SEARCH

SEARCH BY CITATION

Keywords:

  • diagnosis;
  • smear-negative;
  • tuberculosis;
  • pulmonary;
  • algorithms;
  • decision-making

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Objectives

Diagnosis of smear-negative pulmonary tuberculosis (SNPT) remains a challenge, particularly in resource-constrained settings. We evaluated a diagnostic algorithm that combines affordable laboratory tools and a clinical prediction rule (CPR).

Methods

We derived, based on published evidence, a diagnostic algorithm for SNPT. Sputum concentration constitutes its first step. In suspects with negative results, SNPT probability is classified with a CPR as low (excluded), high (confirmed) or intermediate. For intermediate patients, sputum Middlebrook 7H9 liquid culture is performed, and they are assessed after 2 weeks. If clinically deteriorated, with still negative liquid culture, bronchoscopy is offered. Otherwise, results of Middlebrook 7H9 culture are awaited. We prospectively evaluated this algorithm against a reference standard of solid and liquid cultures in two reference hospitals in Lima, Peru.

Results

670 SNPT suspects were included from September 2005 to March 2008. The prevalence of SNPT was 27% according to the reference standard. The algorithm's overall accuracy was 0.94 (95% CI 0.91–0.95), its sensitivity was 0.88 (95% CI 0.82–0.92) and its specificity, 0.96 (95% CI 0.94–0.98). Sputum concentration, the CPR, Middlebrook 7H9 sputum culture and bronchoscopic samples defined a diagnosis of SNPT according to the algorithm in 57 (37%), 25 (16%), 63 (41%) and 8(5%) of patients, respectively. 65% of patients were diagnosed within 3 weeks.

Conclusions

The algorithm was accurate for SNPT diagnosis. Sputum concentration, CPR and selective Middlebrook 7H9 culture are essential components.

Objectifs

Le diagnostic de la tuberculose pulmonaire à frottis négatif (TPFN) reste un défi, en particulier dans les pays à ressources limitées. Nous avons évalué un algorithme de diagnostic qui combine des outils de laboratoire abordables et une règle de prédiction clinique (RPC).

Méthodes

Nous avons établi, sur base des données publiées, un algorithme de diagnostic pour la TPFN. La concentration des expectorations constitue sa première étape. Chez les suspects avec des résultats négatifs, la probabilité de TPFN est classée avec une RPC comme: faible (exclue), élevée (confirmée) ou intermédiaire. Pour les patients intermédiaires, la culture des crachats dans du milieu liquide Middlebrook 7H9 a été effectuée et évaluée au bout de deux semaines. Dans le cas d'une détérioration clinique avec une culture liquide toujours négative, la bronchoscopie est offerte. Dans le cas contraire, les résultats de culture dans le Middlebrook 7H9 sont attendus. Nous avons évalué prospectivement cet algorithme par rapport à une norme de référence basée sur la culture en milieux solides et liquides dans deux hôpitaux de référence à Lima, au Pérou.

Résultats

670 suspects de TPFN ont été inclus de septembre 2005 à mars 2008. La prévalence de la TPFN était de 27% selon la norme de référence. La précision globale de l'algorithme était de 0.94 (IC 95%: 0.91–0.95), sa sensibilité était de 0.88 (IC 95%: 0.82–0.92) et sa spécificité de 0.96 (IC 95%: 0.94–0.98). La concentration des expectorations, la RPC, la culture des expectorations dans le Middlebrook 7H9 et des échantillons bronchoscopiques ont défini un diagnostic de la TPFN selon l'algorithme chez 57 (37%), 25 (16%), 63 (41%) et 8 (5%) patients respectivement. 65% des patients ont été diagnostiqués endéans 3 semaines.

Conclusions

L'algorithme était précis pour le diagnostic de la TPFN. La concentration des expectorations, la RPC et la culture sélective dans le Middlebrook 7H9 sont des éléments essentiels.

