The tuberculin skin test: a comparison of ruler and calliper readings

INT J TUBERC LUNG DIS 14(10):1266–1271© 2010 The Union

The tuberculin skin test: a comparison of ruler and calliper readings

H. Geldenhuys,*† S. Verver,‡§ S. Surtie,*† M. Hatherill,*† F. van Leth,‡§ F. Kafaar,*† M. Tameris,*† W. Kleynhans,*† K. K. Luabeya,*† S. Moyo,*† W. Sikhondze,*† W. Hanekom,*† H. Mahomed*†

* South African Tuberculosis Vaccine Initiative, University of Cape Town, † School of Child and Adolescent Health, University of Cape Town, Cape Town, South Africa; ‡ KNCV Tuberculosis Foundation, Amsterdam, § Centre for Infection and Immunity, Academic Medical Centre Amsterdam, University of Amsterdam, Amsterdam, The Netherlands

Correspondence to: Hennie Geldenhuys, South African Tuberculosis Vaccine Initiative, Brewelskloof Hospital, Haarlem Street, Worcester, South Africa 6850. Tel: (+27) 23 346 5400. Fax: (+27) 23 626 1822. e-mail: [email protected] submitted 19 February 2010. Final version accepted 30 March 2010.

B A C K G R O U N D : The Mantoux tuberculin skin test

(TST) is widely used to diagnose latent infection with

Mycobacterium tuberculosis. TST skin induration may

be measured either by a transparent ruler or by a pair of

callipers. We hypothesised that the type of instrument

used may affect the reading.

O B J E C T I V E : To determine whether variability in Man-

toux TST measurement is affected by the type of read-

ing instrument.

M E T H O D : A TST (Mantoux method) was performed

among healthy adolescents. The indurations were read

with among ruler and calliper by two independent read-

ers. Limits of agreement and Kappa (κ) scores at TST

positivity cut-off points were calculated. A Bland-Altman

graph was constructed.

R E S U LT S : The 95% limits of agreement between in-

struments ranged from −5 mm to 3 mm. The limits of

agreement between readers ranged from −5 mm to

4 mm. κ scores between instruments were respectively

0.7 and 0.8 at 15 mm and 10 mm cut-offs.

C O N C L U S I O N : The variability between readers of TST

indurations is not infl uenced by changing the reading

instrument.

K E Y W O R D S : tuberculosis diagnosis; tuberculin skin test; Mantoux

THE TUBERCULIN SKIN TEST (TST) is widely used to diagnose latent infection with Mycobacte-rium tuberculosis. The test may also help in the diag-nosis of active tuberculosis (TB).1–3 In the Mantoux TST, a purifi ed protein derivative (PPD) of M. tuber-culosis is injected intradermally and the resulting skin induration is measured after 48–96 h. This measure-ment can be performed using either a transparent ruler or a calliper.1–4 A result above a certain cut-off, usually 10 mm or 15 mm, is considered positive.

We hypothesised that the type of instrument used to measure the induration may infl uence the inter-reader variability in reading the result. This may be particularly important in surveys or research settings where data are aggregated, resulting in amplifi cation of small differences. With the possibility of system-atic biases in reading TST indurations, estimates of the burden of latent TB infection within a population may be prone to error.

In practice, the ruler is the reading instrument most commonly used. It is prone to digit preference because the reader can see the scale of the ruler while measuring the induration. The diffi culty in aligning the

scale of the ruler with the often unclear boundaries of the induration may also be problematic. Swelling or tenderness around the induration may prevent the ruler from being placed fl at on the surface of the skin. The line divisions on the ruler may not be clearly visi-ble against the backdrop of the skin, particularly in dark-skinned individuals.

The use of callipers may help prevent these prob-lems. However, because the calliper is a mechanical device with moving parts, it is technically more diffi -cult to use, and may therefore also give rise to intra- and inter-reader variation. In addition, callipers are more expensive (approximately US$30), more prone to breaking and are less readily available, especially in resource-limited settings.

Although inter- and intra-reader variability with TST has been extensively reported,5–10 the effect of the type of TST reading instrument on results has not been evaluated, and guidelines do not defi nitively rec-ommend one instrument over the other.1–3 If the equivalence of the ruler and calliper reading instru-ments, or lack thereof, can be demonstrated, it would have important implications for the cost, method and

S U M M A R Y

Comparison of TST readings using ruler and calliper 1267

interpretation of TST reading. Our objective was there-fore to determine whether variability in TST measure-ments is affected by the type of reading instrument.

