evaluation of a dipstick test for glucose in urine

CLIN. CHEM. 22/2, 205-210 (1976)

CLINICAL CHEMISTRY, Vol. 22, No. 2, 1976 205

Evaluation of a Dipstick Test for Glucose in Urine

J#{216}rnDyerberg,1 Lissi Pedersen,2 and Ole Aagaard2

As an example of qualitative tests, a dipstick analysis

for glucose in urine has been tested for the influence ofmodifying factors on the test result. Two different typesof dipsticks were examined, “Clinistix” and “S-Gluko-test.’ ‘ Used according to manufacturer’s instructions,the latter is more sensitive and selective. By multivar-iance analysis the following variables were. examined:urine samples, inter- and intra-analyst, exposure to light,and dipstick batch. The first three contributed signifi-cantly to the total variation in results, inter-specimenvariation being the most important. With knowledge ofthe frequency of testing urines with a given glucoseconcentration and the probability of the result at thatconcentration, an expression of the probability of theglucose content of a urine sample can be obtained.Even with the tests of the type examined having a sensi-tivity and specificity exceeding 95 % , 1 4 of 100 patientssuspected of having diabetes mellitus on the basis of adipstick examination will be found to have a urinary glu-cose concentration of <2 mmol/liter. These figureswere found when the prevalence of urines with a glu-

cose concentration exceeding 2 mmol/liter was 17.5%.

AddItional Keyphrases: analysis of variance #{149}screening

Tests that give either a positive or a negative me-

sult, the so-called qualitative tests, make up a quite

considerable part of the diagnostic process. Such

tests are used extensively in the clinical chemical

field, and as the binary result decides whether the

patient will be further examined and possibly treat-

ed, it is necessary to know the efficiency and level of

discrimination of the basic analyses, as well as what

effect modifying factors will have on the test results.

Furthermore, these analyses must be controlled, just

as quantitative analyses are.

Qualitative tests performed with “dipsticks” are

widely used because of their simplicity. However, in

our opinion these tests are often performed without

sufficient knowledge of the possibilities of misinter-pretation and without the necessary measures being

taken against modifying factors. We have tried here

to illustrate (a) which modifying factors may influ-

ence results of such tests, and (b) what precautions

must be taken.

1 Clinical Chemical Department, Aalborg Hospital North, Box

561, 9100 Aalborg, Denmark.2 Clinical Chemical Department, Aalborg Hospital South, Box

365, 9100 Aalborg, Denmark.

Received Sept. 3, 1975; accepted Nov. 12, 1975.

As an example of these tests we have chosen the

test for glucose in urine, because this test is one of

the most frequently performed. Also, glucose is a

well-defined substance that can be quantitated spe-

cifically.

Material

We used dipsticks from Ames Co., Miles Lab. Ltd.,

England (“Clinistix”) and from Boehninger

Mannheim Corp., Waldhof, Germany (“5-Gluko-

test”); both were kindly supplied by the manufactur-

ems.

Experimental Design and Results

Factors Affecting Color Change

For both dipsticks, interpretations are based on a

change in color to be observed at a fixed time after

immersing the dipsticks in urine. Clinistix gives a

positive reading when the color changes from med to

blue after 10 s. 5-Glukotest changes from yellow to

green after 30 s. The sensitivity is given as 5-6 mmol/

liter for Clinistix and as 2-3 mmol/litem for 5-Gluko-

test. This accounts for the initial color change of the

dipstick. Although the manufacturers say it is possi-

ble to semiquantify a positive result by the intensity

of the color obtained, we used only the results from

the initial color change.

The purpose of this part of the study was to estab-

lish the effect of modifying factors on the total varia-

tion of the color change: variation among samples

(urine factor), interperson variation (laboratory

technician factor), intra-person variation (time fac-

tor), among batches (batch factor), and finally, varia-

tion of light (environmental factor).

The effect of the above factors on color change has

been examined by multivaniance analysis, in which

the interaction terms are presumed to be negligibleand where the factors have been picked at random-

following a so-called extended Greek Roman square

(1 ). Each factor is present in the square at 13 levels.

