Int J Biometeorol (2012) 56:811–821 DOI 10.1007/s00484-011-0483-9
ORIGINAL PAPER
Reliability of the method of levels for determining cutaneous temperature sensitivity Miroljub Jakovljević & Igor B. Mekjavić
Received: 28 December 2010 / Revised: 20 June 2011 / Accepted: 24 July 2011 / Published online: 21 August 2011 # ISB 2011
Abstract Determination of the thermal thresholds is used clinically for evaluation of peripheral nervous system function. The aim of this study was to evaluate reliability of the method of levels performed with a new, low cost device for determining cutaneous temperature sensitivity. Nineteen male subjects were included in the study. Thermal thresholds were tested on the right side at the volar surface of mid-forearm, lateral surface of mid-upper arm and front area of mid-thigh. Thermal testing was carried out by the method of levels with an initial temperature step of 2°C. Variability of thermal thresholds was expressed by means of the ratio between the second and the first testing, coefficient of variation (CV), coefficient of repeatability (CR), intraclass correlation coefficient (ICC), mean difference between sessions (S1-S2diff), standard error of measurement (SEM) and minimally detectable change (MDC). There were no statistically significant changes between sessions for warm or cold thresholds, or between warm and cold thresholds. Within-subject CVs were acceptable. The CR estimates for warm thresholds ranged from 0.74°C to 1.06°C and from 0.67°C to 1.07°C for cold thresholds. The ICC values for intra-rater reliability ranged from 0.41 to 0.72 for warm thresholds and from 0.67 to 0.84 for cold thresholds. S1-S2diff ranged from −0.15°C to 0.07°C for warm thresholds, and from −0.08°C to 0.07°C for cold thresholds. SEM ranged from 0.26°C to 0.38°C for warm thresholds,
M. Jakovljević (*) Faculty of Health Studies, University of Ljubljana, Ljubljana, Slovenia e-mail:
[email protected] I. B. Mekjavić Department of Automation, Biocybernetics and Robotics, Josef Stefan Institute, Ljubljana, Slovenia
and from 0.23°C to 0.38°C for cold thresholds. Estimated MDC values were between 0.60°C and 0.88°C for warm thresholds, and 0.53°C and 0.88°C for cold thresholds. The method of levels for determining cutaneous temperature sensitivity has acceptable reliability. Keywords Thermotactile quantitative sensory testing . Healthy subjects . Reliability
Introduction Thermotactile quantitative sensory testing (TTQST) of thresholds is increasingly used for evaluation of peripheral nervous system function in the clinical and research domains. Evoked potentials to sensory thermal stimulation of skin may provide objective information about the integrity of thermal sensory afferents as part of the peripheral nerve system, as well as brain response to selective stimulation of certain types of sensory fibers. Thermal threshold testing is a suitable measure of the function of unmyelinated C fibers and myelinated A-δ fibers. There are two predominant classes of methods: reaction time inclusive and reaction time exclusive (RTE) (Yarnitsky 1997). The method of levels is an RTE method that utilizes stimuli of predetermined levels of intensity and duration. Upon termination of the stimulus, the subject is requested to indicate whether the stimulus was perceived. In the event that the stimulus was perceived, the following stimulus is smaller in intensity, whereas following an unperceived stimulus, the intensity of the following stimulus is increased. The magnitude of the change in intensity from one stimulus to the subsequent stimulus is termed a “step”. The precise manner in which the stimulus intensity is altered, the test terminated, and the threshold
812
calculated differs for the various algorithms. Overestimation or increased variability of sensory thresholds is minimized. Despite being the least time consuming of the TTQST methods, it carries a potential for misreading if the subject is not fully alert to each and every stimulus (Yarnitsky 1997). Commercially available devices for testing thermal sensation usually incorporate a thermoelectric module (thermode) that effects temperature changes by the Peltier effect, i.e., when a current is passed through a junction of dissimilar metals, cooling or warming will occur at the junction depending on the direction of the current. The thermode is in contact with the skin, and the subject is requested to report the sensation of temperature change. Skin adaptation temperature is usually maintained at 32°C or close to the level of the surrounding skin area. The most commonly used rate of change of temperature during a test of cutaneous temperature sensitivity is 1°C s−1. Subjects are not necessarily naïve regarding the direction (heating or cooling) of the thermal stimulus. If they are, subsequent to indicating presence or absence of sensation, they are also asked to report its quality (warm or cold). However, the devices also differ regarding hardware, which dictates different capabilities regarding rate of temperature change, range of temperatures, size of the thermode, algorithm used for stimulation, and also regarding price. The latter has doubtlessly an influence on decision about acquisition of such a device. In contrast to other TTQST devices, Biomed T-sensy is low cost, small, light, portable, designed for field TTQST and without additional joining pieces (cooling system, calibration system or patient response device). The key to performing good experiments is to make sure that results are reproducible. Therefore, the aim of the present study was to evaluate the reproducibility of the TTQST test performed with a new, low cost device (described in detail in the Methods section below), applying the standardized approach according to Yarnitsky and Ochoa (1991). Our working hypothesis was that reproducibility of the new device would be acceptable, i.e., comparable to the estimates found in the literature for the more costly devices.
