Reliability of clinical pressure-pain algometric measurements obtained on consecutive days.

Algometry is a method of quantifying soft

tissue tenderness. An algometer registers the

force (in kilograms per square centimeter)

that is applied to the tissues via a small

rubber footplate. The force that is recorded is usually

the amount of pressure that causes pain, called the

pressure-pain threshold (PPT). Normal ranges of PPT have

been established for some muscles and for the sites of

some bony prominences.[1] The PPT has been used with

individuals without pain to assess the hypoalgesic effect

of physical therapy modalities. For instance, laser

therapy has been applied to normal peripheral sensory

nerves, and PPT was compared before and after the

intervention.[2]

Algometry in Delayed-Onset Muscle Soreness

Algometers have been used to measure tenderness

associated with inflammatory conditions.[1] Accumulation of

fluid in intracellular or extracellular spaces as a

consequence of injury raises tissue pressure and lowers PPT.[3,4]

Delayed-onset muscle soreness (DOMS)[3,5] is a condition

that occurs when untrained muscles perform strenuous

exercise. This condition develops between 24 and 48

hours after exercise, and it can be recognized by the

presence of pain on stretching, loss of force, stiffness,

and tenderness in the affected muscles. Blood enzyme

analysis[3] and muscle biopsy[6] reveal that there are

morphological and biochemical changes in untrained

muscles following strenuous eccentric exercise. Disruption

of myofibrils as well as supporting connective tissue has

been noted.[7] Some authors[7] have noted the presence of

cellular infiltrates, including neutrophils, macrophages,

and other inflammatory mediators. Tenderness is

thought to be due to swelling in the myofibrils and

extracellular space.

Delayed-onset muscle soreness can be induced for

experimental purposes. The premise for using DOMS in

research is that it results in a controlled injury being

imposed on the muscles. This control allows the time

course of the response to be examined, either under

different exercise conditions or under different

treatment conditions. We believe that algometers are quick

and safe to administer and are preferred over invasive

procedures as a daily measure of the effects of strenuous

exercise. The absence of any abnormal tenderness in the

muscle prior to inducing DOMS is a prerequisite of the

model.

Algometry has been used to monitor symptoms of

experimental DOMS. Jones et al[3] induced DOMS in the biceps

brachii muscle and used PPT and goniometry to

measure tenderness and stiffness, respectively, following the

exercise. Pressure-pain threshold has also been used to

assess differences in development of DOMS. Newham et

al,[5] for example, induced DOMS in the quadriceps

femoris muscle and used PPT to compare the

distribution of tenderness and the degree of tenderness induced

by two different exercise protocols. Pressure-pain

threshold also has been used to assess the effect of treatment

on DOMS. Hasson and colleagues,[4,8] for example,

investigated the effect of ultrasound and dexamethasone

iontophoresis on DOMS in the quadriceps femoris

muscle. They used PPT and a pain measure to evaluate

DOMS over a 48-hour period after exercise. Intervention

with ultrasound reduced the symptoms of DOMS, as

evidenced by less pain being reported and higher PPT

values in a treatment group than in a control group.

Dexamethasone iontophoresis was found to be

ineffective for the treatment of DOMS using the same outcome

measures. Jones et al,[3] Hasson et al,[4] and Newham et al[5]

reported changes in PPT measurements obtained after

exercise as compared with PPT measurements obtained

before exercise.

In none of the studies discussed was a non-DOMS

control group used to examine whether PPT changed as

a result of measurement procedures alone.

Pressure-pain threshold testing involves forcible probing of the

muscle surface. In individuals without pain, the PPT in

muscle may be as high as 11 kg/[cm.sup.2]. We have observed

that this amount of pressure may cause bruising. We

wondered, therefore, whether algometry at high

pressures over the same site daily might lead to progressive

lowering of PPT. We have not found studies addressing

the reliability of algometric measurements over

consecutive days. Thus, the reliability of PPT as an outcome

measure of DOMS has not been established.