The test plan is illustrated in detail in Figure 1. By

such analysis it is possible to decide the relative con-

tribution of a particular factor to the total variation

of the system. Any combination of two factors will be

encountered only once, and the effect of a particular

factor can be compared at different levels because all

the other factors make the same contribution at a

given level, thus neutralizing each other.

LAST ECu.

URI

POINT OF

CHNIGE TYPE OFtIOL/L HAIIG[

206 CLINICAL CHEMISTRY, Vol. 22, No. 2, 1976

1 2 3 14 5 6 7 8 9 10 11 12 13

T 1 13 12 11 10 9 8 7 6 5 � 3 2b 1 13 12 11 10 9 8 7 6 5 � 3 2[ 1 12 10 8 6 � 2 13 11 9 7 5 3

2 T 2 1 13 12 11 10 9 8 7 6 5 � 3

B 3 2 1 13 12 11 10 9 8 7 6 5 4

E � 2 13 11 9 7 5 3 1 12 10 8 6

3 T 3 2 1 13 12 11 10 9 8 7 6 5 �

i� 5 �4 3 2 1 13 12 11 10 9 8 7 6E 7 5 3 1 12 10 8 6 14 2 13 11 9

L4 T � 3 2 1 13 12 11 10 9 8 7 6 5

B 7 6 5 LI 3 2 1 13 12 11 10 9 8i: 10 8 6 L4 2 13 11 9 7 5 3 1 12

5 T 5 14 3 2 1 13 12 11 10 9 8 7 6

i� 9 8 7 6 5 14 3 2 1 13 12 11 10L 13 11 9 7 5 3 1 12 10 8 6 � 2

6 1 6 5 14 3 2 1 13 12 11 10 9 8 7

B 11 10 9 8 7 6 5 � 3 2 1 13 12E 3 1 12 10 8 6 � 2 13 11 9 7 5

I I 7 6 S L4 3 2 1 13 12 11 10 9 8E 13 12 11 10 9 8 7 6 5 � 3 2 1

E 6 4 2 13 11 9 7 5 3 1 12 10 8

8 T 8 7 6 5 ‘4 3 2 1 13 12 11 10 9

B 2 1 13 12 11 10 9 8 7 6 5 � 3E 9 7 5 3 1 12 10 8 6 L4 2 13 11

9 T 9 8 7 6 5 �4 3 2 1 13 12 11 10b � 3 2 1 13 12 11 10 9 8 7 6 5

L 12 10 8 6 14 2 13 11 9 7 5 3 1

10 1 10 9 8 7 6 5 LI 3 2 1 13 12 11B 6 5 14 3 2 1 13 12 11 10 9 8 7

E 2 13 11 9 7 5 3 1 12 10 8 6 �

11 1 ii 10 9 8 7 6 5 L4 3 2 1 13 12B 8 7 6 5 14 3 2 1 13 12 11 10 9E 5 3 1 12 10 8 6 L4 2 13 11 9 7

12 1 12 11 10 9 8 7 6 5 14 3 2 1 13

B 10 9 8 7 6 5 L4 3 2 1 13 12 11E 8 6 L4 2 13 11 9 7 5 3 1 12 10

13 T 13 12 11 10 9 8 7 6 5 � 3 2 1

B 12 11 10 9 8 7 6 5 14 3 2 1 13[ ii 9 7 5 3 1 12 10 8 6 4 2 13

Fig. 1 . Test plan for analysis of the influence of modifying fac-tors on color changeAs an example. the diagram should be read as follows: The square in fourthrow, column three, means that at time (T) No. 2 laboratory technician No. 3has to read urine No. 4 with a dipstick from batch (B) No. 5 in environment (E)No. 6

.-�-. . . -�-- ++++++ 2.5 �NORMAL

Us - 2.5 �DO(JBT�

,Uc - ‘ . � ..+++.+ �.5 �FALS{ POS.�

UD - . + + . ++,++ 1.5 �FALSE lEG.’ I

0 1 2 3 � 5 6 7 8 9MMOL/L

Fig. 2. Matrix for interpretation of qualitative results into apoint of change, indicating the type of change, i.e. “normal,”“doubtful,” “false-positive,” and “false-negative.” UA, UB,lJ�, and UD indicate four different urines

TABLE 1.