Materials and methods The study was approved by the National Committee for Medical Ethics of the Republic of Slovenia. The thermal test was carried out in healthy male subjects. All were free of neurological illness, did not have any medical condition that may cause peripheral neuropathy, were not taking any prescription medications and had normal neurological examination. They were all familiarized with protocol and gave informed consent to participate in the study
Int J Biometeorol (2012) 56:811–821
Testing conditions Sessions were held in stable environmental conditions. Subjects participated in two sessions separated by at least 48 hours. During the tests, subjects lay in supine position. They received standardized instructions for the test. All tests were performed by the same investigator (the first author). Distractions were minimized, and participants were allowed ten minutes to adapt to the room temperature. Visual access to the equipment was restricted. Total testing time of each session did not exceed 35 minutes in order to prevent decline in attention. Ambient temperature and relative humidity were measured during each test. Skin and tympanic temperature Skin temperature (Tsk) was measured at four sites with thermocouples (Concept Engineering, Old Saybrook, CT, USA). They were attached with adhesive tape (Transpore, 3M, Germany). Weighted mean Tsk was calculated from measurements obtained at the chest (manubrium, Tch), forearm (mid anterior surface Tfa), thigh (mid anterior surface, Tth) and calf (mid medial surface, Tca) sites. Immediately upon completion of the test, tympanic temperature (Tty) of the subject was measured with an infrared thermometer (GentleTemp 510, Omron Healthcare Co., Ltd., Kyoto, Japan). Tested sites and side Cold and warm coetaneous sensitivity thresholds were tested at the volar surface of the mid-forearm, lateral surface of the mid-upper arm and front area of the mid-thigh. All tests were performed on the right side. The order of the cold and warm tests was alternated among the subjects. However, the order of the test sites was always the same, i.e., upper arm, followed by forearm and thigh. Equipment and algorithm Thermal stimuli were produced by the Biomed T-Sensy device (MAK Elektronik, MAK d.o.o., Škofja Loka, Slovenia). The device consisted of a thermode (Peltier element) with a surface area of 12.5 cm2 and a mass of 165 g, thermal sensor for the measurement of surrounding skin (reference temperature, Tref) and a controlling system (Fig. 1a, b, c). The accompanying software enables either automatic or manual testing. The thermode was secured to the test skin site with Velcro type elastic band (Fig. 1d). Since in adults, pretest warming of the test skin site either did not improve threshold detection or (in some subjects) complicated perception of test stimuli, the pretest and reference temperatures were measured on the skin
Int J Biometeorol (2012) 56:811–821
813
Fig. 1 System configuration: host computer (a), Biomed T-Sensy main unit (b) with thermode (c), and illustration of thermode application (d)
surrounding the test site. Skin temperature adaptation was always to the temperature of the adjacent surrounding area. Thermal testing was carried out manually according to the method of levels (Yarnitsky and Ochoa 1991). An initial temperature step of 2°C was administered, with the temperature returning to the adaptation temperature immediately upon stimulus termination. Subjects were requested to respond whether they perceived the stimulus (“yes”) or not (“no”). An affirmative (“yes”) response always caused the subsequent stimulus to be smaller by a half, while a negative (“no”) response led to the subsequent stimulus being 2°C larger. Stimuli continued to be increased by steps of 2°C until the first “yes” response. Stimuli were then decreased by 1°C until the first “no” was recorded. This was continued until stimulus step size attained 0.2°C. The average magnitude of the step stimulus temperature for the last “yes” and “no” responses was defined as the cutaneous temperature sensitivity threshold. Rate of change of temperature during a test of cutaneous temperature sensitivity was 1°C s−1. Statistical analysis Ambient (Ta), mean weighted skin (Tsk), and tympanic (Tty) temperatures obtained in the two trials were compared with Student‘s t-test for paired samples. Correlation between thermal thresholds and skin temperature was