Reliability Issues in Algometry

Fischer[1] studied the reliability of algometric

measurements in 10 muscles of 50 subjects without pain on a

single occasion. On the basis that there was no

difference between single measurements of corresponding

muscles on opposite sides of the body, Fischer

concluded that PPT was reproducible and proposed a range

of normal values. He noted that PPT varied between

individual muscles. The quadriceps femoris and biceps

brachii muscles are the muscles that are examined most

often in DOMS research. Fischer studied the quadriceps

femoris muscle, but we have not found any investigation

of normal PPT in the biceps brachii muscle. Abnormal

tenderness is an exclusion criterion for studies involving

DOMS.

Some authors[9-12] tested the reliability of repeated

measurements of PPT. They demonstrated that, although

measurements were not precise, differences between

trials did not exist. Reliability was confirmed by different

authors for several patterns of repetition of PPT,

including 10 to 50 consecutive measurements,[9,11] trials 45

minutes apart,[9] trials 1 hour apart,[10] and trials 1 week

apart.[11] Marking test sites was thought to be one method

of improving the reliability of PPT measurements.[10] The

reaction time of the examiner and variation in the rate

of pressure increase were other factors that affected

reliability.[9]

Nonelectronic algometers, such as the Fischer

algometer,(*) depend on the operator to control the rate of

pressure increase. Fischer[1] recommended a rate of

1 kg/[cm.sup.2]/s. Jensen et al[9] emphasized the importance

of increasing pressure at a standardized rate, based on

their finding that higher PPT scores were recorded at

higher application rates. Some authors[9-l3] used

electronic algometers to reduce variation in the rate of

pressure increase; the electronic tool provides

examiners with visual cues to improve their timing. Another

advantage of an electronic algometer is that the reaction

time of the examiner is eliminated; on reaching the pain

threshold, the subject activates a button to release

pressure. Jensen et al[9] thought that measurements of PPT

were most reliable when the measurement site was flat,

broad, and bony as opposed to a soft tissue site where the

footplate might slide off the target.

Kosek et al[12] used an electronic algometer and studied

three trials of PPT. In contrast to other authors, they

found a decrease in PPT between trials done 10 seconds

apart and an increase in PPT between trials done 20 to

30 minutes apart. The mean PPT of the three trials,

however, was not different from the mean PPT of three

trials after a 1-week interval. Other investigators[11,13] also

used the mean of multiple trials as a criterion score to

reduce variation across occasions.

Ohrbach and Gale[13] carried out a study to determine the

number of measurements that gave the best estimate of

PPT. They used an electronic algometer and measured

facial muscles five times each, at 4- to 5-minute intervals.

They found that PPT increased and decreased

unsystematically from trial to trial but that there was a correlation

between pairs of trials (Pearson r=.81-.91). Combining

trials showed that the mean of trials 1 and 2 provided a

more reliable estimate of PPT than either trial alone,

and the authors reported that more than three trials was

not justified by their data.

Merskey and Spear[14] investigated PPT using a

non-electronic algometer. They measured PPT twice on two

separate occasions. There were no differences across the

four trials. Their results, however, appear to support the

idea that an electronic algometer provides more reliable

measurements, because they reported a lower

between-trial correlation (Pearson r=.65) than that reported by

Orbach and Gale,[13] who used an electronic instrument.

In the same work, Merskey and Spear[14] examined the

interrater reliability of PPT measurements. They reported

that there was no difference between examiners, although

there was a tendency for one examiner to score higher

than the other examiner. The correlation between

examiners was reported as Pearson r=.59. In spite of the low

Pearson correlation coefficient, the authors stated that the

degree of reliability in their study supported the use of PPT

for investigation of the efficacy of analgesia.

Delaney and McKee[15] also examined the interrater and

intrarater reliability of PPT measurements. They used a

Fischer algometer. The results of their preliminary work

showed lower correlation between examiners (Pearson

r [is less than] .28) than that reported by Merskey and Spear.[14] They

attributed the finding to a difference in rate of pressure

increase, and they addressed the problem prior to

another study by training examiners to apply pressure

while being timed. Pain-pressure threshold was then

measured by two examiners alternately, at 5-minute

intervals, for a total of four trials per point.

Standardizing the timing of force application appears to have been

an effective strategy because high interrater and

intrarater reliability were reported in their final study

(intraclass correlation coefficients [ICCs]=.80-.92).