THE TEST RESULTS FROM THE TWO SQUARE EXPERIMEtITS.

CLINISTIX: A. S-GLUKOTEST: B.

THE FIGURES IUDICATE THE POINTS OF CHANGE (MMOL GLUCOSE PER L)

READ ACCORDUIG TO FIG. 2.

LAB.TECH.

URINE 1 2 3 � 5 6 7 8 9 10 11 12 13

1 A 2.8 2.8 2.8 2.8 0.8 2.8 2.8 2.8 0.8 2.8 1.8 2.8 1.8

B 1,8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 0.8 1.8 0.8

2 A 2.0 6.0 3.0 4.0 2.0 �.O �.O �.O 3.0 5.0 3.0 5.0 2.0B 2.0 2.0 2.0 2.0 1.0 1.0 3.0 2.0 2.0 2.0 3.0 3.0 3.0

3 A �.O 5.0 2.0 �.0 2.0 �4.O 5.0 �.O 2.0 5.0 4.0 5.0 2.0

B 1.0 2.0 2.0 3.0 2.0 2.0 2.0 2.0 1.0 2.0 3.0 3.0 3.0

14 A L�9 3.9 2.9 0.9 2.9 2.9 2.9 3.9 2.9 2.9 1.9 3.9 1.9

B 1.9 1.9 0.9 1.9 1.9 0.9 0.9 1.9 1.9 0.9 1.9 2.9 1.9

S A 3.5 �.5 L4�5 45 2.5 3.5 5.5 �.5 2.5 �.5 2.5 6.5 L�5

� 3.5 3.5 3.� �.5 3.5 3.5 L4,5 3.5 �.5 3.5 �.5 3.5 4.5

6 A 0.7 2.7 1.7 3.7 1.7 1.7 2.7 1.7 3.7 2.7 1.7 2.7 1.7B 1.7 1.7 0.7 1.7 0.7 1.7 0.7 1.7 1.7 1.7 1.7 0.7 1.7

7 A 2.5 3.5 L4,5 3,5 2.5 L4�5 4,5 3,5 1.5 3.5 2.5 3.5 2.5B 2.5 2.5 1.5 2.5 2.5 2.5 2.5 1.5 2.5 2.5 2.5 2.5 2.5

8 A 1.7 1.7 1.7 1.7 0.7 1.7 1.7 1.7 0.7 1.7 1.7 1.7 1.7

B 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7 0.7

9 A 2.8 3.8 3.8 1.8 0.8 2.8 �.8 2.8 3.8 3.8 3.8 1.8 0.8

E 1.8 1.8 0.8 1.8 0.8 O.8� 0.8 0.8 1.8 0.8 1.8 0.8 1.8

10 A 3.1 7.1 3.1 8.1 3.1 5.1 5.1 4.1 �.1 4.1 2.1 4.1 2.1

B 2.1 2.1 1.1 2.1 1.1 2.1 2.1 2.1 2.1 1.1 2.1 2.1 2.1

11 A 2.9 1.9 i.9 i.g 0.9 3.9 L9 1.9 0.9 1.9 0.9 2.3 0.9

B 0.9 0.9 0.9 0,9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9 0.9

12 A 1.5 2.5 1.5 2.5 1.5 2.5 3.5 2.5 2.5 3.5 1.5 0.5 1.5B 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5

13 A 8.8 8.8 5.8 5.8 5.8 5.8 6.8 6.8 6.8 7.8 5.8 2.8 7.8U 2.8 2.8 2.8 3.8 1.8 2.8 1.8 2.8 3.8 1.8 3.8 2.8 2.8

The urine factor is represented by 13 night unines

selected at random from nondiabetic inpatients.

First, the actual glucose content of the urine was de-

termined with an LKB 8600 by use of a hexokinase

reagent kit from Boehringer Mannheim (cat. No.

15931). The concentration range was found to be

0-0.6 mmol/liter.

Next, glucose was added to the urine samples,

creating 10 different concentrations of the same urine

sample, ranging from 0-10 mmol/liter, with intervals

of 1 mmol/liter. The samples were then tested in man-

dom order.

The laboratory technician factor is represented by

13 laboratory technicians with various degrees of ex-

penience in reading Clinistix, but who were not famil-

iar with S-Glukotest.

