown) were similar to the the hypoesthetic group. data for HP:5.

514 DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 Dyck and Associates

Table 3—Average composite severity of impairments and symptoms of DPN of patients in the RDNS cohort by HP thresholds

HP : 5 HP : 0 5 HP : 5 – 0 . 5 2. 5 t h 2. 5 t h – 5 0 t h 2. 5 t h 2. 5 t h – 5 0 t h 2. 5 t h 2. 5 t h – 5 0 t h n Me a n n Me a n P n Me a n n Me a n P n Me a n n Me a n P Im p a i rm e n t s NIS(LL) 4 tests 31 2. 9 0 98 2. 6 3 0. 8 9 9 10 2. 1 0 14 1 2. 8 9 0. 9 0 9 17 4. 4 1 79 3. 1 0 0. 3 5 6 NI S ( L L ) 31 1. 1 6 98 1. 2 9 0. 2 5 6 10 1. 0 0 14 1 1. 3 3 0. 8 1 8 17 3. 0 6 79 1. 5 0 0. 2 3 0 Lo w e r -limb NIS(W) 31 0. 2 9 98 0 0. 0 1 2 10 0 14 1 0. 0 6 0. 7 2 0 17 0. 7 6 79 0. 2 1 0. 0 3 3 Lo w e r -limb NIS(S) 31 0. 4 2 98 0. 3 8 0. 9 6 3 10 0. 2 0 14 1 0. 4 5 0. 7 7 0 17 1. 3 5 79 0. 3 5 0. 0 1 4 Sy m p t o m s NSC severity 31 0. 3 9 98 0. 4 7 0. 5 6 8 10 1. 0 0 14 1 0. 4 8 0. 2 8 7 17 1. 3 5 79 0. 4 1 0. 1 2 2 NS C ( W ) 31 0. 0 3 98 0 0. 0 7 8 10 0 14 1 0. 0 1 0. 8 1 1 17 0. 3 5 79 0. 0 4 0. 0 2 6 NS C ( S ) 31 0. 2 9 98 0. 3 1 0. 9 2 9 10 0. 5 0 14 1 0. 2 7 0. 9 8 3 17 0. 6 5 79 0. 1 9 0. 1 3 1 NS C ( A ) 31 0. 0 6 98 0. 1 6 0. 3 5 2 10 0. 5 0 14 1 0. 2 0 0. 1 8 1 17 0. 3 5 79 0. 1 8 0. 2 0 3 NS C ( L F ) 31 0. 0 3 98 0. 2 0 0. 2 6 4 10 0. 2 0 14 1 0. 1 6 0. 8 0 5 17 0. 2 9 79 0. 0 6 0. 1 8 2 NS C ( S F ) 31 0 98 0. 0 3 0. 4 3 2 10 0 14 1 0. 0 2 0. 7 2 0 17 0. 1 2 79 0. 0 5 0. 4 8 4 NS C ( P ) 31 0. 2 6 98 0. 0 7 0. 3 9 1 10 0. 3 0 14 1 0. 0 9 0. 3 0 4 17 0. 2 4 79 0. 0 8 0. 0 2 8 Data were attained by use of Wil c o x o n ’ s rank-sum test.