In summary, algometric measurements have been shown

to have good interrater and intrarater reliability when

the measurements were performed once or repeatedly

(2-50 repetitions) on a single day, at weekly intervals

(1-5 weeks), and at longer intervals (8-12 weeks).[8-14]

The reliability of measurements taken over consecutive

days has not been studied. Investigators have suggested

that electronic algometers provide more reliable

measurements than do nonelectronic algometers.[9-12] The

latter type of instrument, however, is more convenient to

use and is more commonly available.

Our study was designed to (1) examine the range of

"normal" PPT in the biceps brachii muscle, (2)

reexamine the intertrial and interrater reliability of PPT

measurements using a nonelectronic algometer on

asymptomatic muscle over 3 consecutive days, and (3) establish

the number of measurements needed for the best

estimate of PPT.

Method and Materials

Two examiners participated in the study. They were

physical therapists with many years of clinical experience

but no prior experience using an algometer. One week

prior to the study, the examiners practiced using an

algometer while being timed. The standard was to

increase pressure linearly to 5 kg/[cm.sup.2] over 5 seconds

according to the method recommended by Fischer.[1] Ten

practice trials were performed by each examiner. A

Fischer algometer was used for the practice and test

trials. The instrument has a 1-[cm.sup.2] rubber footplate and

a scale marked from 2 to 20 kg/[cm.sup.2], in increments of

0.2 kg/[cm.sup.2]. A new instrument was acquired for the

purpose of the study. No calibration was performed.

Thirty-five subjects without complaints of pain

volunteered and gave informed consent to participate in the

study. There was an imbalance of female subjects in the

sample (Tab. 1). The PPT of the biceps brachii muscle in

the nondominant arm of each subject was measured on

3 consecutive days by each examiner.

[TABULAR DATA 1 NOT REPRODUCIBLE IN ASCII]

Subjects were seated with their test arm positioned on a

padded support in 90 degrees of horizontal abduction,

with full elbow extension and forearm supination. The

upper arm was measured, and the skin overlying the

biceps muscle belly was marked with indelible ink, at a

point one fourth of the distance from the elbow crease

to the lateral border of the acromion. This mark

established the site for all testing. Each subject's non-test arm

was similarly measured and marked for the purpose of a

practice session to familiarize subjects with the sensation

of PPT.

Standardized instruction was given prior to each trial on

all occasions. Subjects were instructed to "report as soon

as the sensation of pressure changes to pain by saying

`pain,' and I will stop." The footplate of the algometer

was held perpendicular to the muscle belly with the

gauge turned away from the subject and the examiner.

Pressure was increased at a rate of approximately 1

kg/[cm.sup.2]/s until the subject reported "pain." The examiner

then released the pressure and lifted the algometer off

the muscle to read the gauge and record the

measurement. The needle on the gauge was returned to baseline

before each trial using the pressure-release button on

the algometer. Subjects were kept uninformed of their

scores throughout the study to prevent subject bias from

influencing the results.

The first examiner did three practice trials on each

subject's non-test arm. The practice trials were followed

by three trials of measuring PPT on the subject's test

arm, with 10-second intervals between trials. After a

20-minute interval, the procedure was repeated by the

second examiner, using the marked sites on the non-test

arm followed by the marked sites on the test arm. On

days 2 and 3, procedures were repeated on the test arm

only, using the same sequence and timing. Thus, each

subject's test arm was measured three times on each of 3

days (9 trials) by each examiner for a total of 18 trials.

For 20 subjects, the order of daily testing was examiner A

followed by examiner B. The order of examiners was

reversed for 15 subjects. During the study, examiners did

not have access to each other's scores or to their own

scores of previous days. No analysis was done until data

collection was complete.

Data Analysis

Intraclass correlational analyses (Shrout and Fleiss

formula, ICC[2,1], a two-way random-effects layout[16]) were

used to estimate interrater, trial-to-trial, and day-to-day

reliability. Interrater reliability was estimated for each of

the nine trials and for scores derived from the mean

score of various combinations of trials.

Trial-to-trial reliability was estimated by correlating the

trial 1 and trial 2 scores and the trial 2 and trial 3 scores

each day, as well as by computing the correlation among

all three trials on each day. Day-to-day reliability was

estimated by computing the correlation between single

trials of like number in the sequence of daily trials.

Correlations were also calculated for day-to-day scores

derived from the mean score of a combination of

like-numbered trials in the sequence of daily trials.