assessed with Pearson’s correlation coefficient. Variability was expressed as the ratio between second and first (S2/S1) measurements of thermal threshold. Coefficients of variation were calculated within the sample (CVsa) and within the subjects (CVsu). Descriptive statistics are reported as mean ± standard deviation. Statistical significance was set at p≤0.05. Most of the reliability studies for the TTQST (Chong and Cros 2004) report the coefficient of repeatability (CR). To estimate the magnitude of variability, the individual differences (test 2–test 1) were calculated for each point. The CR was computed according to the formula: CR ¼ 2 SDdiff
ð1Þ
where SDdiff is the standard deviation of the individual differences. Approximately 95% of test–retest differences are expected to be in the range from –CR to+CR. In contrast, intra-class correlation coefficients (ICCs) have been reported to be the best method for reliability analysis in test–retest situations (Huck and Cormier 1996). Reliability was tested using a repeated measures analysis of variance (ANOVA) and ICC. An ICC (2,1) was calculated to determine the agreement in terms of consistency of observer across the two test occasions. It was decided a priori that a reliability coefficient of more than 0.75 is
814
Int J Biometeorol (2012) 56:811–821
considered to represent “excellent” reliability, while an ICC of more than 0.40 is an index of “good” reliability (Merchut and Toleikis 1990). Other important outcome measures in reliability studies are typical error and change in the mean values between trials (Hopkins 2000), which were also calculated. A typical error is synonymous with the within-subject standard deviation, or the standard error of measurement (Hopkins 2000). Based on typical error, i.e., standard error of measurement (SEM), the minimally detectable changes (MDC) for temperature thresholds were calculated. The SEM or the error associated with a single measure was computed with the following formula (Stratford et al. 1996): SEM ¼ SD
p
ð1 RÞ
ð2Þ
where SD is the standard deviation of the thermal thresholds and R is the test–retest reliability coefficient (ICC 2,1). In statistical terms, the MDC, also called smallest detectable change or smallest real change, indicates which changes fall outside the measurement error. The MDC was defined as follows: MDC ¼ 1:64
p
2 SEM
ð3Þ
The value represents the 90% confidence interval and defines the possible range of the measurements because of error. It is an estimate of MDC because a change larger than it represents a change that is unlikely to be a result of error. Bland-Altman analyses were performed to assess the agreement between thermal threshold measurements for individual participants (Bland and Altman 1986). Using the limits-of-agreement plot rather than the conventional test– retest scatterplot, a rough indication of systematic bias and random error is provided by examining the direction and magnitude of the scatter around the zero line, respectively.
Results During the study, ambient temperature of 25.9±1.3°C (range, 25–28°C) and relative humidity of 22.3±1.8% (range, 20–25%) were similar in both trials. Nineteen healthy male subjects participated in the test–retest study. Their mean age was 22±3 years (range, 18–31 years), mean body weight was 79.2±14.9 kg (range, 57.9–102.2 kg), mean height was 180±9 cm (range, 161–197 cm) and mean body mass index was 24.3± 3.0 kg/m2 (range, 20.2– 33.6 kg/m2). Mean Tsk in the first session was 33.6±0.6°C and in the second session 33.7±0.5°C. The difference between mean Tsk in the two trials was not statistically significant. Skin temperature (both local and average) was not correlated with any of the thresholds (correlation coefficients ranged between −0.30 and 0.19 with p-values between 0.25 and 0.94). Mean Tty during the first session was 36.5±0.3°C, and 36.5±0.2°C during the second session. The difference was not statistically significant. There were no statistically significant changes between sessions for either the warm or the cold thresholds (Table 1). There was also no significant difference between warm and cold thresholds in all tested sites. CVsa for warm thresholds was smaller than that for cold thresholds, and was reduced in the second trial only for cold thresholds. CVsu ranged from 21.4% for cold thresholds to 28.8% for warm thresholds on the thigh (Table 1). These are acceptable values for coefficients of variation. Mean CVsu for warm thresholds were the highest for thigh and the lowest for upper arm. In contrast, CVsu for cold thresholds were the highest for upper arm and the lowest for thigh (Table 1). The second warm threshold can be expected to lie on average between 36.4% (forearm) and 471.4% (thigh) of the first measurement, while a cold threshold may vary between 44.4% (thigh) and 420.0% (upper arm) (Table 2). The mean of differences ranged from 0.01°C for the thigh cold stimuli to 0.15°C for the thigh warm stimuli. The CR estimates for warm thresholds ranged from 0.74°C (forearm) to 1.06°C