Objetivos

El diagnóstico de la tuberculosis pulmonar con baciloscopia negativa (TPBN) continúa siendo un gran reto, particularmente en lugares con pocos recursos. Hemos evaluado un algoritmo de diagnóstico que combina herramientas de laboratorio asequibles y una regla de predicción clínica (RPC).

Métodos

Basándonos en evidencia publicada, derivamos un algoritmo diagnóstico para TPBN. El primer paso es concentrar el esputo. En pacientes con resultados negativos pero con sospecha de TB la probabilidad de TPBN se clasifica con una RPC como baja (excluída), alta (confirmada) o intermedia. Para pacientes con una clasificación intermedia, se realiza un cultivo de esputo en medio líquido Middlebrook 7H9 y se evalúan dos semanas más tarde. Si están clínicamente deteriorados, con un cultivo líquido aún negativo, se realiza una broncoscopia. Evaluamos de forma prospectiva este algoritmo frente a un estándar de referencia de cultivos sólido y líquido en dos hospitales de referencia en Lima, Perú.

Resultados

Se incluyeron 670 pacientes con sospecha de TPBN entre Septiembre 2005 y Marzo 2008. La prevalencia de TPBN era un 27% según el estándar de referencia. La precisión del algorithmo era de 0.94 (IC 95% 0.91–0.95), la sensitibilidad era de 0.88 (IC 95% 0.82–0.92) y la especificidad de 0.96 (IC 95% 0.94–0.98). La concentración de esputo, el RPC, el cultivo de esputo Middlebrook 7H9 y las muestras de la broncoscopia definieron un diagnóstico de TPBN según el algoritmo en 57 (37%), 25 (16%), 63 (41%) y 8(5%) de los pacientes respectivamente. Un 65% de los pacientes fueron diagnosticados en un plazo de tres semanas.

Conclusiones

El algoritmo era preciso para el diagnóstico de TPBN. La concentración de esputo, la RPC y el cultivo en Middlebrook 7H9 eran componentes esenciales.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

For decades, the diagnosis of pulmonary tuberculosis (TB) has relied on the search for acid-fast bacilli (AFB) in sputum smears, a procedure that has a sensitivity of only 50–60% (Aber et al. 1980; Lipsky et al. 1984; Siddiqi et al. 2003). A case of smear-negative pulmonary tuberculosis (SNPT) is defined, according to WHO (2010), as a clinical suspect with at least two negative AFB smears and a positive culture, or two negative smears, radiographical abnormalities consistent with active pulmonary TB, no improvement with a course of broad-spectrum antibiotic (if HIV-negative), and the decision to treat for TB.

The existing guidelines for diagnosing SNPT in resource-constrained settings, published by the International Union Against Tuberculosis and Lung Disease (Ait-Khaled et al. 2010), WHO (2010) and the American Thoracic Society (Blumberg et al. 2003), are somewhat vague. In general, they hinge on ‘strong clinical evidence’, ‘medical officer's judgment’ or ‘additional investigations’, without detailing the diagnostic work-up. Nowadays, some affordable and accessible options are available for a more refined diagnosis of SNPT. Direct examination of concentrated sputum is more sensitive than conventional microscopy (Steingart et al. 2006). Fluorescence microscopy improves the diagnostic yield of sputum samples (WHO 2011a). Cultures in solid media, classically the reference standard, have been superseded by liquid cultures in terms of accuracy and time to diagnosis (WHO 2007), and liquid cultures have become more widely available. Work performed on clinical prediction rules (CPR) for SNPT also shows encouraging results (Bah et al. 2002; Mello et al. 2006; Soto et al. 2008b).

We aimed to construct and evaluate an efficient diagnostic algorithm for SNPT in resource-constrained settings that combines these laboratory tools with a CPR.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Setting and study population

The study was conducted in Lima, Peru. Peru has a tuberculosis incidence of 101/100.000, a 2.9% prevalence of HIV infection among TB cases and 22.5% smear negativity among pulmonary cases (WHO 2012). We recruited patients from the outpatient internal medicine and pneumology clinics, the emergency room and the pneumology and internal medicine wards in two-third-level public hospitals, Hipólito Unanue and Cayetano Heredia. They are referral centres for a population of about two million.

