Vous êtes sur la page 1sur 7

INT J TUBERC LUNG DIS 15(4):483488

2010 The Union


doi:10.5588/ijtld.10.0229

Evaluation of light emitting diode-based uorescence


microscopy for the detection of mycobacteria in
a tuberculosis-endemic region
S. Shenai,* J. Minion, V. Vadwai,* T. Tipnis,* S. Shetty,* A. Salvi,* Z. Udwadia,* M. Pai,
C. Rodrigues*
* Microbiology Department, Parmanand Deepchand Hinduja National Hospital and Medical Research Centre, Mumbai,
India; Department of Epidemiology and Biostatistics, and Respiratory Epidemiology and Clinical Research Unit,
McGill University, Montreal, Quebec, Department of Medical Microbiology and Immunology, University of Alberta,
Edmonton, Alberta, Canada
SUMMARY
OBJECTIVES:

To evaluate fluorescence microscopy (FM)


using light emitting diode (LED) technology for the detection of acid-fast bacilli at a tertiary referral centre in
Mumbai, India, a tuberculosis-endemic country.
D E S I G N : LED FM was introduced into a laboratory
experienced with Ziehl-Neelsen (ZN) microscopy but unfamiliar with FM. It was evaluated in parallel with routine ZN microscopy services and compared with mycobacterial culture as a reference standard.
R E S U LT S : A total of 1357 pulmonary and 917 extrapulmonary specimens were examined during the study.
LED FM had 78.3% sensitivity and 92.0% specificity
against mycobacterial culture when using pulmonary
specimens, and 34.0% sensitivity and 88.8% specificity

for extra-pulmonary specimens. The mean time per smear


examination was 2.48 min for ZN vs. 1.41 min for LED
FM. Several biases in study design and operation identified during analysis, which are likely to lead to underestimates of LED FM accuracy, are discussed in the context of future LED FM evaluations.
C O N C L U S I O N S : Although LED FM has significant benefits over both ZN microscopy and conventional FM, its
implementation and validation may be prone to difficulties which could hamper evaluation of its performance.
Adequate training and detailed standard operating procedures are important to maximise accuracy.
K E Y W O R D S : tuberculosis; microscopy; light emitting
diode; accuracy

FLUORESCENCE MICROSCOPY (FM) for the detection of tuberculosis (TB) has been in use since the
mid-1940s and is the method of choice for large laboratories in industrialised nations.1,2 A systematic review has shown that FM is on average 10% more
sensitive than conventional light microscopy, has comparable specificity and takes significantly less time to
read smears.3,4 The main barrier to the widespread
implementation of FM has traditionally been the cost
of fluorescent microscopes and their maintenance.5
Light emitting diode (LED) fluorescent microscopes
have been identified as an alternative to conventional
FM. Compared to mercury vapour fluorescent microscopes, LED microscopes are less expensive, have lower
maintenance requirements, require less power and are
able to run on batteries.57 LED bulbs have a very long
half-life, there is no ultraviolet light production and
they are reported to perform equally well without a
darkroom. These qualities make them feasible for use
in low-resource settings, and recent World Health
Organization (WHO) policy has recommended that

conventional FM be replaced by LED microscopy,


and that LED microscopy be phased in as an alternative to Ziehl-Neelsen (ZN) light microscopy.8
The present study was undertaken to evaluate the
performance of LED microscopy compared to ZN
light microscopy using mycobacterial culture as a reference standard.

MATERIALS AND METHODS


Samples
Consecutive patient specimens (pulmonary and extra-pulmonary) received for acid-fast bacilli (AFB)
smear and culture from January to June 2009 underwent routine ZN microscopy and LED FM at the
mycobacteriology laboratory of the Parmanand Deepchand Hinduja National Hospital and Medical Research Centre, Mumbai, India, a tertiary care TB
referral centre.
For each pulmonary TB case, three morning sputum specimens per patient were received for ZN and

Correspondence to: Camilla Rodrigues, P D Hinduja National Hospital and Medical Research Centre, Veer Savarkar Marg,
Mahim (West) Mumbai, India. Tel: (+91) 22 2444 7795. Fax: (+91) 22 2444 9151/2318. e-mail: dr_crodrigues@hinduja
hospital.com
Article submitted 30 April 2010. Final version accepted 29 September 2010.