Table 1 Mean, standard deviations (SD), coefficients of variation within sample (CVsa), and coefficients of variation within subject (CVsu) for warm and cold temperature thresholds estimated in the first (S1) and second (S2) session Location
S1 (°C) Mean (SD)
S1 CVsa (%)
Upper arm Forearm Thigh
1.0 (0.8) 0.8 (0.4) 0.8 (0.3)
78.2 52.3 41.5
Upper arm Forearm Thigh
0.7 (0.4) 0.9 (1.1) 1.0 (0.5)
61.0 121.1 53.0
S2 (°C) Mean (SD) Warm thresholds 1.0 (0.7) 0.7 (0.4) 1.0 (0.5) Cold thresholds 0.8 (0.4) 0.9 (0.7) 1.0 (0.5)
S2 CVsa (%)
Mean CVsu (%) (range)
65.9 62.6 51.4
22.5 (0.0–50.7) 24.2 (0.0–66.0) 28.8 (0.0–91.9)
51.0 81.0 50.3
25.1 (0.0–87.0) 22.7 (4.0–54.4) 21.4 (2.0–54.4)
Int J Biometeorol (2012) 56:811–821 Table 2 Retest as percentage (%) of test, mean differences between sessions (S1-S2 diff), coefficient of repeatability (CR), and intraclass correlation coefficients (ICC) with 95% confidence limits (95% CL) for warm and cold temperature thresholds estimated in the first (S1) and second (S2) session
815
Location
Retest as % of test (range)
Upper arm Forearm Thigh
51.4–211.8 36.4–257.1 43.8–471.4
Upper arm Forearm Thigh
48.6–420.0 63.1–225.0 44.4–169.2
(upper arm), and for the cold thresholds from 0.67°C (thigh) to 1.07°C (forearm) (Table 2). The ICC values for intrarater reliability were all above 0.40; they ranged from 0.41 (thigh) to 0.72 (upper arm) for warm thresholds, and from 0.67 (upper arm) to 0.84 (forearm) for cold thresholds (Table 2). The latter were therefore more reliable. In comparison with cold thermal thresholds, SEM was higher for warm thermal thresholds for upper arm and thigh and lower for forearm (Table 3). SEM for warm stimuli was the largest for the upper arm and smallest for the forearm, and SEM for cold stimuli was the largest for the forearm and smallest for the thigh. The minimal detectable change shows that in hypothetical applications of the measurement in healthy male subjects, only changes larger than 0.60°C to 0.88°C for warm thresholds and 0.53°C to 0.88°C for cold thresholds (Table 3) can be interpreted as real changes in TTQST. Bland-Altman analyses are presented in Fig. 2. The bias (mean difference between two measurements) for cold thresholds was from −0.08°C (upper arm) to 0.07°C (forearm); for warm thresholds, it ranged from −0.15°C (thigh) to 0.07°C (forearm).
Discussion Both temperature sensitivity thresholds (Zwart and Sand 2002) and test–retest differences seem to increase with age. We concur with the results of previous studies (Meier et al.
Table 3 Standard error of measurement (SEM) with 95% confidence limits (95% CL), and minimal detectable change (MDC) for warm and cold temperature thresholds
Location
Upper arm Forearm Thigh
S1-S2 diff (°C) Warm thresholds −0.02 0.07 −0.15 Cold thresholds −0.08 0.07 −0.01
CR (°C)
ICC (95% CL)
1.06 0.74 0.91
0.72 (0.46–0.87) 0.61 (0.29–0.81) 0.41 (0.02–0.69)
0.70 1.07 0.67
0.67 (0.37–0.84) 0.84 (0.66–0.92) 0.80 (0.60–0.91)
2001) that significant differences between tests and re-tests are generally not found. Moreover, despite the fact that warm receptors lie in deeper skin layers than cold thermoreceptors (Hensel et al., 1974), and that they are also outnumbered by the cold sensors (Guyton and Hall 2000), our data did not show a difference in thermal thresholds between cold and warm stimuli as others (Valensi et al. 1993) have reported. Strong dependence of warm sensation on area of stimulation exists (Kojo and Pertovaara 1987), i.e., the smaller the area of stimulation, the higher stimulus intensity needed for threshold sensation. In the present study, we used a 12.5 cm2 thermode with area obviously large enough to cover a sufficient number of the skin’s hot spots. Thermode baseline temperatures are usually kept at 32°C (Meier et al. 2001), or in the range of 31–33°C for the upper limbs, and in the range of 30–34°C (Valensi et al. 1993) or higher (Claus et al. 1990) for the lower limbs. We did not preheat the stimulated area because the environmental temperature ensured that the skin temperature was in the range of 30–34°C. In any case, no differences in temperature sensitivity were noted for measurements conducted in the warm and neutral environment (Strigo et al. 2000; Hagander et al., 2000). In the present study, the sample was relatively homogenous, thus the difference observed in TTQST reliability for the different sites is most likely due to variations in the distribution of thermoreceptors between subjects and test sites.