The time factor indicates the order in which the

urine samples were tested by each laboratory techni-

cian.

The environmental factor is represented by 13 dif-

fement locations with various combinations of day-

light, neon light, and incandescent lamps.

The batch factor is represented by 13 different

batch numbers.

The calculations were based on the average con-

centrations of the last negative and the following pos-

itive reading (Figure 2). In the case of dubious read-

ings, the procedure was as described in Figure 2.

The test was performed twice, first with Clinistix

and then with 5-Glukotest, under identical condi-

tions. Between the tests the urine samples were kept

frozen at -60 #{176}C.

Source ofvariation

Table 2. Analysis of Variance Table for the Square Experiment with Two DipsticksSum of squares of Degrees of

the deviations freedom

A. Clinistix,

F PVariance

readafter lOs

Urine 4.8251 12 0.40209 18.59 <0.0005

Lab.technician 1.6678 12 0.13899 6.42 <0.0005Environment 0.6205 12 0.05171 2.39 0.005 < P < 0.01

Batch 0.3320 12 0.02767 1.28 0.1 < P < 0.3

Time 0.2573 12 0.02144 0.99 0.3 < P < 0.5Remainder 2.3365 108 0.02163

Total 10.0392 168

Urine 9.9036

B. S-G/ukotest, read after 30 s

12 0.82530 71.82 <0.0005

Lab.technician 0.5644 12 0.04703 4.09 <0.0005Environment 0.2754 12 0.02295 2.00 0.025 < P < 0.05

Batch 0.2351 12 0.01959 1.70 0.05 < P < 0.1Time 0.1853 12 0.01544 1.34 0.1 < P < 0.3

Remainder 1.2411 108 0.01149

Total 12.4049 168

1�O

0.9

08

0,7

0.6

0.5

a)>

In00.

0

C

0U(Ii

U-

The influence of the factors could now be analyzed,

the median point of color change calculated, and the

results from the two squares compared by a Wilcoxon

test, assuming gaussian-distributed rank sums.

Results

Results of the Variance Analyses

The log-normal distribution was used for the van-

ance analysis of the two test squares. The test results

appear in Table 1. Table 2 (a and b) shows the me-

sults of the F-test. Evidently, not only the urine fac-

tom but also the laboratory technician and environ-

mental factors contribute significantly to the vania-

tion in the point of color change. Clinistix read after

10 5 and S-Glukotest after 30 s show a median point

of color change of 2.8 and 1.8 mmol/litem, respective-

ly. The “95-90” tolerance interval for these points of

change was calculated to 0.5-0.8 mmol/liter for Clin-

istix and 0.5-4.5 mmol/liter for 5-Glukotest. Owing

to the discrete figure distribution, the tolerance in-

tervals were calculated according to the nonparame-

tric tolerance method (2).

Figure 3 shows the distribution of positive/nega-

tive answers from 2 X 1690 single results from the

test squares, summed up for each of the 10 concen-

tration intervals.

The points of change established by the two test

squares, compared by the Wilcoxon test, proved that

the sensitivity of Clinistix was significantly poorer

than that of 5-Glukotest. The rank sum for Clinistix:

36 159 (P << 0.001). Consequently, we extended the

time for reading the Clinistix to 30 s during the suc-

ceeding tests, thereby increasing the sensitivity.

As already mentioned, we have arrived at the re-

sults presented so far by “translating” the positive/

negative answers to a urine glucose concentration.

For this we used the matrix shown in Figure 2. Dubi-

04

0.3

02

0.l.J-

_��0 1 2 3 4 5 6 7 8 9 10

Urine glucoseconcentration (mmol It)

Fig. 3. Distribution of positive results from the 1690 individualresults of the two test squares, summed for each of the 10concentration intervalsCIlnistix (-) were read after 10 s, S-Glukotest (- - -) after 30 s

ous results were obtained only with Clinistix. Among

each of the 169 points of change there were 24

“doubtful” cases, four “false-positive” cases, and

seven “false-negative” cases. All the results obtained

by 5-Glukotest were “normal” (Figure 2). This dif-

ference in occurrence of dubious answers is statisti-

cally significant (P < 0.01) according to the binomial

distribution.