484

The International Journal of Tuberculosis and Lung Disease

LED smears; however, due to cost constraints only one


(generally the first ZN-positive) specimen was processed for mycobacterial culture. If all three specimens were ZN-negative, the last specimen was processed for culture. Only clinical specimens processed
for mycobacterial culture were included in the final analysis.
Methods
Two direct smears were prepared immediately from
each specimen received for mycobacterial culture, one
for routine ZN microscopy (using an Olympus CX31
light microscope, Olympus, Tokyo, Japan) and the
other for examination of auramine-O stained smears
with FM using the LuminTM (LW Scientific, Lawrenceville, GA, USA) LED attachment. Cerebrospinal fluid
(CSF) and other sterile body fluids were concentrated
by centrifugation at 3500 g for 15 min prior to smear
preparation and culture inoculation. Specimens from
non-sterile sites were digested and decontaminated
by the standard N-acetyl-L-cysteine 2% sodium hydroxide (NALC-NaOH) method.9 Tissues were homogenised and urines were centrifuged at 3500 g for
15 min prior to decontamination. Depending on the
quantity of the specimen, two concentrated smears
were prepared using standard NALC-NaOH from
the decontaminated sediment, one for ZN staining
and one for auramine-O staining.
Mycobacterial culture
Briefly, 0.5 ml of decontaminated/concentrated sediment was inoculated into a Mycobacteria Growth
Indicator Tube (MGIT) 960 vial (BD Diagnostics,
Sparks, MD, USA) and 0.2 ml onto Lwenstein-Jensen
(LJ) media. Inoculated MGIT vials were incubated in
a BACTEC MGIT 960 for 6 weeks and inoculated LJ
slants were incubated at 37C in a CO2 incubator. LJ
slants were inspected daily for growth for the first
week and then weekly for the next 10 weeks. Positive
growth was confirmed for Mycobacterium tuberculosis complex or non-tuberculous mycobacteria (NTM)
by the p-nitrobenzoic acid (PNBA) assay.
AFB smears
One set of smears were stained by the hot ZN technique.9 Premade fluorescent stain kits for mycobacteria from HiMedia Laboratories (Mumbai, India) were
used for fluorescence staining. Smears were flooded
with auramine-O for 15 min, then rinsed with water,
destained with acid-alcohol for 2 min, then rinsed
again with water. They were finally counterstained
with potassium permanganate for 24 min, then
rinsed with water and allowed to air dry.
The ZN-stained slides were read by experienced
laboratory technologists as part of routine laboratory
work. Each auramine-stained slide was read by two
independent research fellows, hereafter referred to as
LED1 and LED2, using the Lumin LED attachment

to an Olympus CH2 light microscope using 200


magnification. LED1 was a newly trained microscopist inexperienced in smear microscopy, whereas
LED2 was experienced in ZN microscopy but inexperienced with FM. All the microscopists were blinded
to all other smear results. The number of AFB observed
was quantified according to WHO guidelines.10 The
time required to read both ZN and auramine-stained
slides was estimated by logging the time at the start
and at the end of reading a group of slides. This time
was averaged for the number of slides read. Two preliminary pilot studies over a total of 7 days were conducted for the readers to gain experience, and agreement was found to be 93% ( = 0.85) before data
collection for the study began.
Quality control
Quality control was performed daily and with each
new batch of ZN and auramine stain with positive
(H37RV strain) and negative control (American Type
Culture Collection strain of Escherichia coli) slides.
For external proficiency testing a set of stained ZN
and auramine slides was sent to a reputed hospital
with about 20 years of experience in FM.
Statistical analysis
Sensitivity and specificity were calculated for LED
FM and ZN using mycobacterial culture as a reference standard. Subgroup analysis was performed according to source of specimens (pulmonary vs. extrapulmonary) and direct vs. concentrated slides. The
average time required to read the slides was compared
between LED FM and ZN. Inter-rater agreement between the two research technologists was calculated
using a weighted measure. Confidence intervals (CIs)
were calculated using Fishers exact methods, and statistical significance was defined as non-overlapping
95%CIs.