Warm threshold (°C)
Cold threshold (°C)
SEM (95% CL) (°C)
MDC (°C)
SEM (95% CL) (°C)
MDC (°C)
0.38 (0.30–0.52) 0.26 (0.21–0.36) 0.32 (0.25–0.46)
0.88 0.60 0.74
0.25 (0.20–0.34) 0.38 (0.30–0.52) 0.23 (0.19–0.33)
0.58 0.88 0.53
816
Int J Biometeorol (2012) 56:811–821
WARM THRESHOLDS
COLD THRESHOLDS
Fig. 2 Bland-Altman analysis (limits-of-agreement plots) of warm and cold temperature threshold measurements at upper arm (UA), forearm (FA) and thigh (T)
Although several thermal sensation reliability studies have been published (Chong and Cros 2004), it is difficult to compare the results between them due to different subject populations, detection methods, sites and particularly data analysis methods. Correlation has been used in the past, but is inappropriate, since it is a measure of the strength of the relation between two variables and not a measure of agreement. Despite that fact, some authors (Claus et al. 1990) used correlation techniques to compare results between separate sessions. Others (Yarnitsky and Sprecher 1994) used calculation of the CR, for which there is a 95% confidence interval that two determinations made on the same subject would differ by less than CR. Only intersession differences larger than CR would be considered “different”. Huck and Cormier (1996) reported intraclass correlation coefficients to be the best method for reliability analysis in test–retest situations. Nevertheless, Hopkins (2000) stated that in the reliability study the most important outcome measures are typical error and the change in the mean between trials. It appeared as the best solution to us to employ different reliability statistics, with intention to compare the results of the present study with as many previous (and future) ones as possible. ICC is an accepted method of quantifying reproducibility. However, using ICC, the variance between subjects is usually considered as variance of interest, whereas with respect to longitudinal changes the magnitude of within-subjects
variance over time is relevant (Roebroeck et al. 1993), so the MDC are better suited for this purpose (Hopkins 2000). The MDC of instrument (and algorithm) can be used to determine whether the change in thresholds indicates true and meaningful changes in follow-up or outcome studies (Beaton et al. 2002). Thus the MDC is a better benchmark than traditional usage of statistical significance to determine whether the change is important (Hsieh et al. 2007). We found no data in the literature to compare our MDC with others. Both cold and warm retest thresholds were expressed as a percentage of the test values, and were expected to vary between 30% and either 230% (Becser et al. 1998) or 300% (Claus et al. 1990) of the first measurement. Our results confirmed the expectations regarding the lower limit, and the majority of the values also regarding the upper limit, with the exception of the thigh warm thresholds (451%) and upper arm cold thresholds (420%) (Table 4). In normal subjects, the within subject CVsu of the thermal threshold can be as much as up to 71% (Armstrong et al. 1991), and within subject CVsu in diabetic patients is even higher (Arezzo et al. 1986). In the present study, the within subject CVsu did not exceed 30% for both thresholds. The results are in agreement with those reported by Valensi et al. (1993), who claimed that in diabetic patients CVsu were similar for both thresholds. Previous studies (Claus et al. 1990; Valensi et al. 1993) have also reported
Site
Behind the right medial malleolus
Five cephalic points, thenar, dorsal hand
Mid-forearm, lateral surface of the mid-upper arm and front area of the mid-thigh
Behind the right medial malleolus
Sole of the foot
Reference
Claus et al. 1990
Becser et al. 1998
Jakovljević and Mekjavić, present data
Claus et al. 1990
Armstrong et al. 1991
Table 4 Reproducibility studies on quantitative thermo tactile sensory testing
Healthy: 7 monthsDiabetics: 1 month
Three consecutive days
Two days
Twice at different days, within a 7-day limit
Three consecutive days
Interval
Thermal Stimulator (method of levels)
PATH-Tester MPI 100 (method of limits and force choice test)
Biomed T-sensy (method of levels)
Somedic (method of limits)
PATH-Tester MPI 100 (method of limits and force choice test)
Equipment (algorithm)