METHOD

ParticipantsA cross-sectional study design was used to compare TST readings taken with the ruler and readings taken with the calliper. Healthy adolescents aged between 12 and 18 years were recruited from a single high school in an area with high TB incidence. This popu-lation was used because previous experience suggested that more indurations can be expected in this group than in younger children.

TST administrationA TST (Mantoux method) was administered to par-ticipants. Participants were excluded if they had a his-tory of severe reaction to a previous TST, were medi-cally ill on the day of TST administration or knew they would not be available for both the administra-tion and reading study sessions. Two tuberculin units (TU) of PPD RT23 (Statens Serum Institut, Copen-hagen, Denmark) in a volume of 0.1 ml was injected intradermally into the volar surface of the forearm according to the Mantoux technique. Administration was performed by one of 10 dedicated research nurses who were trained and experienced in the technique.

TST readingThe maximum transverse diameter of TST induration was read 72 h after administration. Participants’ TST indurations were initially visually inspected by a sin-gle reader independent of the main study readers. This reader determined whether the induration was greater than zero. Participants with an induration greater than zero were included for a formal reading by each of two study readers. As a quality control measure, sequential participants with an induration initially assessed as zero were also included up to 15% of the fi nal study sample size.

Data were collected in two rounds at a single sit-ting. Each participant’s TST induration was measured twice by each reader, once with a ruler and once by a calliper (Table 1). The readers were trained and expe-rienced in both reading techniques. The same manu-

facturer and model of a transparent, fl exible ruler and dial vernier calliper were used by both readers. To delineate the induration, palpation only was used instead of marking with a pen. Results were read to a person who recorded results. Each reader was blinded to previous readings by herself or by the other reader. Participants were randomised prior to each round of reading for the order in which they were read and which reader read their induration fi rst. Readings were recorded for each participant per reader and per instrument used.

Participant care and ethicsEthical approval for this study was given by the Uni-versity of Cape Town Faculty of Health Science’s Re-search Ethics Committee. Parental consent and par-ticipant assent was obtained. After-care advice and treatment of adverse effects were provided if neces-sary. Participants with an induration of ⩾15 mm were referred by a medical offi cer to a TB clinic for exclusion of active TB.

Data analysisIn the absence of sample size calculation methods for a Bland-Altman11 type analysis, we estimated that we would need to enrol 300 participants to obtain an adequate sample size of 100 who had indurations greater than zero. As the primary analysis was to as-sess agreement between two reading methods and not differences, a formal sample size calculation is not useful for this analysis. For the additional analysis re-garding the difference in mean induration between the two methods, we could only perform a post hoc assessment of the minimal detectable statistical dif-ference. Based on the study results, this was 2.5 mm in a paired assessment, at a signifi cance level of 0.05 and a power of 80%.

Data were entered into a Microsoft Access data-base (version 2003, MicroSoft, Redmond, WA, USA). Readings were rounded off to the nearest 1 mm. Dif-ferences between instruments and readers were anal-ysed by calculating limits of agreement, which is equiv-alent to the 95% distribution as described by Bland and Altman.11 A Bland-Altman graph11 was plotted with the difference between ruler and calliper readings on the y-axis, and the average ruler-calliper reading on the x-axis. To determine agreement between ruler and calli per readings at the conventional clinical cut-off values of 10 mm and 15 mm for a positive result, the percentage disagreement and kappa (κ) score were calculated.

RESULTS

Ninety-seven participants underwent the full set of comparative readings (Figure 1). The median age of the participants was 15 years, with an interquartile range (IQR) of 1 year.

Table 1 Summary of reading of TST indurations by reader and round of reading*

Ruler Calliper

Reader 1 Round 1 Round 2Reader 2 Round 2 Round 1

* Each participant’s TST induration was read twice by each reader in two rounds. In the fi rst round Reader 1 read by ruler and Reader 2 by calliper. In round 2, the readers switched instruments so that Reader 1 read by calliper and Reader 2 by ruler.TST = tuberculin skin test.