RESULTS
During the 6-month study period, 3729 clinical specimens (2781 pulmonary and 948 extra-pulmonary)
from 2306 patients were received for smear microscopy. Of these, 2275 specimens (1358 pulmonary and
917 extra-pulmonary) were processed for mycobacterial culture and were included in our analysis.
Among the pulmonary specimens, 1050 (91%) were
sputum and the remaining 308 (9%) included bronchoalveolar lavage (BAL), bronchial washings, tracheal secretions and other non-sputum specimens from
the respiratory tract. Extra-pulmonary specimens included abscesses (n = 296), biopsy/aspirates (n =
247), lymph nodes (n = 140), bone marrow (n = 3),
specimens from the gastro-intestinal system (n = 43),
the genitourinary system (n = 7), urine (n = 70), adrenal (n = 2), CSF (n = 35), joint (n = 17), thoracic

LED FM for detection of mycobacteria

Table 1

485

Accuracy of ZN and LED FM using mycobacterial culture as a reference standard

Pulmonary
ZN
Direct
Concentrated
LED1
Direct
Concentrated
LED2
Direct
Concentrated
Mean LED
Direct
Concentrated
Extra-pulmonary
ZN
Direct
Concentrated
LED1
Direct
Concentrated
LED2
Direct
Concentrated
Mean LED
Direct
Concentrated

Smear+/culture+
n/N

Sensitivity (95%CI)

Smear/culture
n/N

617/746
531/743

0.827 (0.7980.854)*
0.715 (0.6810.747)*

586/611
593/611

0.959 (0.9400.973)
0.971 (0.9550.983)

555/746
488/743

0.744 (0.7110.775)
0.657 (0.6210.775)

577/611
583/611

0.944 (0.9230.961)
0.954 (0.9340.969)

613/746
556/743

0.822 (0.7920.849)*
0.748 (0.7150.779)*

546/611
554/611

0.894 (0.8660.917)
0.907 (0.8810.928)

584/746
522/743

0.783 (0.7510.812)*
0.703 (0.6680.735)*

561.5/611
569/611

0.920 (0.8950.940)
0.931 (0.9080.950)

90/241
48/228

0.373 (0.3120.438)*
0.211 (0.1590.247)*

578/676
550/584

0.855 (0.8260.881)*
0.942 (0.9100.969)*

54/241
42/228

0.224 (0.1730.282)
0.184 (0.1350.242)

634/676
548/584

0.938 (0.9170.955)
0.938 (0.9030.969)

110/241
72/228

0.456 (0.3920.522)*
0.316 (0.2550.382)*

566/676
530/584

0.837 (0.8070.864)*
0.908 (0.8710.941)*

82/241
57/228

0.340 (0.2810.404)
0.250 (0.1950.314)

600/676
539/584

0.888 (0.8610.910)
0.923 (0.8870.955)

Specicity (95%CI)

* Indicates statistical signicance between direct and concentrated smear readings, based on non-overlapping 95%CIs.
ZN = Ziehl-Neelsen; LED = light emitting diode; FM = uorescent microscopy; + = positive; = negative; CI = condence interval.

(n = 17), wound (n = 15), spine (n = 9), ascitic (n =


8), gastric aspirate (n = 4), and ear/nose/throat (n =
4), with one specimen per patient.
Among the 1358 pulmonary specimens, 746 were
culture-positive, 611 were negative and one was repeatedly contaminated. The contaminated specimen
was excluded from the final analysis. Of the 917
extra-pulmonary specimens, 241 were positive and
676 were negative by mycobacterial culture. The rate
of culture positivity was 55% for pulmonary specimens and 26% for extra-pulmonary specimens.
LED FM and ZN results were evaluated against
mycobacterial culture (MGIT and LJ; Table 1). Of
746 culture-positive pulmonary specimens, 617 were
detected by ZN microscopy (82.7%, 95%CI 79.8
85.4), 555 were detected by LED1 (74.4%, 95%CI
71.177.5) and 613 were detected by LED2 (82.2%,
95%CI 79.284.9). Of 611 culture-negative pulmonary specimens, 586 were identified as negative by
ZN (95.9%, 95%CI 94.097.3), 577 were identified
as negative by LED1 (94.4%, 95%CI 92.396.1) and
546 were identified as negative by LED2 (89.4%,
95%CI 86.691.7). The mean sensitivity estimate of
LED FM for direct AFB smears was 78.3% (95%CI
75.181.2) and the mean specificity estimate was
92.0% (95%CI 89.594.0), using culture as the reference standard. Of the extra-pulmonary specimens,
241 were culture-positive and 676 were culturenegative. Ninety positive specimens were detected
by ZN microscopy (37.3%, 95%CI 31.243.8) and