Upper limit 313.4 (211.8–471.4) Coefficient of variation Method of limits: Cold: 62.1 Warm: 35.8 Forced choice: Cold: 89.0 Warm: 51.5 Healthy Cold: 71.0 (male); 61.1 (female) Warm: 66.8 (male); 60.3 (female) Diabetics Cold: 8.8–129.5 Warm: 3.2–108.1
Percentage of the test value: average (range) (%) Method of limits: Cold: up to 300 Warm: up to 250 Forced choice: Cold: up to 350 Warm: up to 250 Cold: Lower limit 44 (39–56) Upper limit 206 (174–244) Warm: Lower limit 50 (35–68) Upper limit 206 (137–220) Cold: Lower limit 52.0 (44.4–63.1) upper limit 271.4 (169.2–420.8) Warm: Lower limit 43.9 (36.4–51.4)
Measure of reproducibility
Int J Biometeorol (2012) 56:811–821 817
Site Index finger, great toe
Dominant side medial malleolus and great toe
Foot dorsum Mid-forearm, lateral surface of the mid-upper arm and front area of the mid-thigh
Thenar eminence, dorsum of the foot
Five cephalic points, thenar, dorsal hand Suprapatellar, medial calf, lateral calf, foot dorsum, lateral foot
Reference
Arrezo et al., 1986
Valensi et al. 1993
Bird et al. 2006
Jakovljević and Mekjavić, present data
Yarnitsky and Sprecher 1994
Becser et al. 1998
Zwart and Sand 2002
Table 4 (continued)
One to two hours
Twice at different days, within a 7-day limit
Two weeks
Two days
Three separate days within 4-week window
Four weeks
Measure of reproducibility
Somedic (method of limits)
Somedic (method of limits)
Warm: 2.94 (1.58–4.30) Cold: 1.04 (0.52–1.54) Warm: 1.56 (0.62–2.66) Symptomatic side Cold: 2.1 (1.3–3.4) Warm: 4.8 (3.5–7.0) Non-symptomatic side Cold: 2.1 (1.2–2.7) Warm: 4.5 (2.8–6.2)
Index finger: 19.0 Great toe: 26.6 Range (8.3–47.1) Thermal Testing System (forced choice) Sample: Cold: 116.6 Warm: 64.5 Intra-subject: Cold: 15.7 Warm: 32.8 CASE IV (4-2-1 stepping algorithm) Diabetics Cold: 30.22 Biomed T-sensy (method of levels) Sample: Cold: 69.6 (50.3–121.1) Warm: 58.7 (41.5–78.2) Intra-subject: Cold: 23.1 (0.0–87.0) Warm: 25.2 (0.0–91.9) Coefficient of repeatability: average (range) °C Medoc TSA (method of levels, Method of levels: method of limits) Cold: 2.03 (1.04–3.02) Warm: 2.17 (0.57–3.76) Method of limits: Cold: 2.87(1.96–3.78)
Equipment (algorithm)
Ten times at intervals of at least a day Thermal Sensitivity Tester (forced choice)
Interval
818 Int J Biometeorol (2012) 56:811–821
Site Thenar, foot dorsum
Mid-forearm, lateral surface of the mid-upper arm and front area of the mid-thigh
Behind the right medial malleolus
Five cephalic points, thenar, dorsal hand Suprapatellar, medial calf, lateral calf, foot dorsum, lateral foot
Foot dorsum Mid-forearm, lateral surface of the mid-upper arm and front area of the mid-thigh
Reference
Moravcová et al. 2005
Jakovljević and Mekjavić, present data
Claus et al. 1990
Becser et al. 1998
Zwart and Sand 2002
Bird et al. 2006
Jakovljević and Mekjavić, present data
Table 4 (continued)
Two days
Three separate days within 4-week window
One to two hours
Twice on different days, within a 7-day limit
Three consecutive days
Two days
One week
Interval
Biomed T-sensy (method of levels)
CASE IV (4-2-1 steppig algorithm
Somedic (method of limits)
Somedic (method of limits)
PATH-Tester MPI 100 (method of limits and force choice test)
Biomed T-sensy (method of levels)
Medoc TSA (method of levels)
Equipment (algorithm)
Cold: 0.63 (0.47–0.75) Warm: 0.56 (0.38–0.80) Symptomatic side Cold: 0.69 (0.40–0.83) Warm: 0.55 (0.35–0.67) Non-symptomatic side Cold: 0.59 (0.27–0.86) Warm: 0.61 (0.43–0.82) Diabetics Cold: 0.70 (0.68–0.73) Cold: 0.77 (0.67–0.84) Warm: 0.58 (0.41–0.72)
Intraclass correlation coefficient: average (range) Method of limits: Cold: 0.78 (0.73–0.83) Warm: 0.70 (−) Forced choice: Cold: 0.89 (−) Warm: 0.80 (0.78–0.82)
Healthy Cold: 0.69 (0.48–0.90) Warm: 1.46 (0.54–2.38) Neuropathic Cold: 2.17 (1.22-3.12) Warm: 1.93 (1.24-2.62) Cold: 0.81 (0.70–1.07) Warm: 0.90 (0.74–1.06)
Measure of reproducibility
Int J Biometeorol (2012) 56:811–821 819
820