1268 The International Journal of Tuberculosis and Lung Disease

Distribution of readings and digit preferenceOf the 97 participants, 69 (71%) had a mean TST in-duration (of all four readings) of >10 mm, and 33 (34%) >15 mm. Most indurations above zero were between 10 and 20 mm (median 13, IQR 7 mm). Us-ing unit intervals of 1 mm, there was no trend for digit preference for either reader with either instru-ment (Figure 2).

Ruler vs. calliperRuler and calliper measurements differed by ⩾3 mm in at least 18% of participants. In 64 (66%) of cases,

the calliper reading was larger than the ruler reading. Of the 17 indurations initially assessed as zero, ruler and calliper agreed on all, except in one case where a single ruler reading was 5 mm.

The mean difference between measurements was −1 mm (ruler minus calliper) for each of the readers. The limits of agreement for the difference were −5.7 to 3.7 mm for Reader 1, and −5.6 to 3.5 mm for Reader 2 (Table 2). The Bland-Altman graph showed that the differences between the ruler and calliper do not follow a trend in relation to the size of the indu-ration (Figure 3).

For a cut-off point of >15 mm TST positivity, the κ score for agreement between ruler and calliper for each of the two readers was 0.7 (good) and at >10 mm the score was 0.8 (very good).The percentage disagree-ment for TST positivity remained similar for each of the separate readers at the 15 mm cut-off point (Reader 1 15%, Reader 2 12%). At the 10 mm cut-off, the per cent disagreement for separate readers differed marginally (Reader 1 4%, Reader 2 8%; Table 3). At

Figure 1 Flowchart of study design. TST = tuberculin skin test.

Figure 2 Distribution of TST induration measurements in interval units of 1 mm to determine digital preference. There is no pattern of distribution peaks evident at signifi cant terminal digit cut-offs. Readings are reported separately per reader and instrument. TST = tuberculin skin test.

Table 2 Mean difference between TST induration measurements taken by the ruler and calliper and 95% limits of agreement (N = 97)*

Mean difference between ruler and calliper

mm (95%CI)

95% upper level of agreement mm (95%CI)

95% lower level of agreement mm (95%CI)

Reader 1 −1.0 (−1.5–−0.5) 3.7 (3.5–3.9) −5.7 (−5.9–−5.6)Reader 2 −1.0 (−1.6 –−0.6) 3.5 (2.7–4.3) −5.6 (−6.4–−4.8)

* Confi dence intervals around the levels of agreement are reported as de-scribed by Bland and Altman.11 Results are reported separately for different readers.TST = tuberculin skin test; CI = confi dence interval.

Comparison of TST readings using ruler and calliper 1269

both cut-off points, more participants were desig-nated as positive using the calliper measurement than with the ruler measurement (at 15 mm 1% more by Reader 1, 4% more by Reader 2; at 10 mm 4% more by Reader 1 and 8% more by Reader 2).

Inter-reader variation with the same instrumentReadings on the same participant taken by the two readers differed by >3 mm in 15 (15%) participants for ruler readings and 20 (20%) for calliper readings. The mean difference in readings between the two readers (inter-reader variability) was 0.1 mm for both ruler and calliper readings. The limits of agreement for the mean difference between readers ranged be-tween −5.0 mm and 5.2 mm (Table 4).

DISCUSSION

In this study, we assessed agreement in TST readings between readers and instruments using both indica-tors of precision (difference in means, κ scores at cut-off points) and variability (95% limits of agreement). The mean difference in readings using different in-struments was small, and the κ scores at clinical cut-offs were good. Even so, at both the 10 mm and 15 mm clinical cut-off points for TST positivity, the calliper designated more participants as positive (up to 8% more). This difference would have important implications in a survey setting. Moreover, the limits of agreement between readers with the same instru-ment and between instruments used by the same reader were wide. This suggests considerable variability. Variability between instruments was similar for both readers and that between readers was similar for both instruments. The wide limits of agreement between readers imply that inter-reader variation is as large and equally relevant as inter-instrument variation.

Figure 3 Bland-Altman graph for TST indurations for Reader 1 and Reader 2 (N = 97). Ruler minus calliper measurements on the y-axis are plotted against the average of ruler and calliper measurements on the x-axis. The horizontal lines denote the mean difference and the upper and lower levels of agreement. TST = tuberculin skin test.