I..-’

I,J

r�

aQ5.�

C

.2 Q4U

U- 0,3

Q2

Qi

0.00�1 � �3 4 � 6

Urine gLucoseconcentration(mmol /1)

Fig. 4. Distribution of positive results from 77 night urines en-riched with glucose (1 mmol/liter intervals), Clinistix (-) andS-Glukotest (- - -), read after 30 s

‘N

>

N,

C

U’N,C

Fig. 5. Distribution of glucose concentrations in urines re-ceived in the laboratory from the hospital departments for glu-cose screening during one week (n 343)

‘, 0,2 0.4 Q6 Q8 1,0 1,5 20 �0 L�O �0 �0 10 15 20 30 50 80 120 170 350

Urine giucoseconcentration (mmol/I)


Level and Ability of Discriminationunder Favorable Analytical Conditions

As 13 randomly selected urine samples were not

considered representative of all the urine samples

tested in a clinical chemical laboratory, and as it was

desirable to determine more accurately the discrimi-

nation level and ability of the dipsticks, 77 urine

samples were enriched with glucose and tested under

the most favorable conditions, i.e., by skilled labora-

tory technicians in the best possible light (neon light

and incandescent lamps). For this part of the test,

dipsticks of the same batch number were used, and

the urine samples were tested before the other analy-

ses that day. The urine samples were collected from

nondiabetic inpatients chosen without conscious

bias, and the glucose content was determined by the

hexokinase kit method. Their glucose concentrations

ranged from 0 to 0.9 mmol/liter. The urine samples

were enriched with glucose so that six concentrations

were prepared from each sample, ranging from 0 to 6

mmol/liter with intervals of 1 mmol/liter, and tested

in random order.

As stated above, the point of color change occurred

at a considerably higher glucose concentration for

Clinistix than for 5-Glukotest, but as this difference

was lessened by reading the Clinistix after 30 s-and

not after 10 s as recommended by Ames-both dip-

sticks were read after 30 s during this experiment.

Figure 4 shows the distribution of positive/nega-

tive results from 77 night urines enriched with glu-

cose at 1 mmol/liter intervals; both Clinistix and 5-

Glukotest were read after 30 s.

A comparison of Figures 3 and 4 shows the im-

proved sensitivity when Clinistix is not read until

after 30 S.

Interpretation of a Dipstick Result

We used the hexokinase method to measure glu-

cose in urine samples from inpatients, received in the

laboratory for glucose screening during one week.

By combining the information about a probable

positive/negative dipstick result in the different con-

centration intervals with the probability of finding

urines in the same intervals, we are able to express

the value of the information obtained from a dipstick

examination, i.e., to predict the probable glucose con-

tent of the urine sample. The expression is given by

Bayes’ formula (3):

P(C, � C < C+1�neg. statement)

P(neg. statementlC1 � C < C��1) X P(C1 � C < C1�1)

� P(neg. statementlC1 < C < C��1) X P(C1 � C < C�#{247}1)

where Cj = 0, 1, 2, 3 , n mmol glucose/liter,

and C�+i = 1, 2, 3, 4 , (n + 1) mmol glucose/liter.

Figure 5 shows the distribution of the glucose con-

centrations of unines received in the laboratory dur-

ing one week, as measured by the hexokinase reac-

tion.

Table 3 shows the results when Bayes’ formula was

used to calculate the probability of a urinary glucose

value being within certain concentration limits when

a negative dipstick reading has been obtained.

The above may picture the capacity of the dipstick

method too optimistically because of the relatively

large number of normal urines used for the calcula-

tion. Thus we have included the same figures in a

table of sensitivity and specificity (Table 4).