that the within-sample CVsa for cold thresholds are larger than for the warm thresholds. Our results confirmed this only for the forearm and thigh. Our results are in agreement with those of Becser et al. (1998) who noted that the CVsa for the warm thresholds of the hand dorsum and the upper arm was higher than CVsa for cold thresholds. We can conclude that within-sample and within-subject variability values of the warm and cold thresholds in our study were in the middle of the mentioned studies (Table 4). Yarnitsky and Sprecher (1994) calculated the CR in the foot to be 3.8°C for warm and 4.3°C for cold thresholds. Zwart and Sand (2002) found a smaller CR for cold thresholds than for warm thresholds. This was confirmed by Moravcová et al. (2005), who also reported lower CRs for the hand thenar (0.48°C and 0.54°C for cold and warm thresholds, respectively) compared to the hand dorsum (0.90°C and 2.38°C for cold and warm thresholds, respectively). Warm-insensitive areas in the forearm (e.g. areas of at least 4.84 cm2 with thresholds larger than 41°C) have been demonstrated in healthy volunteers, and it is suggested that this is probably caused by a very low warm receptor density in some normal skin areas (Gelber et al. 1995). Becser et al. (1998) reported that CR was higher for warm thresholds (1.92±0.89°C) than for cold thresholds (0.78±0.32°C) for the dorsal hand. Conversely, Dyck et al. (1991) emphasized that cold thresholds were more reliable than warm thresholds. In the present study, the CR for warm and cold thresholds was similar (Table 4). Claus et al. (1990) reported that ICCs for 55 normal subjects behind the right medial maleolus ranged from 0.73 to 0.83 for warm thresholds and ICC was 0.71 for cold thresholds. In the dorsal hand, good reproducibility was observed for both warm (ICC=0.66) and cold (ICC=0.71) thresholds (Becser et al. 1998). Zwart and Sand (2002) reported reliability in sciatica patients for warm thresholds as ICC=0.35–0.82 and for cold thresholds as ICC=0.27– 0.86. Bird et al. (2006) demonstrated that cold thermal thresholds on the left dorsal foot of diabetics (N=1100), as measured with CASE IV instrumentation, was highly reproducible (ICC=0.68–0.73). In our study, ICC showed good reliability (>0.40) for warm thresholds on all sites and cold thresholds on the upper arm, and excellent reliability (>0.75) for cold thresholds on forearm and thigh. In general, warm threshold ICC values were lower (reaching only good repeatability). Nevertheless, our ICC estimates suggest TTQST repeatability to be acceptable both for cold and warm thresholds (Table 4). Furthermore, Bland-Altman plots (Fig. 1) show that there was no systematic bias (i.e., trend or increasing or decreasing spread across subjects), there were only one or two subjects with notably different measurements (i.e., points outside the limits of agreement) at either site for warm and cold temperature thresholds, and all the estimated
Int J Biometeorol (2012) 56:811–821
limits of agreement include the desired zero difference. The reasons for the relatively large difference observed in a single subject (on upper arm regarding warm and cold threshold and on forearm regarding cold threshold) most probably lie in the limitations discussed below, i.e., malfunctioning of the device after repeated application or a drop of the subject’s attention. Despite establishing relatively good reliability, we realized some limitations. Causes of some poorer reliability indicators can be sought in a variety of factors, which can be grouped into those resulting from the environment, the proposed algorithm, the Biomed T-Sensy system and the subjects. Since the environmental factors were stable and had no influence on skin thermal receptors, and the proposed algorithm was followed strictly, the last two factors deserve more attention. On some occasions, we experienced a problem with slow thermode cooling or warming to the level of surrounding skin temperature. Additionally, periodical malfunctioning arose from the software and/or the hardware. The consequence was that the periods between stimuli were sometimes longer, which most likely caused a decrease in the subjects’ attention. As already mentioned in the introduction, if the subject is not fully alert to each and every stimulus, the chances for misreading are higher and the measurement reliability is lower.
Conclusion The function of peripheral small fibers of the upper and lower extremity and their central projections can be assessed by TTQST. Becser et al. (1998) emphasize that knowledge of test–retest reliability is important in research planning. We ascertain that TTQST with the method of levels using the low cost Biomed T-sensy device has acceptable reliability, at least in the setting of a scientific intervention study in healthy individuals.