Assuming an overall accuracy of the tested algorithm of 80%, 95% confidence level and 10% loss to follow-up, the sample size needed was 546 patients. From September 2005 to March 2008, all patients aged 18 years or more with clinical suspicion of pulmonary tuberculosis and at least two negative AFB sputum smears were eligible for inclusion. Clinical suspicion was defined as cough for a minimum of 2 weeks and fever, weight loss of more than 3 kg or dyspnoea.

The algorithm

The algorithm for SNPT was developed by the research team and local experts, based on the literature review and using the Delphi panel method (Collecting group data 2010). Sputum concentrate AFB smear examination constitutes its first step (Figure 1). If positive, patients are referred to the National TB programme for treatment. If negative, a previously developed CPR (Soto et al. 2008b) is applied to assess the probability of tuberculosis. It considers as clinical predictors of TB the presence of hemoptysis (2 points), weight loss (1 point), ability to expectorate (−1 point) and age older than 45 years (−1 point) and as radiographical predictors the presence of apical infiltrate (3 points) and miliary pattern (4 points). A score less than 0 points indicates a very low probability of having SNPT (except for HIV positive patients, who are then considered to have intermediate probability). The patient is, according to the algorithm, deemed not to have tuberculosis. A score of more than 4 points corresponds to a high probability of the disease (except in patients with a history of previous tuberculosis, who are then considered to have intermediate probability). The patient is classified as having tuberculosis and referred to the national TB programme for treatment.

image

Figure 1. Flow of smear-negative pulmonary tuberculosis (SNPT) suspects through and diagnosis with the SNPT algorithm. Lima, Peru 2005–2008. Right (black) boxes represent patients diagnosed with SNPT and left (grey) boxes patients with SNPT excluded according to the algorithm.

Download figure to PowerPoint

Patients with scores from 0 to 4 points or otherwise intermediate probability of SNPT are given a course of antibiotics (doxycycline 100 mg bid for 10 days), and a liquid culture in Middlebrook 7H9 (hereafter referred as liquid culture) taken from their sputum concentrates is performed. They are evaluated after 2 weeks, when preliminary liquid culture results are available. If the culture is negative, the clinical status is assessed through patient's appreciation and clinical examination. Patients who improved are no longer followed up, but they are traced back if their liquid culture result eventually turns out to be positive. Patients with unchanged clinical condition are given an appointment 4 weeks later, when the final results of liquid culture are available. Patients with clinical deterioration are offered fiberoptic bronchoscopy (FBO) with bronchial aspirate for AFB smear and liquid culture.

Study Procedures

For all eligible patients that gave informed consent, we recorded signs and symptoms, demographic information and medical history. HIV counselling and testing was offered, and those testing positive were referred to the national HIV programme for corresponding management.

Microbiological investigation included Ziehl-Neelsen and auramine-O (Van Deun et al. 2008) staining of concentrated sputum specimens. Liquid culture in Middlebrook 7H9 was a step in the algorithm, but to construct the reference standard, it was performed in all patients entering the algorithm, along with mycobacteria growth indicator tube (MGIT) and solid (Ogawa) culture. Sputum decontamination and concentration, as well as manual MGIT (Becton-DickinsonTM) cultures were performed as previously reported (Soto et al. 2008a). Ogawa cultures followed conventional methodology (Kent & Kubica 1985). Due to stock shortage from September 2005 to January 2006 and from July to September 2006, MGIT could not be performed for 130 patients. Middlebrook 7H9 (DIFCOTM) vials were prepared according to manufacturer instructions and used in conjunction with OADC (oleic acid, albumin, dextrose and catalase) supplement and the antimicrobial combination PANTA (polymyxin B, amphotericin B, nalidixic Acid, trimethoprim and azlocillin). The appearance of lump-like colonies was considered as a positive result. Positive cultures in MGIT or Middlebrook 7H9 were recultured in Ogawa for M. tuberculosis typification. Cultures were evaluated twice weekly for up to 6 weeks for liquid cultures and 8 weeks for Ogawa.