A plot of the data showing the relationship in scores

between trials and between examiners suggested that the

data varied considerably from the line of equality. We

considered it necessary, therefore, to further analyze the

data using graphical techniques, as recommended by

Bland and Altman[17] and as described.

Graphical analysis: interrater reliability of single trials.

A subject's scores in a single trial recorded by each of the

two examiners were paired for comparison. The mean of

each pair was plotted against their difference. The

overall (n=35) mean difference (d) and standard

deviation of the differences were calculated for each of the

nine trials.

Using the SAS Univariate procedure,(dagger) differences were

found to be normally distributed. Most of the differences

(95%) could, therefore, be expected to lie between the

mean difference and approximately two standard

deviations (d [+ or -] 1.96SD), which was interpreted as the "limits

of agreement."[17]

Graphical analysis: interrater reliability of repeated

measurements. Each examiner calculated each

subject's mean score for two consecutive trials within the

same day. Means were computed for trials 1 and 2 and

trials 2 and 3 daily. Each subject's mean scores, derived

from like-numbered trials by each examiner, were paired

for comparison. The mean of the paired scores was

plotted against their difference. Overall mean difference

(n=35) and limits of agreement were calculated as

described previously.

A correction, according to the method of Bland and

Altman,[17] was applied to calculate the standard deviation

of the differences between paired scores, which were

derived from the means of multiple measurements. The

correction was to compensate for removal of some of the

measurement error. The graphical technique was not

used for analysis of the mean of more than two repeated

measurements because of the risk of underestimation of

the standard deviation of the differences.

Graphical analysis: trial-to-trial and day-to-day reliability

of single trials. The scores of each examiner were

analyzed separately. A subject's score in one trial was

paired for comparison with the score in the subsequent

trial on the same day to assess trial-to-trial reliability and

with the score in the like-numbered trial in the sequence

of trials on the subsequent day to assess day-to-day

reliability. Means and differences were plotted, and the

overall mean difference (n=35) and limits of agreement

were calculated as described previously.

Effect of order of examiners. A one-way analysis of

variance (ANOVA) on the mean difference between

examiners in trial 1 was used to assess whether the order

of examiners affected the results.

Results

A total of 630 PPT scores were collected. The mean PPT

and standard deviation are shown for each trial and each

examiner in Table 2. Some subjects had bruising at the

measurement site by the third day of the study.

Table 2.

Repeated Measurements of Pressure-Pain Threshold (PPT) for

35 Subjects

        PPT (kg/[cm.sup.2])

        Examiner A            Examiner B

Trial   X      SD             X      SD



Day 1

  1     3.41   0.98           3.27   1.37

  2     3.50   1.09           3.23   1.36

  3     3.54   1.23           3.26   1.46

Day 2

  1     3.41   1.32           3.05   1.16

  2     3.39   1.38           2.98   1.24

  3     3.39   1.44           3.12   1.48

Day 3

  1     3.39   1.28           3.01   1.15

  2     3.41   1.31           3.00   1.37

  3     3.44   1.34           3.09   1.52

Normative PPT Values for the Biceps Brachii Muscle

The mean PPT in the biceps brachii muscle was 4.63

kg/[cm.sup.2] (range=2.04-10.32) for the male subjects (90

scores) and 3.05 kg/[cm.sup.2] (range=1.81-6.80) for the

female subjects (540 scores).

Interrater Reliability

Examiner A recorded higher scores than those recorded

by examiner B in about 70% of the paired

measurements. As mean PPT increased, however, examiner A

tended to score increasingly lower than examiner B did.

At a mean PPT of approximately 7.0 kg/[cm.sup.2], examiner

A recorded a score that was 2.5 kg/[cm.sup.2] lower than the

score recorded by examiner B. On balance, however, the

mean difference between examiners (0.14 kg/[cm.sup.2] in

trial 1) was small.

Table 3 shows interrater ICCs (2,1) for single trials and

for mean scores derived from various combinations of

trials; all correlations were significant at P [is less than] .0001. Each

day reliability was lowest for the first of the single trials

and highest for the third of the single trials (ICC=.74-.89).

Table 3.