References Arezzo JC, Schaumburg HH, Laudadio C (1986) Thermal sensitivity tester: device for quantitative assessment of thermal sense in diabetic neuropathy. Diabetes 35:590–592 Armstrong FM, Bradbury JE, Ellis SH, Owens DR, Rosen I, et al. (1991) A study of peripheral diabetic neuropathy. The application of age related reference values. Diabet Med 8 Symposium: S94-9 Beaton DE, Boers M, Wells GA (2002) Many faces of the minimal clinically important difference (MCID): a literature review and directions for future research. Curr Opin Rheumatol 14 (2):109–114 Becser N, Sand T, Zwart J-A (1998) Reliability of cephalic thermal thresholds in healthy subjects. Cephalgia 18:574–582
Int J Biometeorol (2012) 56:811–821 Bird SJ, Brown MJ, Spino C, Watling S, Floyt HL (2006) Value of repeated measures of nerve conduction and quantitative sensory testing in diabetic neuropathy trial. Muscle Nerve 34:214–224 Bland JM, Altman DG (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1(8476):307–310 Chong PS, Cros DP (2004) American Association Of Electrodiagnostic Medicine Review: Quantitative sensory testing equipment and reproducibility studies. Available at: http://www.aanem.org/ documents/qstReview.pdf. Accessed on December 21, 2010 Claus D, Hilz MJ, Neuendoerfer B (1990) Thermal discrimination thresholds: a comparison of different methods. Acta Neurol Scand 81:533–540 Dyck PJ, Kratz KM, Lehman KA, Karnes JL, Melton LJ et al (1991) The Rochester diabetic neuropathy study: design criteria for types of neuropathy, selection bias, and repeatability of neuropathic tests. Neurology 41:799–807 Gelber DA, Pfeifer MA, Broadstone VL, Munster EW, Peterson M et al (1995) Components of variance for vibratory and thermal threshold testing in normal and diabetic subjects. J Diabetes Complications 9:170–176 Guyton AC, Hall JE (2000) Textbook of medical physiology. W.B. Saunders Company, London Hagander LG, Midani HA, Kuskowski MA, Parry GJ (2000) Quantitative sensory testing: effect of site and skin temperature on thermal thresholds. Clin Neurophysiol 111(1):17–22 Hensel H, Andres KH, von Düring M (1974) Structure and function of cold receptors. Pflugers Arch 352:1–10 Hopkins WG (2000) Measures of reliability in sports medicine and science. Sports Med 30(1):1–15 Hsieh YW, Wang CH, Wu SC, Chen PC, Sheu CF et al (2007) Establishing the minimal clinically important difference of the Barthel Index in stroke patients. Neurorehabil Neural Repair 21(3):233–238 Huck SW, Cormier WH (1996) Reading statistics and research, 2nd edn. Harper Collins, New York
821 Kojo I, Pertovaara A (1987) The effects of stimulus area and adaptation temperature on warm and heat pain thresholds in man. Int J Neurosci 32(3–4):875–880 Meier PM, Berde CB, DiCanzio J, Zurakowski D, Sethna NF (2001) Quantitative assessment of cutaneous thermal and vibration sensation and thermal pain detection thresholds in healthy children and adolescents. Muscle Nerve 24:1339–1345 Merchut MP, Toleikis C (1990) Aging and quantitative sensory thresholds. Electromyogr Clin Neurophysiol 30:293–297 Moravcová E, Bednařík J, Svobodník A, Dušek L (2005) Reproducibility of thermal threshold assessment in small-fibre neuropathy patients. Scripta Medica (Brno) 78(3):177–184 Roebroeck ME, Harlaar J, Lankhorst GJ (1993) The application of generalizability theory to reliability assessment: an illustration using isometric force measurement. Phys Ther 73:386–401 Stratford PW, Finch E, Solomon P et al (1996) Using the RolandMorris questionnaire to make decisions about individual patients. Physiother Can 48:107–110 Strigo IA, Carli F, Bushnell MC (2000) Effect of ambient temperature on human pain and temperature perception. Anesthesiology 92:699–707 Valensi P, Attali J-R, Gagant S (1993) Reproducibility of parameters for assessment of diabetic neuropathy. French Group for Research of Diabetic Neuropathy. Diabet Med 10:933–939 Yarnitsky D (1997) Quantitative sensory testing. Muscle Nerve 20 (2):198–204 Yarnitsky D, Ochoa JL (1991) Warm and cold specific somatosensory systems. Psychophysical thresholds, reaction times and peripheral conduction velocities. Brain 114:1819–1826 Yarnitsky D, Sprecher E (1994) Thermal testing: normative data and repeatability for various test algorithms. J Neurol Sci 125:39– 45 Zwart J-A, Sand T (2002) Repeatability of dermatomal warm and cold sensory thresholds in patients with sciatica. Eur Spine J 11:441– 446