Table 3 Agreement of positivity classifi cation at 15 mm and 10 mm cut-off (N = 97)

Calliper >15 mm positive

n (%)

Calliper ⩽15 mm negative

n (%)

Reader 1, 15 mm cut-off Ruler >15 mm positive 27 (28) 7 (7) Ruler ⩽15 mm negative 8 (8) 55 (57)

Reader 2, 15 mm cut-off Ruler >15 mm positive 32 (33) 4 (4) Ruler ⩽15 mm negative 8 (8) 53 (55)

Reader 1, 10 mm cut-off Ruler >10 mm positive 67 (69) 0 Ruler ⩽10 mm negative 4 (4) 26 (27)

Reader 2, 10 mm cut-off Ruler >10 mm positive 60 (62) 0 Ruler ⩽10 mm negative 8 (8) 29 (30)

Table 4 Mean difference between TST induration measurements taken by Reader 1 and Reader 2 and 95% limits of agreement (N = 97)*

Mean difference mm (CI)†

95% upper level of agreement mm (CI)

95% lower level of agreement mm (CI)

Ruler 0.1 (−0.4–0.6) 4.4 (4.1–4.7) −4.2 (−4.4–−4.1)Calliper 0.1 (−0.4–0.6) 5.2 (4.9–5.4) −5.0 (−5.3–−4.8)

* The confi dence intervals around the levels of agreement are reported as described by Bland and Altman.11 Results are reported separately for differ-ent instruments.† Reader 1 minus Reader 2.TST = tuberculin skin test; CI = confi dence interval.

1270 The International Journal of Tuberculosis and Lung Disease

This variability in TST reading is not unlike that found by other researchers, and may be inherent to the TST. Studies using a single participant with a known induration size and many readers,6,9 as well as studies with a number of participants and readers using a standard technique and instrument,5,7,10 have shown variation in TST readings. The extent of the variation has differed, and is complicated by different methods of reporting agreement. A study that used a similar Bland-Altman type sub-analysis in addition to other correlation coeffi cients, for example, demon-strated inter-reader limits of agreement of −2.7 to 3 mm.5 This is slightly narrower than our fi ndings, but has similar clinical implications.

Although other studies have explored the reliabil-ity of TST reading by determining inter- and intra-o bserver variation and repeatability, these studies have used a single instrument modality during reading.5–7,9 No direct comparisons have been made between dif-ferent reading instruments. One study found good re-liability for readings taken among adults by a sliding ruler similar to a calliper, but did not compare this device with the conventional ruler.7 Most other stud-ies used the ruler alone,6–8 fewer the calliper alone,10 and where the calliper and ruler were used together,5 a comparison of instruments was not the main focus of the analysis. The question of consistent bias or in-accuracies due to the instrument used, independently of other intra-reader and inter-reader variation, has thus remained unresolved up until now.

Some researchers have suggested that using the ball-point pen technique to determine the boundaries of the induration is more accurate than the palpation technique.5–7 However, this advantage may be more evident among inexperienced readers only.6 Others have found comparable variation using either tech-nique.8 In our study, both readers were experienced, and we preferred to use the palpation technique for two reasons: pen marks on the skin could not be re-moved easily and would lead to bias for the second and subsequent readings; and repeatedly marking the skin could cause irritation and discomfort and make the reading more diffi cult and more prone to error.

Limitations to this study include that the study de-sign did not allow for the assessment of intra-reader variation per instrument because indurations were not read more than once by each reader with a particular

instrument. The sample size chosen was approximated and possibly inadequate.

CONCLUSION

Despite a theoretical advantage to using a calliper in-stead of a ruler, the variability associated with the TST is not affected by the reading instrument. This implies that either can be used. In deciding which in-strument to use, other factors need to be considered such as training and the confi dence of the reader in a particular instrument, and its availability and cost.

AcknowledgementsThe authors thank the school board, headmaster, staff, and learn-ers of Breërivier High School for their invaluable support. This study was funded independently by the South African Tuberculosis Vaccine Initiative (SATVI).

References 1 American Thoracic Society. Diagnostic standards and classifi -

cation of tuberculosis in adults and children. Am J Respir Crit Care Med 2000; 161: 1376–1395.

2 British Thoracic Society Joint Tuberculosis Committee. Con-trol and prevention of tuberculosis in the United Kingdom: code of practice 2000. Thorax 2000; 55: 887–901.

3 Arnadottir Th, Rieder H L, Trébucq A, Waaler H T. Guidelines for conducting tuberculin skin test surveys in high prevalence countries. Tubercle Lung Dis 1996; 77 (Suppl 1): S1–S19.

