Sensory perception testing by monofilaments in the digits of controls and workers with havs

Analysis and statistics

For each study population, the ‘normal’ cut-off was esti-

mated using the 95th percentile of the sensory thresholds 

of each digit. As the bend forces chosen for this study are 

not truly continuous, it would not be appropriate to assume 

a particular distribution so the 95th percentile was taken 

directly from the data. The 95% confidence intervals (95% 

CI) were bootstrapped with the subject as the resampling 

unit, thereby taking into account any clustering due to mul-

tiple digits per person. Results were analysed by sex, age, 

hand and digit for each study population. Because of the 

method used to estimate the 95th percentiles and CIs, no 

formal statistical tests were undertaken to compare the 95th 

percentiles across groups. Instead it was assumed that if the 

CIs did not overlap, then the two values were significantly 

different at the 

p

 < 0.05 level. The 

t

 test was used to compare 

group ages.

To determine intra-digit variability, 20 office workers and 

10 heavy manual workers were re-tested in the same way and 

by the same investigator two to four weeks later. The degree 

of agreement between the two measurement occasions was 

assessed using the percentage agreement and linear weighted 

kappa. The 95% CIs were bootstrapped with the subject as 

the resampling unit.

For workers with HAVS, the mean sensory perception 

threshold of the two digits with the highest thresholds was 

ascertained. This was compared with the number of abnor-

mal thermal (hot and cold) and vibration perception (31.5 

and 125 Hz) thresholds in the index and little fingers of the 

same hand. On the same day, thermal aesthesiometry was 

undertaken according to the method described by Lindsell 

and Griffin 

2003

 and vibration perception according to 

ISO 13091–2 (

2003

) and as described (Poole et al. 

2016

). 

Thresholds > 48.5 °C for hot or < 19.0 °C for cold; > 0.4 m/s

2

 

for 31.5 Hz or > 1.0 m/s

2

 for 125 Hz were taken as abnormal 

(HSE 

2005

; Lindsell and Griffin 

2003

), making the maxi-

mum QST score per finger four.

The relationship between the mean SWM threshold of 

the two digits with the highest thresholds and the number of 

QST abnormalities in the same hand of subjects with HAVS 

was modelled using mixed effects Poisson regression with 

the subject as the random effect and using a restricted cubic 

spline with three knots. The small number of results ≥ 10 g-f 

were excluded. The utility of SWM for identifying abnor-

malities of QST was assessed using Receiver Operating 

Characteristic (ROC) curves and sensitivity and specificity 

calculated at different sensory thresholds.

Results

There were 300 office workers of whom 155 were male, 

144 female and one unstated. Their mean age was 42 (range 

19–68) years. There were 272 who declared their right-hand 

and 27 their left-hand to be dominant, with one declaring 

ambidexterity. The ambient temperature of the laboratory 

varied between 22–24 °C and humidity between 30 and 

35%. Table 

1

 shows sensory perception broken down by 

sex, age, digit and hand-dominance. The 95th percentile 

was 0.16 g-f and did not vary by sex, hand-dominance or 

digit. Women < 30 had a significantly lower 95th percentile 

for sensory perception than women ≥ 30 years. Men had a 

higher 95th percentile for the dominant thumb (0.40 g-f) 

compared with the other digits, but as the CIs overlapped the 

difference did not reach statistical significance.

The frequency distribution of sensory perception for the 

office workers showed the median to be 0.07 g-f with no 

digit with a threshold greater than 0.6 g-f (Table 

2

). The 

data were positively skewed towards the lower sensory 

thresholds.

There were 115 heavy manual workers all of whom 

were male. Their mean age was 40 (range 18–66) years, 

which was not significantly different from the office workers 

(

p

 = 0.114). There were 99 who declared their right-hand and 

14 their left-hand to be dominant, with one declaring ambi-

dexterity and one unstated. There was no difference in sen-

sory perception by hand dominance. The 95th percentile for 

the dominant thumb was 1.00 g-f (95% CI 1.00–1.40), which 

was significantly higher than the middle 0.60 g-f (95% CI 

0.60–0.60) and the ring fingers 0.60 g-f (95% CI 0.40–0.60) 

on the dominant hand.

Table 

2

 shows the frequency distribution of the sensory 

perception thresholds for the office and heavy manual work-

ers. The median threshold for heavy manual workers was 

0.16 g-f and there was no threshold > 2.00 g-f. The 95th 

percentile was 1.00 g-f (95% CI 0.60–1.00), which was 

significantly greater than for office workers 0.16 g-f (95% 

CI 0.16–0.16). Heavy manual workers ≥ 50 years had the 

highest 95th percentile of 1.4 g-f (95% CI 1.00–2.00), 

but there was no consistent trend with age; 40–49 years 

0.60 g-f (95% CI 0.40–1.00); 30–39 years 0.40 g-f (95% CI 

0.28–0.60); < 30 years 0.60 g-f (95% CI 0.60–1.00).

Table 

3

 shows the degree of agreement between 

repeated measurements on the same digits. There was per-

fect agreement for 198/300 (66%) of digits; 92/300 (31%) 

differed by one filament; 10/300 (3%) differed by two 

---

 

International Archives of Occupational and Environmental Health

1 3

filaments. No finger differed by more than two filaments. 

The weighted kappa statistic for intra-subject reliability 

was 0.63 (95% CI 0.53–0.70). The degree of agreement 

for the office workers was not substantially different from 

that of the heavy manual workers.

There were 62 cases of neurological HAVS all of whom 

were male. Their mean age was 51 (23–69) years. They 

were older than the office (

p

 < 0.001) and the heavy man-

ual workers (

p

 < 0.001). There were 55 who declared their 

right-hand and 7 their left-hand to be dominant. There 

were no significant differences in the 95th percentiles of 

sensory perception by age, digit or hand dominance. Of the 

124 hands that were classified and neurologically staged 

according to the ICC (Poole et al. 

2019

); 4 were stage 0; 

66 were stage 1; 36 were stage 2 and 18 were stage 3.

For workers with HAVS the median threshold of per-

ception by SWM was 0.6 g-f and 11 digits had a thresh-

old ≥ 10  g-f. The 95th percentile was 4.0  g-f (95% CI 

2.0–6.0), which was significantly greater than the office 

workers and the heavy manual workers (confidence intervals 

not overlapping). Table 

4

 shows the frequency distribution 

of sensory perception thresholds for workers with HAVS.

Figure 

1

 shows the spread of sensory perception in the 

digits by SWM for the three study populations. The range 

increased from office workers (0.04–0.60 g-f), to heavy man-

ual workers (0.04–2.0 g-f), to HAVS cases (0.04 to ≥ 10 g-f).