Intraclass Correlation Coefficients (ICC[2,1]) Between Two Examiners

for Measurements of Pressure-Pain Threshold for Single Trials or for

Scores Derived From the Mean of Multiple Trials for 35 Subjects

        Single   Interrater                     Interrater

        Trials   ICC          Multiple Trials   ICC



Day 1   1        .74           1, 2    (mean)   .81

        2         84           1, 2, 3 (mean)   .85

        3        .89           2, 3    (mean)   .88



Day 2   1        .75           1, 2    (mean)   .82

        2        .84           1, 2, 3 (mean)   .84

        3        .84           2, 3    (mean)   .86



Day 3   1        .78           1, 2    (mean)   .82

        2        .82           1, 2, 3 (mean)   .84

        3        .86           2, 3    (mean)   .85

Reliability improved when the mean

score of the three daily trials was used

rather than the score of the first or

second trial of the day. The highest

reliability, however, was seen when the

score of the first trial of each day was

omitted and the mean of the second

and third trials of the day

(ICC=.85-.88), or the score of the third trial alone

(ICC=.84-.89), was used as the

criterion score.

Figure 1 shows the relationship

between examiners of the scores

recorded in trial 1. The line of equality

is shown on which all points would lie if

the two examiners recorded identical

scores. The variation from the line of

equality shown in Figure 1 is typical of

the results of all the single trials and

prompted the additional analyses using

Bland and Altman's methods.[17]

[FIGURE 1, GRAPH OMITTED]

Figures 2 and 3 show the agreement

between examiners using Bland and

Altman's methods.[17] Each subject is

represented by a point (n=35) that

shows the difference in scores between

examiner's against their mean score.

Points on the zero line show perfect

agreement. The overall mean

difference between examiners is shown by a

broken line. The limits of agreement

are also shown (d [+ or -] 1.96SD).

[FIGURE 2 & 3, GRAPH OMITTED]

Figure 2 illustrates the added benefit of

using Bland and Altman's methods[17]

for the data obtained in trial 1. The

mean difference between examiners in

this trial was 0.14 kg/[cm.sup.2] (SD=0.86).

From the limits of agreement, it can be

projected that if one examiner

measures PPT once, a second examiner

would score the same subjects within

1.55 kg/[cm.sup.2] below and 1.83 kg/[cm.sup.2]

above the first examiner's

measurement 95% of the time. Agreement

between examiners was better in some

later trials than in trial 1. The limits of

agreement for day 2 of trial 3 for

example, were from -1.22 to +1.78 kg/[cm.sup.2].

There was little change in agreement between examiners

when the measure was derived from the mean score of

the first two trials rather than the scores of the first trial

each day. For example, when the measurement was

based on subjects' mean score for day 1 of trials 1 and 2,

the limits of agreement between examiners were from

-1.29 to +1.71 kg/[cm.sup.2].

On all 3 days, however, agreement between examiners

was best when the measurement was derived from the

subjects' mean score of trials 2 and 3. Figure 3 shows the

results for day 1; the limits of agreement lie between

-0.97 and +1.47 kg/[cm.sup.2].

Trial-to-Trial Reliability

Tables 4 and 5 show the trial-to-trial and day-to-day

reliability (ICC[2,1]) of measurements of PPT. With the

exception of day 1, trial-to-trial reliability was higher

between trials 2 and 3 than between trials 1 and 2.

Day-to-day reliability for a single measurement of PPT

was highest in trial 3, and day-to-day reliability for a

measurement derived from the mean of multiple trials

was highest for the mean of trials 2 and 3.

Table 4.

Trial-to-trial intraclass Correlation Coefficients (ICC[2,1]) for

Pressure-Pain Threshold,(a) Rated by Examiner B, for 35 Subjects

        Trials    ICC   Trials    ICC   Trials             ICC



Day 1   1 and 2   .98   2 and 3   .96   All trials (1-3)   .96

Day 2   1 and 2   .94   2 and 3   .95   All trials (1-3)   .93

Day 3   1 and 2   .94   2 and 3   .98   All trials (1-3)   .95

(a) Correlations are between single trials on the same day.

Table 5.