The associations of QST stimuli by use of pred e t e r mined algorithms in the RDNS cohort, and, in this cohort, hyperesthesia with NIS and the CASE IV. QST results were several neuropathic end points were signifi- These associations were tested only for the ex p r essed as a percentile value or as a nor- cantly greater than those in patients whose RDNS cohort because hyperesthesia was too mal deviate, which allowed individual th r esholds fell in the low-normal range. in f r equent in the rhNGF and PRT cohorts to th r esholds to be characterized as abnor- The strong and consistent corre l a t i o n allow meaningful testing. Because hyperes - mally low (hyperesthetic, low-normal, or between sensory loss (thresholds at or thesia was found only for HP and not for h i g h - n o rmal) or as abnormally high above the 97.5th percentile) and other VDT or CDT, only associations of impair- (hypoesthetic). Studies of this kind had not markers of DPN and even of symptoms of ments and symptoms with HP were com- pr eviously been possible, because large po l y n e u r opathy provides additional vali- pa r ed (Table 3). Generally, in the RDNS co h o r ts had not previously been character- dation that QST hypoesthesia is an indica- co h o r t, impairments or symptoms were not ized by these approa c h e s . tor of DPN (15). consistently larger or smaller in patients with Our studies have confirmed that these The grea t e r -than-expected freq u e n c y HP:5 or HP:0.5 in patients whose thres h o l d s QST approaches are useful not only for the of patients with sensation thre s h o l d s we r e at or below the 2.5th percentile, as characterization of sensation alterations in between the 50th and 97.5th percentiles, as co m p a r ed with those patients whose thres h - diabetes, but also for the detection and co m p a r ed with that of patients with sensa- olds were between the 2.5th and 50th per- characterization of DPN. The findings of tion thresholds between the 2.5th and 50th centiles. All four neuropathic impairme n t s our study are as follows: 1) hypoesthesia is p e rcentiles, provides evidence of an and all seven composite symptom scores recognized by an increased frequency of unequivocal subclinical functional abnor- we r e greater in patients whose HP:5–0.5 patients with an unequivocal abnorma l i t y ma l i t y . Such evidence was observed in all we r e at or below the 2.5th percentile than in at or above the 97.5th percentile and by an th r ee cohorts. This shift in severity spectra patients whose thresholds were between the i n c reased frequency of patients with of QST abnorma l i t y , which was not prev i - 2.5th and 50th percentiles. The diffe re n c e s th r eshold values that fall between the 50th ously described, implies a subtle change in we r e significant for lower-limb NIS(W) (P = and 97.5th percentiles; 2) among modali- n e rve function, presumably before the 0.033), for NSC(W) (P = 0.026), and for ties of sensation, vibration threshold was development of minimal criteria for the NSC(P) (P = 0.028). most frequently abnormal when polyneu- diagnosis of DPN. By assessing NCs and ropathy was mild (as in the RDNS cohort) , variations in HBDB, we found a similar CO N C L U S I O N S — The unique stud- but all modalities tended to be abnorma l subclinical shift to abnorma l i t y . At the last ies characterizing vibration, cooling, and when polyneuropathy was more severe (as evaluation of the RDNS cohort, we found HP thresholds of the foot and their associ- in the rhNGF and PRT cohorts); 3) patients that peroneal motor nerve conduction ation with severity of DPN, as rep o rt e d with hypoesthesia of VDT or CDT have velocity was more frequent in the patients he r e, were made possible because of the significantly more impairments and symp- between the 2.5th and 50th perce n t i l e s following reasons: 1) data from three dia- toms of DPN than do patients with high- ( 60%) than in the patients between the betic cohorts that consisted of 1 , 5 0 0 no r mal thresholds; 4) increased freq u e n c i e s 50th and 97.5th percentiles ( 10%). It patients; 2) all patients of each cohort had of vibration or cooling hyperesthesia were should be noted, unlike hypoesthesia, severity of DPN quantitated using a com- not encountered in any of the cohorts stud- ab n o r mality of conduction velocity and pr ehensive score of severity and diffe re n t ied, but testing of cooling was not in the HBDB is in the lower tail of the distribu- symptom scores; and 3) vibration, cooling, cold-painful range; and 5) increased fre- tion. Likewise, 63% of HBDB res p o n s e s and HP thresholds of the foot were assessed quencies of patients with HP hyperes t h e s i a fell into the 2.5th–50th percentile category using graded, quantitated, and invariant (hyperalgesia) were observed most notably as compared with 27% in the 50th–97.5th

DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 515 Hypoesthesia and hyperalgesia in diabetic polyneuropathy pe r centile category. Early investigators did demonstrated only in the RDNS cohort, the 6. Ver dugo R, Ochoa JL: Quantitative somato- not categorize NC as a percentile abnor- mildest spectrum of DPN. The mechanisms se n s o r y thermotest: a key method for func- ma l i t y , but they did demonstrate that the underlying HP hyperalgesia are not com- tional evaluation of small caliber affe re n t conduction velocities of diabetic patients pletely understood, but electrop h y s i o l o g i c a l channels. Brain 115:893–913, 1992 without findings of DPN were significantly and chemical (pain peptides and cytokines) 7. Ochoa J: Thermal hyperalgesia as a clinical lower than those of control subjects (22). events in the periphery or in the dorsal symptom. In Hyperalgesia and Allodynia. William WD Jr., Ed. New York, Raven Both their demonstration and our demon- ho r n of the spinal cord have been impli- Pr ess, 1992, p. 151–165 stration of subtle QST, NC dysfunction, cated (8,25,26). Irrespective of the site or 8. Gracely RH: Experimental pain models. In and autonomic dysfunction, before the mechanisms involved, it is now clear that Advances in Pain Research and Therapy. Vol . development of overt DPN, undoubtedly ne u r opathic abnormality from QST must be 18. Max M, Portenoy R, Laskn E, Eds. New have biological, disease, and tre a t m e n t in f e r red not only from hypoesthesia but York, Raven Press, 1991, p. 33–47 implications. Earlier detection of subclini- also from hyperalgesia. 9. Low PA, Schmelzer JD, War d KK: The effe c t cal abnormality and more intensive trea t - The demonstration that hyperalgesia of age on energy metabolism and res i s t a n c e ment may later result in prevention of an (o r , in our study, a steeper stimulus res p o n s e to ischemic conduction failure in rat adverse clinical outcome (23,24). slope) is a useful marker of injured human peripheral nerve. J Physiol (Lond)3 7 4 : 2 6 3 – 272, 1986 Our observation that abnormality of ne r ve has been emphasized by Lindblom 10. Se r vice FJ, Rizza RA, Daube JR, O’Brien PC, VDT is much more frequent than abnor- and Ochoa (5), Ver dugo and Ochao (6), Dyck PJ: Near normoglycaemia improv e d mality of CDT or HP:5 in the RDNS cohort, and Ochoa (7). The present studies suggest ne r ve conduction and vibration sensation even though the frequency of these abnor- that it may be worthwhile to assess for in diabetic neuro p a t h y. Diabetologia 2 8 : malities is nearly identical in the rhN G F hyperalgesia also in patients with DPN, 722–727, 1985 and PRT cohorts, req u i r es an explanation. even when they do not rep o r t allodynia 11. Dyck PJ: Quantitative sensory testing: a con- It may be that this phenomenon simply (pain from stimuli that are not usually sensus rep o r t from the Peripheral Neurop a - reflects a milder involvement of a length- painful) or hypersensitivity. thy Association. Ne u r ology 43 : 1 0 5 0 – 1 0 5 2 , dependent polyneuropathy in the RDNS 19 9 3 co h o r t versus that in the other two cohorts . 12. American Diabetes Association and Ameri- can Academy of Neurology: Report and rec - Thus, in the RDNS cohort, sensation A c k n o w l e d g m e n t s — This study was sup- ommendations of the San Antonio ab n o r mality would be found in the toe (the po r ted in part by a grant from the National co n f e r ence on diabetic neuropathy (Consen- site of vibratory testing) but not on the Institute for Neurological Disorders and Strok e sus Statement). Diabetes Care 11 : 5 9 2 – 5 9 7 , dorsum of the foot (the site of cooling and (36797) and funding from Genentech, Inc. 19 8 8 HP testing), whereas in the rhNGF and (South San Francisco, CA). 13. Dyck PJ: Quantitating severity of neurop a - PR T cohorts, sensation abnormality would We gratefully acknowledge Jenny Davies, th y . In Peripheral Neuro p a t h y. 3rd ed. Dyck occur at all sites. The alternative explana- Jane Norell, and Peter Compton, MA, for help- PJ, Thomas PK, Griffin JW, Low PA , ing with data analysis; Mary Lou Hunziker and tion is that the vibratory receptors and/or Poduslo JF, Eds. Philadelphia, WB Saun- Ca r ol Overland for helping with preparation of ders, 1993, p. 686–697 fibers are affected in mild neuro p a t h y, the manuscript; and L. Joseph Melton, III, MD, wh e r eas all sensory fibers are affected in 14. Williamson RT: The vibrating sensation in for offering helpful suggestions. The laboratory af fections of the nervous system and in dia- mo r e severe neurop a t h y . Regardless of the of Phillip A. Low provided the HBDB res u l t s . betes. Lancet i:855–856, 1905 explanation, the diffe r ence in the pattern of 15. Dyck PJ, Bushek W, Spring EM, Karnes JL, QST abnormalities by modality provides a Litchy WJ, O’Brien PC, Service FJ: Vib r a - useful indicator of severity. Re f e re n c e s to r y and cooling detection thresholds com- This study provides new informa t i o n 1. O’Brien PC, Dyck PJ: Proc e d u r es for setting pa r ed with other tests in diagnosing and on hyperesthesia in DPN. First, hyperes - no r mal values. N e u rology 45:17–23, 1995 staging diabetic neurop a t h y . Diabetes Care thesia was found only for HP and not for 2. Dyck PJ, Litchy WJ, Lehman KA, Hokan- 10:432–440, 1987 vibration or cooling. Because we do not son JL, Low PA, O’Brien PC: Variables influ- 16. Sosenko JM, Boulton AJM, Kubrusly DB, routinely test cooling to the cold-pain range encing neuropathic endpoints: the Weintraub JK, Skyler JS: The vibratory per- by use of the CASE IV, we do not know Rochester Diabetic Neuropathy Study of ception threshold in young diabetic patients: associations with glycemia and whether cold-pain hyperalgesia would have Healthy Subjects (RDNS-HS). N e u ro l o g y 45:1115–1121, 1995 pu b e rt y . Diabetes Care 8:605–607, 1985 been found if we had tested for it. Secondly, 3. Steiness IB: Vibration perception in norma l 17. Zaslansky R, Yar nitsky D: Clinical applica- stimulus-induced HP hyperesthesia (hyper- subjects and in diabetics: a biothesiometric tions of quantitative sensory testing (QST). algesia) was more frequent than expected, st u d y . Acta Med Scand158:315–331, 1957 J Neurol Sci153 (Suppl. 2):215–238, 1998 but only in the RDNS cohort. Conceptually, 4. La m b e r t EH, Dyck PJ: Compound action 18. Dyck PJ, Zimmerman IR, O’Brien PC, Ness this hyperesthesia might be due to either potentials of sural nerve in vitro in periph- A, Caskey PE, Karnes J, Bushek W: Intro- im p r oved but normal sensation (supersen- eral neurop a t h y . In Peripheral Neuro p a t h y. duction of automated systems to evaluate sitivity) or to abnormal sensation (hyperal- 3r d ed. Dyck PJ, Thomas PK, Griffin JW, t o u c h - p re s s u re, vibration, and therm a l gesia). Because patients with this HP Low PA, Poduslo JF, Eds. Philadelphia, WB cutaneous sensation in man. Ann Neurol4 : hy p e r esthesia had significantly greater neu- Saunders, 1993, p. 672–684 502–510, 1978 19. Dyck PJ, Zimmerman I, Gillen DA, John- ropathic impairment and symptoms, we 5. Lindblom U, Ochoa J: Somatosensory function and dysfunction. In Diseases of the son D, Karnes JL, O’Brien PC: Cool, warm, infer that hyperesthesia is a manifestation of Ne r vous System: Clinical Neurob i o l o g .y Vol. 1. and heat-pain detection of receptors: test- hyperalgesia. The best indicator of hyperal- As b u r y AK, McKhann GM, McDonald WI, ing methods and inferences about anatomic gesia was the HP:5–0.5 slope. Furth e r , it is Eds. Philadelphia, WB Saunders, 1986, distribution of receptors. N e u rology 4 3 : of interest that hyperalgesia could be p. 28 3 – 2 9 8 1500–1508, 1993