Analysis

For evaluating the performance of the SNPT algorithm, the reference standard for the diagnosis of SNPT was defined as a positive M. tuberculosis culture in Ogawa or a positive culture in MGIT or Middlebrook 7H9 confirmed by reculture in Ogawa.

Overall accuracy was defined as the percentage of correct results obtained by the algorithm. Sensitivity, specificity, predictive values and likelihood ratios for the algorithm were calculated and reported according to international recommendations (Bossuyt et al. 2003). We conducted a stratified analysis for patients with or without previous TB history and with or without HIV infection. We also evaluated the diagnostic performance of sputum concentration and compared the diagnostic yield and growing times of Middlebrook 7H9, manual MGIT and Ogawa media. Proportions and means were compared with chi-squared and t-test, respectively, using paired approaches when necessary. P-values < 0.05 were considered significant. Data analysis was performed with STATA statistical software (Release 10.0; Stata Corporation, College Station, TX, USA).

Ethics

The research protocol was approved by the ethics committees of the participating hospitals, Cayetano Heredia University and the Institute of Tropical Medicine, Antwerp. Written informed consent was obtained from all patients included in the study.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

A total of 1010 patients were eligible for the study. Of these, 670 entered the algorithm. The reasons for exclusion are detailed in Figure 2. Seven patients had contaminated results in all three culture media and were excluded from the evaluation of the algorithm. Of the remaining 663 patients, 182 (27.5%) had, according to the reference standard, a final diagnosis of SNPT (Figure 2). Clinical, sociodemographic, laboratory and radiological findings are summarised in Table 1. Patients with SNPT were younger and more likely to have hemoptysis, fever, weight loss, an abnormal Chest X-ray and apical infiltrates than patients without SNPT. Previous tuberculosis, being able to expectorate and clubbing were negatively associated with a final diagnosis of tuberculosis. The presence of an abnormal physical lung examination and a positive HIV serology were not associated with SNPT.

Table 1. Sociodemographic characteristics, signs, symptoms and results of examinations in smear-negative pulmonary tuberculosis suspects. Lima, 2005–2008
VariableAll (n = 663)SNPT(n = 182)No TB (n = 481)P value
Sociodemographic characteristics
Mean Age (SD)41.4 (17.2)36.3 (15.9)43.3 (17.3)<0.001
TB contact -no. (%)314 (47.4)91 (50.0)223 (46.4)0.40
Male gender (%)370 (55.8)108 (59.3)262 (54.5)0.26
History of previous TB (%)239 (36.1)43 (23.6)196 (40.8)<0.001
Comorbidities
Alcoholism (%)41 (6.2)11 (6.0)30 (6.2)0.93
Diabetes (%)16 (2.4)6 (3.3)10 (2.1)0.36
Drugs (%)29 (4.4)10 (5.5)19 (4.0)0.39
Renal failure (%)10 (1.5)1 (0.6)9 (1.9)0.21
Symptoms and physical examination
Weight loss (%)417 (62.9)130 (71.4)287 (59.7)0.01
Expectoration (%)443 (66.8)108 (59.3)335 (69.7)0.01
Hemoptysis (%)205 (30.9)77 (42.3)128 (26.6)<0.001
Abnormal lung examination (%)435 (65.6)122 (67.0)313 (65.1)0.64
Lymphadenopathy (%)35 (5.3)14 (7.7)21 (4.4)0.09
Hepatomegaly (%)45 (6.8)18 (9.9)27 (5.6)0.05
Mean blood pressure (SD)84.0 (9.22)84.1 (9.6)84.0 (9.1)0.91
Mean Temperature (SD)36.7 (0.5)36.9 (0.6)36.7 (0.4)<0.001
Mean BMI (SD)22.4 (4.14)22.0 (4.0)22.5 (4.2)0.30
Clubbing (%)54 (8.1)4 (2.2)50 (10.4)<0.01
Ancillary examinations
Mean WBC count × 1038.38 (3.6)8.83 (3.4)8.21 (3.7)0.06
Mean Lymphocyte count × 1032.01 (1.1)1.98 (1.2)2.02 (1.0)0.66
Chest X-ray findings(%)
Abnormal CXR (%)538 (81.2)165 (90.7)373 (77.6)<0.001
Apical infiltrate (%)379 (57.2)132 (72.5)247 (51.4)<0.001
Miliary infiltrate (%)5 (0.7)3 (1.7)2 (0.4)0.10
ELISA HIV serology
Positive (%)98 (14.8)23 (12.6)75 (15.6)0.08
Negative (%)310 (46.8)98 (53.9)212 (44.1)
Refused (%)255 (38.5)61 (33.5)194 (40.3)
image