Intraclass Correlation Coefficients (ICC[2,1]) for Day-to-Day

Pressure-Pain Threshold,(a) Rated by Examiner B, for 35 Subjects

Trial                            ICC



Trial 1 x 3 days                 .88

Trial 2 x 3 days                 .88

Trial 3 x 3 days                 .89

Trials 2 and 3 (mean) x 3 days   .90

(a) Correlation are between single trials of like number in the sequence of

trials on different days or scores derived from the mean of multiple trials on

one day correlated with the mean of multiple trials of like number in the

sequence of trials on subsequent days.

Figure 4 shows the trial-to-trial agreement of the scores

recorded by examiner B on day 1 of trials 2 and 3, using

Bland and Altman's method of graphical analysis.[17] The

results were similar on days 2 and 3. A point is plotted for

each subject (n=35) showing the difference in scores

between trials 2 and 3 against their mean. Perfect

agreement (zero line) and overall mean difference

(-0.03 kg/[cm.sup.2], SD=0.42) between the two trials are

shown. The limits of agreement lie between -0.86 and

+0.79 kg/[cm.sup.2]. Thus, the PPT in the third trial was

between 0.86 kg/[cm.sup.2] below and 0.79 kg/[cm.sup.2] above the

PPT in the second trial 95% of the time.

[FIGURE 4, GRAPH OMITTED]

The same methods of analysis for the scores recorded by

examiner A in trials 2 and 3 showed an overall mean

difference between trials of -0.04 kg/[cm.sup.2] (SD=0.35).

Limits of agreement were between -0.72 and +0.64

kg/[cm.sup.2].

Order of Examiners

The ANOVA revealed that the order in which the

examiners measured PPT had no effect on the

differences between their scores (P=.33).

Discussion

The purpose of our study was to establish the normal

range of PPT values in the biceps brachii muscle because

this muscle is frequently used in studies of

experimentally induced DOMS and to examine interrater,

trial-to-trial, and day-to-day reliability of algometric

measurements in healthy muscle. If PPT proved to be a stable

measure in the absence of pathology, then it could be

used as an outcome measure in studies of experimentally

induced DOMS.

We observed PPT in a group of individuals without pain

whose ages were within a fairly restricted range. The

group included a disproportionate number of female

subjects. These factors might limit the applicability of

our results. The height and weight of the subjects were

fairly representative of the adult population. All subjects

tolerated 18 measurements over 3 days, although some

subjects showed bruising at the measurement site.

Fischer[1] suggested that PPT is reproducible between

individual subjects. He calculated PPT from the mean of

two measurements taken on contralateral sides of the

body and examined the distribution of values in a study

of nine healthy muscles. Based on the standard deviation

from the average logarithmic values of the PPT findings

in his study, Fischer proposed that for diagnostic

purposes and for quantifying pain, 84.1% of mean PPT

should be considered a cutoff value for "normal."

Fischer did not study PPT in the biceps brachii muscle.

In our study, mean PPT for the biceps brachii muscle in

the female subjects was 3.05 kg/[cm.sup.2]. Using 84.1% of

mean PPT as a cutoff, as recommended by Fischer, the

lowest normal value for the biceps brachii muscle in

female subjects would be 2.44 kg/[cm.sup.2]. This value should

be taken into consideration when screening subjects for

admission to studies of DOMS involving the biceps

brachii muscle.

Intraclass correlation coefficients appear frequently in

the literature as an index of reliability. The examiners in

our study were not experienced in algometric

techniques. Thus, they represent examiners as broadly

defined, not a particular set of examiners, and we believe

that the random-effects model (ICC[2,1]) applies.[16]

Using a measurement derived from the

mean score of trials 2 and 3 daily, PPT

appears to yield reliable measurements

of muscle tenderness over a 20-minute

period and over 3 consecutive days,

according to ICC analyses. Moreover,

the ICCs suggest that two examiners

could be used interchangeably to

measure PPT.

Delaney and McKee,[15] using the

Fischer algometer on muscle, also

reported lower reliability for the first of

two trials. They considered their

examiners to be experienced in algometric

techniques, and they timed their

examiners' rate of pressure application in an

attempt to improve reliability. The

reported reliability (ICC=.80-.92) was

similar to ours, and our examiners were

not timed during the testing. Ohrbach

and Gale[13] similarly concluded that

measurements obtained in their first

trial did not agree well with the mean

of five trials of PPT. On the basis of the

95% confidence interval around the mean of five trials,

they recommended the use of data recorded during trial

2 or the mean of trials 1 and 2 to estimate PPT. Our

results did not support the use of trial 2 alone or the

mean score of trials 1 and 2.