516 DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 Dyck and Associates

20. Dyck PJ, Davies JL, Litchy WJ, O’Brien PC: Diabetic Neuropathy Study. N e u rology 43 : with diabetes mellitus: a clinical and elec- Longitudinal assessment of diabetic 817–824, 1993 tr omyographic study of 103 unselected dia- po l y n e u r opathy using a composite score 22. Skillman TG, Johnson EW, Hamwi GJ, betic patients. Ne u r ology11:275–284, 1961 in the Rochester Diabetic Neuro p a t h y Driskill HJ: Motor nerve conduction veloc- 25. Carstens E: Quantitative experimental Study cohort. Ne u r ology49:229–239, 1997 ity in diabetes mellitus. Diabetes 10 : 4 6 – 5 1 , assessment of pain and hyperalgesia in ani- 21. Dyck PJ, Kratz KM, Karnes JL, Litchy WJ, 19 6 1 mals and underlying neural mechanisms. Klein R, Pach JM, Wilson DM, O’Brien PC, 23. Downie AW, Newell DJ: Sensory nerve con- P rog Brain Res110:17–31, 1996 Melton LJ, Service FJ: The prevalence by duction in patients with diabetes mellitus 26. Mizumura K: Natural history of nociceptor staged severity of various types of diabetic and controls. N e u rology11:876–882, 1961 sensitization: the search for a peripheral ne u ro p a t h y , ret i n o p a t h y , and nephrop a t h y 24. Mulder DW, Lambert EH, Bastron JA, mechanism of hyperalgesia (Abstract). Pain 5 in a population-based cohort: the Rochester Sprague RG: The neuropathies associated (Suppl. 2):59–82, 1998

DIABETES CARE, VOLUME 23, NUMBER 4, APRIL 2000 517