Figure 2. Comparison of the results of the algorithm with smear-negative pulmonary tuberculosis (SNPT) diagnosis according to reference standard. Lima, Peru 2005–2008. *44 lost to follow-up and 15 refusals of bronchoscopy.

Download figure to PowerPoint

The patient flow through the algorithm is shown in Figure 1. By the second visit, 59 of the 670 patients (8.8%) had a positive concentrated sputum smear. 57 of these had SNPT according to the reference standard and constituted 31% of all SNPT patients. The remaining 611 patients were evaluated with the clinical prediction rule (CPR) during the same visit, which allowed taking a decision in 104 (15.5%) further patients. 39 had a high probability of SNPT and were classified, following the algorithm, as having tuberculosis. 25 of these (64.1%) had TB according to the reference standard. On the other hand, 65 patients had a low probability and were classified as not having SNPT. Two of these (3.1%) had TB according to the reference standard.

The 507 patients that had an intermediate CPR probability received a course of antibiotics and were scheduled for a follow-up visit 2 weeks later (week 3). By then 12 of these had a positive liquid culture; for another 12 the culture was contaminated. The latter were classified as non-conclusive by the algorithm, but three of them eventually had a reference standard diagnosis of TB. 44 patients did not show up and were classified as non-evaluable by the algorithm (5 of them eventually were diagnosed with TB according to the reference standard).

439 patients with intermediate CPR probability and negative cultures by week 3 were available for clinical re-evaluation. 293 had improved. Liquid culture became positive in 32 (10.9%) by week 7. The remaining 261 were classified as not SNPT by the algorithm, but nine (3.4%) had tuberculosis according to the reference standard. 104 patients had unchanged clinical status, of whom 17 (16.3%) eventually had a positive and 87 (83.7%) had a negative liquid culture. Of the latter, 10 (11.5%) had tuberculosis according to the reference standard.

Forty-two patients clinically deteriorated and were offered bronchoscopy. Fifteen did not accept the procedure and were classified as non-evaluable by the algorithm. None of them had TB according to the reference standard. Of the 27 patients who consented, six had a positive AFB from bronchial aspirate, two subsequently had a positive liquid culture from the initial sputum sample and two had a positive culture from bronchial aspirate only. Liquid cultures from sputum and bronchial aspirate samples of the remaining 17 patients were negative and they had non-conclusive results according to the algorithm. Their cultures were also negative in MGIT and Ogawa media, and they had no TB according to the reference standard.

After excluding non-evaluable patients (n = 59) and those with non-conclusive results (n = 29) according to the algorithm, 576 patients with non-contaminated cultures were available for evaluation of the algorithm performance (Table 2). The overall accuracy, sensitivity, specificity, positive and negative likelihood ratios of the algorithm as a whole were 0.94, 0.88, 0.96, 22.1 and 0.13, respectively. The area under the ROC curve was 0.92 (95% CI 0.89–0.94) and the diagnostic Odds Ratio 175.8 (95% CI 89.3–345.8). Algorithm accuracy was better in patients with history of TB (0.98 vs. 0.91; P = 0.001) but was not affected by HIV status. The cumulative percentage of the 169 patients diagnosed with SNPT by the algorithm attained 58.1% and 65.1% by the end of weeks 1 and 3, respectively. Losses to follow-up cannot have a significant impact on the estimated accuracy of the algorithm. Even under the extreme assumption that all losses were incorrectly classified, the overall accuracy only drops from 0.94 to 0.87.