Some authors[9-12] have argued that the reliability of

measurements of PPT was improved by the use of

electronically controlled instruments. In our study, we

used a nonelectronic instrument. In spite of the

difficulty of maintaining the recommended rate of pressure

increase of 1 kg/[cm.sup.2]/s using our type of algometer,

there was still a positive correlation between trials in our

study, and our reliability was comparable to that

obtained with electronic algometers.

The use of correlation coefficients to assess the repeatability

of measurements is misleading, according to Bland and

Altman.[17] These authors noted that correlation coefficients

measure the relationship between two measurements, not

the agreement between them. Because the examiners in

our study measured the same subjects, and measured them

repeatedly, we would expect the scores between examiners

and between trials to be strongly related. In accordance

with the recommendations of Bland and Altman,[17] we

plotted the measurements of one examiner against those

of the other examiner to assess visually whether the data

varied from the line of equality. Bland and Altman[17] noted

that two sets of measurements that agree perfectly lie on

the line of equality; measurements that are highly correlated

lie along any straight line. A plot of our data (Fig. 1)

showed considerable variation around the line of equality,

especially at higher values of PPT. We believed, therefore,

that additional methods of analysis were indicated.

When we used graphical methods to assess agreement,

our findings supported the opinion of Bland and

Altman[17] that high correlations can exist with concurrent

lack of good agreement between measurements (Figs.

1-4). Graphical analysis of the data provided

information about trials and raters that was not

obvious from the ICCs.

From the distribution of the

measurements around zero, it was apparent that

although the examiners agreed well on

average (ie, small mean difference),

there were quite large differences

between them for individual subjects

(Figs. 2-4). Furthermore, examination

of the relationship between differences

and means showed that the differences

were affected by the size of the

measurement, in that differences between

examiners were larger and in an

opposite direction, at high mean values of

PPT than at low mean values. The

change of direction of differences

between the examiners as mean PPT

increased was unexpected and warrants

some explanation. We speculated that

at low mean PPT, examiner A used a

faster rate of pressure application than

examiner B used. Jensen et al[9] noted

that higher rates tended to produce a

higher PPT. We think that the reason for this finding is

that examiner reaction time is slowed by high rates of

pressure increase, leading to overestimation of PPT. Why

then did examiner A not overestimate at high levels of

PPT? A self-limiting factor might have been that

examiner A was not able to maintain the high rate of pressure

increase at the highest levels of force encountered in our

study. This explanation would account for the change of

direction of differences between examiners.

Graphical methods, in contrast to ICCs, demonstrate

measurement error in units that are clinically

meaningful (eg, kilograms per square centimeter) so that the

consequence of differences between methods can be

assessed. For example, Figure 2 shows that the limits of

agreement between examiners based on a single rating

of PPT were -1.5 to +1.8 kg/[cm.sup.2]. Figure 3 shows that by

using the mean of multiple measurements, the

measurement error between examiners is reduced to -1.0 to

+1.5 kg/[cm.sup.2]. In our opinion, however, a difference

between examiners of up to 1.5 kg/[cm.sup.2] on a

measurement that has a "true" value (mean measurement of two

examiners) of 3 to 5 kg/[cm.sup.2] is large, especially in view of

our finding that the measurement error between trials

by one examiner was from -0.9 to +0.8 kg/[cm.sup.2] (Fig. 4).

Our findings from the graphical analysis suggest that

one examiner should perform all measurements of PPT.

Graphical analysis of trial-to-trial agreement (Fig. 4)

showed that there was no systematic bias of one trial

relative to another trial (consistent upward or downward

shift in PPT). Because the results were similar for all

trial-to-trial comparisons, we conclude that there was no

effect of repeated use of the algometer. There was also

no effect of size of measurement on the size or direction

of trial-to-trial differences. Trial-to-trial results were

similar for both examiners. Thus, we conclude that

examiners A and B were consistent in their individual

techniques, even though there were differences between

their scores.