578 negative specimens were correctly classified by


ZN microscopy (85.5%, 95%CI 82.688.1). LED1
correctly identified 54 culture-positive specimens and
634 culture-negative specimens (sensitivity 22.4%,
95%CI 17.328.2 and specificity 93.8%, 95%CI
91.795.5). LED2 correctly identified 110 culturepositive and 566 culture-negative specimens (sensitivity 45.6%, 95%CI 39.252.2 and specificity 83.7%,
95%CI 80.786.4).
Concentrated smears were available for 1354/1357
pulmonary and 812/917 extra-pulmonary specimens
(Table 1). A decrease in the sensitivity of both ZN

Table 2 Time spent on smear examinations (min/slide)*

ZN
Direct
Concentrated
LED1
Direct
Concentrated
LED2
Direct
Concentrated
Mean LED
Direct
Concentrated

Mean (95%CI)

Median [IQR]

2.48 (2.382.59)
1.11 (1.051.17)

2.40 [2.162.89]
1.05 [0.941.24]

1.08 (0.901.26)
0.42 (0.390.45)

0.90 [0.811.17]
0.41 [0.340.51]

1.73 (1.631.84)
0.74 (0.650.83)

1.71 [1.392.04]
0.66 [0.560.81]

1.41 (1.301.51)
0.58 (0.540.63)

1.39 [1.131.58]
0.53 [0.470.63]

* Estimates for direct smears measured over 115 days; estimates for concentrated smears measured over 72 days with ZN and 62 days with LED.
CI = condence interval; IQR = interquartile range; ZN = Ziehl-Neelsen;
LED = light emitting diode.

486

Table 3

The International Journal of Tuberculosis and Lung Disease

Inter-rater agreement

Comparison
Direct
LED1LED2
LED1ZN
LED2ZN
Concentrated
LED1LED2
LED1ZN
LED2ZN

Agreement
%

Kappa*

77.7
75.8
75.6

0.92
0.87
0.90

82.6
74.5
70.1

0.91
0.75
0.77

* Calculated using quadratic weights for distribution of smear positivity.


LED = light emitting diode; ZN = Ziehl-Neelsen.

and LED microscopy was observed with concentrated


smears in both pulmonary and extra-pulmonary specimens, with several comparisons reaching statistical
significance. Nearly all specificity estimates for both
ZN and LED microscopy were slightly higher with
concentrated smears, although only two comparisons
reached statistical significance.
The time spent on smear examination is reported
in Table 2. The mean time to read a direct smear was
2.48 min for ZN microscopy vs. 1.41 min for LED
microscopy, whereas the mean time to read a concentrated smear was 1.11 min for ZN microscopy and
0.58 min for LED microscopy.
Table 3 describes the inter-rater agreement between
the different readers for direct and concentrated
smears. Agreement was highest between the two LED
readers, and was better when using direct smears
compared to concentrated smears for all readers.

DISCUSSION
The goal of this study was to implement and evaluate
LED microscopy for the direct detection of AFB in a
high-incidence setting. Similar to many high-incidence
laboratories, our technologists are highly skilled and
experienced with ZN microscopy, but are unfamiliar with fluorescent microscopy. Despite the wellestablished benefits of FM, it has not been adopted
widely in low- and middle-income countries, mainly
because of the high maintenance equipment and upfront costs associated with traditional mercury-vapour
fluorescent microscopes. Through the introduction of
LED technology, FM has now become a feasible modality for TB diagnosis.11 Combined with the recent
WHO recommendations for LED microscopy to replace conventional FM and be phased in as an alternative for conventional ZN microscopy, it is anticipated that many TB laboratories currently using ZN
microscopy will consider switching to LED FM in the
near future.8
Here we report our experience in introducing LED
microscopy, and our attempts to validate its performance in a real-world setting. Through this process,
many pitfalls were encountered and lessons were
learned that are often unappreciated or unreported