P(Cj �C<C1+

1 negativestatement

S-GIu kotest b

0.957

0.041

0.003

0.0001

0

. ClinistixNegative5 Giukotest

ClinistixPositive S-Glukotest

82.5 17.5


Table 3. The Probability of Finding the UrineGlucose Content within a Certain ConcentrationInterval when the Dipstick Reading is Negativea

P(cj � C < Cj +

1 negativestatement

Clinistix b

0.953

0.043

0.004

0.0004

0

Ci � C < Cj + 1(mmol/Iiter)

o��;c< 1

1 :�;C< 2

2�C<3

3�C<4

4�C

a For meaning of symbols, see textb Both dipsticks read after 30 s

Table 4. Outline of Specificities at a SelectedDiscrimination Level of 2 mmol/litera

<2mmol >2mmol

79.8 0.479.6 0.2

2.7 17.1

2.9 17.3

17.1 x 100 �

Sensitivity {S-Glukotest 17.3 x 100 � = 98.9%Cl inistix. . . 17.5

17.5

Clinistix 79.8 x 100 � = 96.7%

SPecificitYf 82.582.579.6 x 100 � = 96.5%

17.1 x 100 %=86.4%Clinistix

Predictive { Gluk 19.8value 17.3x 100 %=85.6%

- otest 20.2

a Figures are percent of total

Discussion

The F-test shows that much of the variation in test

readings for both types of dipsticks is attributable to

inter-sample variation. The lowest positive urine and

the highest negative urine had glucose concentrations

of 0.3 and 9.3 mmol/liter, respectively.

Like other authors (4), we experienced a considera-

bly smaller sensitivity when testing glucose in urine

than when experimenting with glucose in water,

probably because urine contains a number of com-

pounds that tend to inhibit the analytical reactions

applied in the experiment (4, 5), particularly certain

drugs (4). We have not studied the effect of drugs

further, as we judge it unwarranted and find it sim-

pler to determine the glucose content of drug-con-

taming urines by quantitative methods.

Unlike the urine factor, the other factors tested

can be modified. Among these factors we experienced

a significant difference between the laboratory tech-

nicians, which illustrates the need for very careful in-

struction before even as simple a test as the dipstick

test is done.

The environment factor has a weak but significant

influence on the dipstick readings; it is necessary to

do the tests under good and consistent lighting.

However, we found no significant difference be-

tween results for various batches of dipsticks or be-

tween one particular laboratory technician at differ-

ent times. We conclude that the two brands of dip-

sticks are prepared from reasonably consistent com-

pounds and that one laboratory technician can justi-

fiably do analyses in series of up to 130.

However, the similarity between two dipsticks

ends here, as illustrated in Figure 3 and by the Wil-

coxon test. We conclude that when the two are used

according to the manufacturers’ instructions, S-Glu-

kotest is more sensitive and gives a more consistent

interpretation.

Values in the literature vary greatly on the subject

�; of the physiological limits of glucose in urine. For ex-

80.2 ample, Renschler et al. (6) fix the upper limit at 1.6779.8 mmol/liter, and Pileggi and Szustkiewicz (7) at 0.55

19.8 mmol/liter. The point of change obtained by 5-Glu-

20. 2 kotest is closer to these values than those obtained by

100.0 Clinistix. By increasing the time before reading theClinistix to 30 s, we improved its sensitivity to com-

pare with that of 5-Glukotest, both now showing an

average point of change between 1 and 2 mmol/liter

and a range between 1 and 4 mmol/liter (Figure 4).

Another reason for preferring a sensitive method of

excluding umines that are normal as far as glucose

concentration is concerned is that we have a quanti-

tative method of differentiating all positive tests and

thus establishing “false positive” results.

Table 3 states the probable glucose concentration

in any particular urine sample when the dipstick test

is negative. Note that the calculation has been based

on the distribution of urines in Figure 5, which is

based on local data and not necessarily universally

representative. When a result is negative there is a

greater than 95% probability that the glucose concen-

tration is less than 1 mmol/liter and more than 99%probability that it is less than 2 mmol/liter.

The above information seems to justify use of the

dipsticks for screening. One must bear in mind, how-

ever, that of urine samples with a glucose content of 2

to 3 mmol/liter, Clinistix may give a “negative” an-

swer for 15% and 5-Glukotest for 9%. For urines with

a glucose content between 3 and 4 mmol/liter, 4% and

2%, respectively, may be so overlooked (Figure 4).

This is true even if the sensitivity of Clinistix is im-

proved by reading after 30 s.