Our study lends support to previous work that has shown

that measurements of PPT are highly reliable in

individuals without pain.[9-13] Reliability improves when three

trials are performed and data from the last two trials are

used to determine PPT. Variation in rate of pressure

increase may be the factor most affecting reliability. To

minimize this effect, we believe that testing should be

performed by one examiner.

The six PPT ratings that we performed daily for 3 days,

in two sets of three with a 20-minute interval each day is

typical of measurement in intervention studies involving

DOMS. Our testing procedure did not, in itself, effect a

change in PPT. We have shown that PPT can be used as

an outcome measure in the treatment of persons with

DOMS.

Conclusion

Measurements of PPT in healthy muscle obtained with a

simple nonelectronic algometer were reliable from trial

to trial within the same day and from day to day over 3

consecutive days. Measurements by one examiner were

more reliable than measurements between examiners.

We have demonstrated that reliability is improved when

the first of three trials is excluded for estimating the

"true" PPT. The algometer appears to have potential for

measuring day-to-day changes in soft tissue tenderness in

persons with DOMS.

(*) Pain Diagnostics and Thermography Inc, 233 E Shore Rd, Suite 106, Great

Neck, NY 11023.

(dagger) SAS Institute Inc, SAS Campus Dr, Cary, NC 27513.

References

[1] Fischer A. Pressure algometry over normal muscles: standard values,

validity, and reproducibility of pressure threshold. Pain.

1987;30:115-126.

[2] Wylie L, Baxter GD, Walsh DM, Robinson L. The hypoalgesic effects

of low-intensity infrared laser therapy upon mechanical pain

threshold. Lasers Surg Med Suppl. 1995;17:9.

[3] Jones DA, Newham DJ, Clarkson PM. Skeletal muscle stiffness and

pain following eccentric exercise of the elbow flexors. Pain.

1987;30:233-242.

[4] Hasson S, Mundorf R, Barnes W, et al. Effect of pulsed ultrasound

versus placebo on muscle soreness perception and muscular

performance. Scand J Rehabil Med. 1990;22:199-205.

[5] Newham DJ, Mills KR, Quigley BM, Edwards RHT. Pain and fatigue

after concentric and eccentric muscle contractions. Clin Sci.

1983;64:55-62.

[6] Stauber WT, Clarkson PM, Fritz VK, Evans WJ. Extracellular matrix

disruption and pain after eccentric muscle action. J Appl Physiol.

1990;69:868-874.

[7] MacIntyre DL, Reid WD, McKenzie DC. Delayed muscle soreness: the

inflammatory response to muscle injury and its clinical implications.

Sports Med. 1995;20:24-40.

[8] Hasson S, Wible C, Reich M, et al. Therapeutic effect of

iontophoretically delivered dexamethasone on delayed muscle soreness. Phys Ther.

1989;69:389. Abstract.

[9] Jensen K, Andersen HO, Olesen J, Lindblom U. Pressure-pain

threshold in human temporal region: evaluation of a new pressure

algometer. Pain. 1986;25:313-323.

[10] Vatine J, Shapira SC, Magora F, et al. Electronic pressure algometry

of deep pain in healthy volunteers. Arch Phys Med Rehabil.

1993;74:526-530.

[11] Brennum J, Kjeldsen M, Jensen K, Jensen T. Measurements of

human pressure-pain thresholds on fingers and toes. Pain.

1989:38:211-217.

[12] Kosek E, Ekholm J, Nordemar R. A comparison of pressure-pain

thresholds in different tissues and body regions. Scand J Rehabil Med.

1993;25:117-124.

[13] Ohrbach R, Gale EN. Pressure-pain thresholds in normal muscles:

reliability, measurement effects, and topographic differences. Pain.

1989;37:257-263.

[14] Merskey H, Spear FG. The reliability of the pressure algometer.

British Journal of Social and Clinical Psychology. 1964;3:130-136.

[15] Delaney GA, McKee AC. Inter- and intra-rater reliability of the

pressure threshold meter in measurement of myofascial trigger point

sensitivity. Am J Phys Med Rehabil. 1993;72:136-139.

[16] Shrout P, Fleiss JL. Intraclass correlations: uses in assessing rater

reliability. Psychol Bull. 1979;86:420-428.

[17] Bland JM, Altman DG. Statistical methods for assessing agreement

between two methods of clinical measurement. Lancet. February 8,

1986:307-310.