4 Huebner R E, Schein M F, Bass J B. The tuberculin skin test. Clin Infect Dis 1998; 113: 1175–1177.

5 Pouchot J, Grasland A, Collet C, Coste J, Esdaile J M, Vince-neux P. Reliability of tuberculin skin test measurement. Ann Intern Med 1997; 126: 210–214.

6 Carter E R, Lee C M. Interpretation of the tuberculin skin test reaction by pediatric providers. Pediatr Infect Dis J 2002; 21: 200–203.

7 Perez-Stable E J, Slutkin G. A demonstration of lack of vari-ability among six tuberculin skin test readers. Am J Public Health 1985; 75: 1341–1343.

8 Longfi eld J N, Marsileth A M, Golden S M, Lazoritz S, Bohan S, Cruess D F. Interobserver and method variability in tubercu-lin skin testing. Pediatr Infect Dis 1984; 3: 323–326.

9 Kendig E L, Kirkpatrick B V, Carter W H, Hill F A, Caldwell K, Entwistle M. Underreading of the tuberculin skin test reaction. Chest 1998; 113: 1175–1177.

10 Singh M D, Sutton C, Woodcock A. Tuberculin test measure-ment variability due to the time of reading. Chest 2002; 122: 1299–1301.

11 Bland J M, Altman D G. Statistical methods for assessing agree-ment between two methods of clinical measurement. Lancet 1986; 1: 307–310.

C O N T E X T E : Le test tuberculinique cutané de Mantoux

(TST) est largement utilisé pour le diagnostic de l’infec-

tion latente par Mycobacterium tuberculosis. L’indura-

tion cutanée du TST peut être mesurée soit par une règle

transparente soit par un compas. Nous avons émis l’hypo-

thèse que le type d’instrument utilisé peut infl uencer la

lecture.

O B J E C T I F : Déterminer dans quelle mesure les variations

des mesures du TST de Mantoux sont affectées par le

type d’instrument de lecture.

M É T H O D E : On a réalisé un TST par la méthode de Man-

toux chez des adolescents sains. Les indurations ont été

lues à la fois par une règle et un compas, par deux lec-

teurs indépendants. Les limites de concordance et les

R É S U M É

Comparison of TST readings using ruler and calliper 1271

scores kappa (κ) ont été calculés aux valeurs limites de

positivité du TST et un graphique de Bland-Altman a

été élaboré.

R É S U LTAT S : Les limites à 95% des concordances entre

instruments ont été de −1 mm à 3 mm. Les limites de

concordance entre lecteurs ont varié entre −5 mm et

4 mm. Les scores κ entre instruments ont été de 0,7 et

0,8 aux limites respectives de positivité de 15 mm et de

10 mm.

C O N C L U S I O N : Le changement d’instrument de lecture

n’infl uence pas les variabilités des lectures d’induration

de la TST entre lecteurs.

R E S U M E N

M A R C O D E R E F E R E N C I A : La prueba cutánea de la tu-

berculina (TST) es de amplio uso en el diagnóstico de la

infección latente por Mycobacterium tuberculosis. La

induración de la reacción TST se puede medir ya sea

con una regla transparente o con un calibrador. Se plan-

tea la hipótesis de que el instrumento de medida usado

puede afectar la lectura.

O B J E T I V O S : Determinar si en la variabilidad de la

medición de la reacción de TST interviene el tipo del in-

strumento de lectura.

M E T H O D : Se practicó la TST (método Mantoux) en ado-

lescentes sanos. Dos lectores independientes midieron la

induración con una regla o con un calibrador. Se calcu-

laron los límites de concordancia y el coefi ciente κ con

cada valor discriminatorio. Se construyó un gráfi co de

Bland-Altman.

R E S U LTA D O S : Los límites de concordancia al 95% en-

tre los instrumentos oscilaron entre −5 mm y 3 mm.

Los límites de concordancia interlectores oscilaron entre

−5 mm y 4 mm. Los coefi cientes κ entre los instrumen-

tos fueron 0,7 con un umbral de 15 mm y 0,8 con un

umbral de 10 mm.

C O N C L U S I Ó N : El cambio de instrumento de medida no

mostró ninguna infl uencia sobre la variabilidad entre

los lectores de la reacción de induración a la TST.