in rigidly controlled diagnostic evaluations, and which


we feel are valuable to describe.
The first decision, which likely had a negative impact on our results, was to use the Lumin LED attachment at a magnification of 200. Similar to many
laboratories currently using ZN microscopy, we have
already invested in purchasing light microscopes.
There is thus great appeal in using simple, inexpensive objective lens attachments, such as Lumin, which
transform the existing equipment into fluorescent LED
microscopes. In the absence of recommendations to
the contrary, our microscopists chose the 200 magnification to screen slides as it was felt to be the most
comfortable viewing field and provided the most rapid
time to read smears. It is probable that using the
lower magnification of the LED device for screening
specimens resulted in lower detection rates as well as
lower specificity.6,12 Furthermore, studies comparing
the Lumin LED attachment with other LED devices
(Fraen FluoLED, Fraen, Reading, MA, USA6 and Zeiss
iLED, Oberkochen, Germany13) have reported inferior image quality and ease of focusing. Adding to
this potential problem, the light microscopes used with
the LED attachment were not identical to those used
by the ZN microscopists, but were older models, as
our study was run in parallel to ongoing laboratory
operations.
Perhaps the most significant problem encountered
in our study was the discrepancy in experience between our microscopists. It is a common misconception to consider the results of diagnostics as purely
objective, as it is common for estimates of accuracy to
be cited as discrete values, with sensitivity and specificity considered static for a given test.14 On the contrary, the performance of smear microscopy in all its
forms is highly dependent upon the setting, the patient
population and the operator performing the test.1519
As TB laboratories contemplate switching from ZN
to LED FM, training of technologists needs to be a
central consideration. The learning curve experienced
when introducing a new technique to a laboratory is
difficult to characterise, as is defining a threshold
where performance could be deemed adequate to begin using test results for patient care decisions. Our
laboratory was experienced in the use of ZN microscopy, but had no personnel experienced with FM,
and we likely underestimated the amount of training
required for our microscopists to become proficient
in FM.
Other aspects of our study also likely biased results in favour of ZN performance. In our routine
laboratory protocol, only one specimen per patient
undergoing investigation for TB diagnosis was subjected to mycobacterial culture, even though multiple
specimens are often submitted and examined by microscopy. This serves to minimise costs for our patients while still providing them with culture-based
TB diagnostics. However, using these culture results

LED FM for detection of mycobacteria

as our reference standard for the evaluation of LED


microscopy led to selection bias in our data. Although
only specimens that underwent culture were included
in this study, those specimens were often selected based
on the results of direct ZN microscopy. If only one
specimen of three submitted was positive using ZN
microscopy, that specimen would have been sent for
mycobacterial culture, whereas the results of reading
the smears with the investigational LED microscope
were not considered when selecting specimens to be
cultured. This would tend to inflate the sensitivity estimates of ZN microscopy, and may lower the specificity estimates of LED microscopy.
It is not clear why we found lower sensitivity when
concentrated smears were examined for both ZN and
LED fluorescent microscopy. It is possible that bacilli
were lost during the decontamination and centrifugation procedure. Unfortunately, this was likely confounded by the shorter time spent by microscopists
examining concentrated smears. The fact that concentrated smear examinations tended to yield somewhat higher specificity estimates suggests that there
was at least a component of an implicit threshold effect during the readings.
Finally, several issues arose from using mycobacterial culture as our reference standard. For example,
as we were attempting to perform operational research
using specimens submitted routinely, there was a mixture of both diagnostic and follow-up specimens. Patients receiving anti-tuberculosis treatment are more
likely to have positive sputum smears due to the ongoing shedding of dead and dying bacilli, while these
organisms will not result in positive cultures.20 This
could have affected ZN and LED specificity estimates
differentially, as fluorochrome dyes appear to stain
AFB damaged by anti-tuberculosis drugs at a higher
rate than carbol fuchsin.21 Due to the high volume of
specimens submitted to our laboratory, it was not possible for us to collect reliable data regarding the treatment status of patients at the time they submitted
specimens. This problem was possibly enhanced by the
patient population at our TB referral centre, which
included a large number of cases with chronic TB and
highly drug-resistant TB, both of which can yield
strains that are more difficult to isolate in culture.22
Mycobacterial culture requires a high level of expertise, and even small alterations in protocol can result in decreased yield. Specifically, decontamination is
a necessary step in specimen preparation to reduce the
burden of non-mycobacterial organisms. However, decontamination agents can and do kill mycobacteria
as well, but to a lesser degree.23 Specimens with low
numbers of mycobacteria are therefore particularly
prone to being over-decontaminated and can result
in false-negative cultures. In busy laboratories, there
can be implicit motivation to over-decontaminate:
an over-decontaminated specimen is likely to remain
negative and requires no further intervention, while