Table 2. Performance of the SNPT algorithm according to history of tuberculosis and HIV status. Lima, 2005–2008
 Overall (n = 576)History of TuberculosisHIV status
Yes (n = 206)No (n = 370)Positive (n = 89)Negative/Unknown (n = 487)
  1. SNPT, Smear-negative pulmonary tuberculosis; PPV, Positive predictive value; NPV, Negative predictive value; LR, Likelihood ratio; 95% Confidence intervals between parentheses.

Accuracy0.94 (0.91–0.95)0.98 (0.95–0.99)0.91 (0.88–0.94)0.94 (0.87–0.98)0.93 (0.91–0.95)
Sensitivity0.88 (0.82–0.92)0.95 (0.84–0.99)0.86 (0.79–0.91)0.86 (0.65–0.97)0.88 (0.82–0.93)
Specificity0.96 (0.94–0.98)0.99 (0.96–1.00)0.94 (0.90–0.96)0.94 (0.90–0.99)0.96 (0.93–0.98)
LR positive22.1 (13.6–35.8)78.1 (19.7–310.1)14.6 (8.7–24.3)28.9 (7.3–114.4)21.09 (12.6–35.3)
LR negative0.13 (0.08–0.19)0.05 (0.01–0.19)0.15 (0.10–0.23)0.14 (0.05–0.40)0.12 (0.08–0.19)
Prevalence0.30 (0.26–0.34)0.20 (0.15–0.27)0.36 (0.31–0.41)0.25 (0.16–0.35)0.31 (0.27–0.36)
PPV0.91 (0.85–0.94)0.95 (0.84–0.99)0.89 (0.82–0.94)0.90 (0.70–0.99)0.91 (0.85–0.95)
NPV0.95 (0.92–0.97)0.90 (0.96–1.00)0.92 (0.88–0.95)0.96 (0.88–0.99)0.95 (0.92–0.97)

Regarding microbiological procedures, 49 patients had positive results for both Ziehl-Neelsen and auramine-O staining of sputum concentrates, two for Ziehl-Neelsen and eight for auramine-O staining only. The incremental diagnostic yield of fluorescence microscopy over Ziehl-Neelsen staining was 16% (08/51) (P = 0.10). Non-contaminated samples for comparison between the three culture media were available for 520 patients. No species other than M.tuberculosis were isolated from cultures. The sensitivity of culture in Middlebrook 7H9 (0.77; 95% CI 0.69–0.84) was lower (P = 0.03) than that in manual MGIT (0.86; 95% CI 0.79–0.91) but significantly higher (P = 0.01) than in Ogawa (0.63; 95% CI 0.55–0.71, P < 0.01). The mean growing time (19.4 days; 95% CI 18.1–20.8) was comparable to that for MGIT (20.7 days 95% CI 19.1–22.2) and significantly (P < 0.001) shorter than for Ogawa (31.3 days; 95% CI 28.6–34.0). The combination of Middlebrook 7H9 and MGIT attained a sensitivity of 0.93 (95% CI 0.88–0.97) against the reference standard, while the combination of Middlebrook and Ogawa attained a sensitivity of 0.91(95% CI 0.85–0.95).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The algorithm provides a structured and feasible approach towards the diagnosis of SNPT and permits an efficient use of laboratory resources, which is of particular importance in low and middle income countries. Its diagnostic performance is good. Sputum concentration, the CPR and selective liquid culture are its essential components. The overall accuracy is 0.94 regardless of HIV status. The attained positive (22.1) and negative (0.13) likelihood ratios provide, according to proposed standards (Dujardin et al. 1994; Jaeschke et al. 1994), convincing diagnostic evidence in favour and against disease, respectively. Moreover, in a study population with a prevalence of 27% of SNPT, the positive predictive value was 0.91 and the negative, 0.95. This underlines that the algorithm can be used for clinical decision-making. It even performs better than culture of all sputum samples in solid medium alone (0.88 vs. 0.63 of sensitivity) and allows more timely decision-making.