On the other hand, the predictive value of the tests

is found to be 86%; i.e., 14% of subjects with a “posi-

tive” result will have a urinary glucose concentration

of less than 2 mmol/liter (Table 4). Suppose that all

untreated diabetics have a urinary glucose concentra-

tion above and all nondiabetics a concentration

below this value-which of course is not quite true-

14 of 100 patients suspected to have diabetes mellitus

on the basis of a dipstick examination will in fact be

healthy.


Generally, we should choose the type of analysis

with the highest level of discrimination between well

and sick in relation to the consequences of a positive/

negative result, and with the greatest ability to dis-

criminate; i.e., with the sharpest change at the appro-

priate decision value.

Of course, it would be better specifically to mea-

sure the glucose in all urine samples, thereby avoid-

ing “false-positive” and “false-negative” results.

However, there may be conditions under which only

the simple screening test can be done.

One method for increasing the certainty of the de-

cision in the case of a doubtful result is to increase

the urine glucose content of the specimen by 2 mmol/

liter and . observe whether the doubtful reaction,

checked at 30 s, now becomes clearly positive. If it is

not, the reason may be that inhibitory substances are

present, and the glucose must then be checked by

other methods.

In any case the consequences of a “false-positive”

or “false-negative” result must be considered before

choosing the decision value. For example, when a

urine sample is tested for glucose, the consequence of

a “false-positive” result will usually be only economi-

cally detrimental (the cost of a quantitative determi-

nation of glucose in urine), plus the fact that the pa-

tient may be inconvenienced by an oral glucose-toler-

ance test. It is quite a different matter with some

other qualitative analyses. For instance, to refute a

positive result of a dipstick test for protein in urine,

one must do several quantitative protein analyses

and possibly expose the patient to extensive and

trying tests (urography, for example).

When dipsticks are used to screen urines for glu-

cose, there should be quality control of the test sys-

tem. Because of the great influence of urine composi-

tion on dipstick results, we suggest that each day a

random series of urines be tested, prepared from a

pool of negative urines enriched with glucose to con-

centration intervals of 1 mmol/liter. When changing

to a new batch of dipsticks, results from this batch

should be compared with those for old batch by com-

paring the fraction of positive results obtained when

analyzing 20 times at random a control urine with a

glucose content corresponding to 0.1 and 0.9 of posi-

tive results, respectively.

On the basis of our results we cannot recommend

the relatively crude dipstick examination for glucose

in urine for quantitative purposes. Differentiating

subtle graduations in color intensity at a fixed time

will be susceptible to even greater errors than those

observed in the “yes or no” situation. As a matter of

fact such an attempt was made by the technicians

during the study; the results were such that any sta-

tistical analysis was impossible. Quick and reliable

methods for measuring glucose concentration in

urine are at hand. Clinical decisions that imply thera-

peutic consequences (e.g., alterations in insulin dos-

age) should be based on information having a validity

that justifies these decisions.

References

1. Cox, D. R., Planning of Experiments. Wiley, New York, N.Y.,

1958.

2. Reed, A. H., Henry, R. J., and Mason, W. B., Influence of statis-

tical method used on the resulting estimate of normal range. Clin.Chem. 17, 276 (1971).

3. Wonnacott, T. H., and Wonnacott, R. J., Introductory Statis-

tics for Business and Economics. Wiley, New York, N.Y., 1972, p

363.

4. Jaksch, W., and Tschepper, P., Erfahrungen mit Schnelltests

zur Harnuntersuchung beim Hund. Wien. Tieraerztl. Monatschr.61, 83 (1974).

5. Blaedel, W. J., and UhI, J. M., Nature of materials in serumthat interfere in the glucose oxidase-peroxidase-o -dianisidine

method for glucose, and their mode of action. Clin. Chem. 21, 119

(1975).

6. Renschler, H. E., Weicker, H., and Baeyer, H. von, Die obere

Normgrenze der Glukosekonzentration im Urin Gesunder. Dtsch.Med. Wochenschr. 90, 2349 (1965).

7. Pileggi, V. J., and Szustkiewicz, C. P., Carbohydrates. In Clini-cal Chemistry: Principles and Techniques, 2nd ed., R. J. Henry,

D. C. Cannon, and J. W. Winkelman, Eds. Harper and Row, New

York, N.Y., 1974, p 1295.

evaluation of a dipstick test for glucose in urine

Documents