487

an under-decontaminated specimen requires repeat


processing and culture and delays results being sent
out to clinicians. While most laboratories aim for an
overall contamination rate of 25% in mycobacterial cultures, this does not completely eliminate the
possibility that some low-burden specimens will yield
negative results.9 The single specimen reported here
as contaminated remained so after repeated decontamination and re-culture attempts. The initial rate
of contaminated specimens during the study period
was 2.1%.
The choice of reference standard is key to any diagnostic evaluation. Mycobacterial culture is considered the gold standard for the diagnosis of active
TB as it is objective, generally more sensitive than
simpler diagnostics such as smear microscopy, and
highly specific, assuming proper speciation and absence of cross-contamination. However, true mycobacterial culture status will not necessarily correspond to true smear status. As described above, there
are situations where one might expect a specimen to
have a positive smear and a negative culture, and it is
not surprising for low-burden specimens to have a
negative smear and a positive culture. Furthermore,
cross-contamination can be a significant concern, especially in high-prevalence settings.24 This raises the
reasonable question of whether another reference
standard may be more appropriate when comparing
two or more microscopy methods, techniques or devices. Although culture remains the gold standard for
the diagnosis of patients with active TB, a microscopybased reference standard involving rechecking or consensus examinations may be more relevant for studies
designed to optimise smear microscopy and its processes, especially under real-world field conditions.
The WHO has recommended that LED microscopy
be phased in as an alternative for conventional ZN
microscopy in both high- and low-volume laboratories, and that the switch to LED FM be carried out
through a carefully phased implementation plan, using LED technologies that meet WHO specifications.8
The WHO is currently developing guidance for implementation of LED, including specifications for
approved LED technologies (including optics, light
source, magnification, etc.) and recommendations for
external quality assurance of LED microscopy for TB
diagnosis. We would also suggest that laboratories
planning on validating and implementing LED microscopy in their settings pay particular attention to
the training their technologists receive, as well as the
stains and protocols used for smear examination.

CONCLUSIONS
LED FM has significant benefits over ZN microscopy
and conventional FM. To realise these benefits, careful implementation is needed when introducing LED
FM into laboratories unfamiliar with the technique.

488

The International Journal of Tuberculosis and Lung Disease

Ensuring adequate training, detailed standard operating procedures and quality assurance programmes
will maximise accuracy. While evidence from implementation studies is very valuable, internal validation
of diagnostic devices through operational research
can be difficult and presents challenges often not encountered in tightly controlled research environments.
Acknowledgements
The authors thank Parmanand Deepchand Hinduja National Hospital and Medical Research Centre for their encouragement and
support. This study was supported by grants from the Canadian
Institutes of Health Research (CIHR MOP-89918) and European
Commission (TBSusgent, EU-FP7). MP was supported by a CIHR
New Investigator Award, and JM was supported by a Quebec Respiratory Health Training Fellowship. None of these agencies were
involved in the conduct or review of this study.

References
1 Lempert H. Fluorescence microscopy in the detection of tubercle bacilli. Lancet 1944; 244: 818822.
2 Marais B J, Brittle W, Painczyk K, et al. Use of light-emitting
diode fluorescence microscopy to detect acid-fast bacilli in sputum. Clin Infect Dis 2008; 47: 203207.
3 Steingart K R, Henry M, Ng V, et al. Fluorescence versus conventional sputum smear microscopy for tuberculosis: a systematic review. Lancet Infect Dis 2006; 6: 570581.
4 Bennedsen J, Larsen S O. Examination for tubercle bacilli by
fluorescence microscopy. Scand J Respir Dis 1966; 47: 114
120.
5 Van Deun A, Chonde T M, Gumusboga M, Rienthong S. Performance and acceptability of the FluoLED Easy module for
tuberculosis fluorescence microscopy. Int J Tuberc Lung Dis
2008; 12: 10091014.
6 Affolabi D, Torrea G, Odoun M, et al. Comparison of two
LED fluorescence microscopy build-on modules for acid-fast
smear microscopy. Int J Tuberc Lung Dis 2010; 14: 160164.
7 Anthony R M, Kolk A H, Kuijper S, Klatser P R. Light emitting
diodes for auramine O fluorescence microscopic screening of
Mycobacterium tuberculosis. Int J Tuberc Lung Dis 2006; 10:
10601062.
8 World Health Organization. Fluorescent light emitting diode
(LED) microscopy for diagnosis of tuberculosis: policy statement. Geneva, Switzerland: WHO, 2010. http: //www.who.int/
tb/dots/laboratory/who_policy_led_microscopy_july10.pdf
Accessed July 2010.
9 Kent P T, Kubica G P. Public health mycobacteriology: a guide
for the level III laboratory. Atlanta, GA, USA: US Department
of Health and Human Services, Public Health Service, Centers
for Disease Control, 1985.

10 De Kantor I N, Kim S J, Frieden T, et al. Laboratory services in


tuberculosis control. Geneva, Switzerland: Global Tuberculosis
Programme, World Health Organization, 1998.
11 Minion J, Sohn H, Pai M. Light-emitting diode technologies
for TB diagnosis: what is on the market? Expert Rev Med Devices 2009; 6: 341345.
12 Minion J, Brunet L, Pai M. Fluorescent light emitting diode
(LED) microscopy for the detection of Mycobacterium tuberculosis: a systematic review and meta-analysis. Approaches to
improve sputum smear microscopy for tuberculosis diagnosis;
Expert Group Meeting. Geneva, Switzerland: World Health
Organization, 2009. http://www.who.int/tb/dots/laboratory/
egmreport_microscopymethods_nov09.pdf Accessed January
2010.
13 Minion J. Evaluating light-emitting diode fluorescence microscopy for the diagnosis of active tuberculosis. MS thesis. Montreal, Canada: McGill University, 2010.
14 Mulherin S A, Miller W C. Spectrum bias or spectrum effect?
Subgroup variation in diagnostic test evaluation. Ann Intern
Med 2002; 137: 598602.
15 Matee M, Mtei L, Lounasvaara T, et al. Sputum microscopy
for the diagnosis of HIV-associated pulmonary tuberculosis in
Tanzania. BMC Public Health 2008; 8: 68.
16 Bruchfeld J, Aderaye G, Palme I B, et al. Evaluation of outpatients with suspected pulmonary tuberculosis in a high HIV
prevalence setting in Ethiopia: clinical, diagnostic and epidemiological characteristics. Scand J Infect Dis 2002; 34: 331
337.
17 Perkins M D, Roscigno G, Zumla A. Progress towards improved
tuberculosis diagnostics for developing countries. Lancet 2006;
367: 942943.
18 Slosarek M, Cend D, Dabrowska L, et al. Comparison of microscopic positivity in smears from sputa stained according
to Ziehl-Neelsen in different modifications. J Hyg Epidemiol
Microbiol Immunol 1977; 21: 715.
19 Kubica G P. Correlation of acid-fast staining methods with culture results for mycobacteria. Bull Int Union Tuberc 1980; 55:
117124.
20 Stone B L, Burman W J, Hildred M V, Jarboe E A, Reves R R,
Wilson M L. The diagnostic yield of acid-fast-bacillus smearpositive sputum specimens. J Clin Microbiol 1997; 35: 1030
1031.
21 Somoskovi A, Hotaling J E, Fitzgerald M, ODonnell D, Parsons L M, Salfinger M. Lessons from a proficiency testing event
for acid-fast microscopy. Chest 2001; 120: 250257.
22 Bastian I, Portaels F, eds. Multidrug-resistant tuberculosis.
London, UK: Kluwer Academic Publishers, 2000.
23 Burdz T V, Wolfe J, Kabani A. Evaluation of sputum decontamination methods for Mycobacterium tuberculosis using viable colony counts and flow cytometry. Diagn Microbiol Infect
Dis 2003; 47: 503509.
24 Frieden T, ed. Tomans tuberculosis: case detection, treatment,
and monitoringquestions and answers. Geneva, Switzerland:
World Health Organization, 2004.

LED FM for detection of mycobacteria

RSUM
C O N T E X T E : Laboratoire de mycobactriologie dun centre de rfrence de la tuberculose (TB) Mumbai, Inde.
O B J E C T I F S : Evaluer lexamen microscopique par fluorescence (FM) au moyen de la technologie des diodes
lectroluminescentes (LED) pour la dtection des bacilles
acido-rsistants dans un centre de rfrence tertiaire dans
un pays o la TB est endmique.
S C H M A : On a introduit la FM LED dans un laboratoire expriment en ce qui concerne lexamen microscopique au Ziehl-Neelsen (ZN), mais peu habitu FM.
On a valu FM LED paralllement avec les services de
routine dexamen microscopique ZN et on la compar
avec les cultures mycobactriennes servant de standard
de rfrence.
R S U LTAT S : Les chantillons provenant de 1357 cas
pulmonaires et de 917 cas extrapulmonaires ont t examins au cours de ltude. Nous avons trouv pour FM
LED une sensibilit de 78,3% ainsi quune spcificit de
92,0% par rapport la culture mycobactrienne pour

les chantillons dorigine pulmonaire et une sensibilit de


34,0% ainsi quune spcificit de 88,8% pour les chantillons dorigine extrapulmonaire. La dure moyenne
dexamen du frottis a t de 2,48 min pour le ZN vs.
1,41 min pour FM LED. On a pu identifier au cours de
lanalyse diffrents biais dans le schma de ltude et dans
sa mise en uvre, biais qui sont susceptibles dentraner
une sous-estimation de la prcision de FM LED et font
lobjet dune discussion dans le contexte dvaluations
futures de FM LED.
C O N C L U S I O N S : Bien que FM LED comporte des avantages significatifs par rapport lexamen microscopique
ZN et la FM conventionnelle, sa mise en uvre et sa
validation peuvent entraner des difficults qui pourraient gner lvaluation de ses performances. Une formation adquate ainsi que des procdures opratoires
standardises et dtailles sont importants pour maximiser la prcision.

RESUMEN
M A R C O D E R E F E R E N C I A : El laboratorio de micobacteriologa en un centro de referencia de tuberculosis en
Mumbai, India.
O B J E T I V O S : Evaluar la microscopia de fluorescencia con
tcnica de diodos electroluminescentes (LED) en la deteccin de los bacilos acidorresistentes en un centro terciario de referencia de un pas con tuberculosis (TB)
endmica.
M T O D O S : Se introdujo la tcnica de microscopia de
fluorescencia (FM) con LED en un laboratorio con experiencia en el estudio sistemtico por microscopia con
coloracin de Ziehl-Neelsen (ZN) y sin competencias
en FM. Se compar la nueva tcnica con el servicio
sistemtico de microscopia con coloracin de ZN, tomando el cultivo de micobacterias como la tcnica de
referencia.
R E S U LTA D O S : En el estudio se examinaron 1357 muestras de origen pulmonar y 917 extrapulmonares. En promedio, la microscopia de fluorescencia con LED ofreci

una sensibilidad de 78,3% y una especificidad de 92,0%


en comparacin con el cultivo en las muestras pulmonares; la sensibilidad fue 34,0% y la especificidad 88,8%
en las muestras extrapulmonares. El tiempo promedio
de examen por cada frotis fue 2,48 minutos con ZN,
contra 1,41 min con la FM con LED. Durante el anlisis
se observaron varios sesgos en los mtodos y el procedimiento, que pueden generar una subestimacin de la
precisin del sistema de FM con LED; estos sesgos se
examinan en el contexto de futuras evaluaciones de esta
nueva tcnica.
C O N C L U S I N : Si bien la FM con LED ofrece ventajas
considerables en comparacin con el examen con coloracin de ZN y la fluorescencia convencional, su ejecucin y validacin pueden presentar dificultades que obstaculicen la evaluacin del rendimiento diagnstico. Una
capacitacin adecuada y la formulacin de procedimientos operativos normalizados son aspectos importantes
en la obtencin de una mxima precisin.