AFB smears of concentrated sputum, a simple and cheap technique, detect 31% of cases of SNPT and substantially reduce the diagnostic delay. A systematic review (Steingart et al. 2006) showed a 17% increase of sensitivity over direct sputum examination, and similar yields for the diagnosis of SNPT have been reported previously (Apers et al. 2004). The false positive rate was very low (3%) and the corresponding patients had wasting and radiographic abnormalities consistent with tuberculosis. Most probably they were culture-negative SNPT cases. Bleach sedimentation has been proposed as an alternative to centrifugation of sputum specimens (Bonnet et al. 2008), but a recent meta-analysis found no convincing evidence for its utility (Cattamanchi et al. 2010).

The clinical prediction rule, based on symptoms and chest X-ray findings, which we use in the algorithm has been evaluated previously as stand-alone tool (Soto et al. 2011a). It identified well patients with high and low probabilities of SNPT and should permit to rationalise the use of liquid culture by reserving it for suspects with intermediate probability. After the CPR step, the algorithm has detected 45% of SNPT cases, but three-quarters of all suspects who entered in the present study still proceeded to culture and antibiotic trial, a higher proportion than we expected. Possibly, many suspects with low SNPT probability never reach our referral hospitals because their symptoms subside. This is in line with the high prevalence of SNPT (30%) in our suspect patients. On the other hand, patients with high probability might have been recognised as such at lower levels of care and put on treatment without further investigations. We hypothesise that the yield of the CPR and the number of cultures foregone would be higher at lower levels of care, but this remains to be verified. At any rate, while the place of the CPR needs further assessment through implementation studies under operational conditions, it hardly produces false negatives (3%), even in the challenging conditions of a referral setting like ours. Furthermore, part of the false positive CPR cases could well be culture-negative SNPT cases, in line with historical studies where 50% of culture-negative SNPT cases diagnosed on clinical and radiographic grounds, became bacteriologically positive during follow-up (Hong Kong Chest Service 1981).

It is of note that the frequency of SNPT differs little between patients with improved (11%) and unchanged (16%) clinical condition after being given an antibiotic course and that a differential approach to their follow-up seems hardly warranted. A high proportion of suspects with intermediate CPR probability responded favourably to antibiotics. Among them, over 10% had a final diagnosis of SNPT. Improvement must be viewed with caution and does not permit to exclude the disease, as also demonstrated in an evaluation of the current WHO algorithm for HIV-negative smear-negative SNPT suspects (Soto et al. 2011b). In the present study, all SNPT patients could eventually be put on antituberculous treatment, but if culture is not performed in suspects who improve under broad-spectrum antibiotics, they are lost or their diagnosis will be seriously delayed.

Deteriorating under an antibiotic trial, however, is associated with an increased probability of TB, even in the event of a negative liquid culture result. For this smaller group of patients, more sensitive but costly diagnostic techniques could be considered. Our results suggest that bronchoscopy can be useful, but it is limited by the small number of eligible patients and the high refusal rate. We feel that the decision to perform bronchoscopy should be individualised based on expert evaluation of the patient's clinical condition until its value for SNPT is more firmly established.

A complementary finding of our study is that the yield of the liquid cultures is much higher in SNPT suspects than the yield of conventional solid cultures. This is in line with results of other studies (Muyoyeta et al. 2009; Shinu et al. 2011) and recent WHO (2007) recommendations that encourage the use of liquid media. Middlebrook 7H9 is cheap and more sensitive than conventional solid cultures. Furthermore, the combination of Middlebrook 7H9 with classic solid media gives better yields than each method alone, based on our results. Surprisingly, in spite of being the basis of methods like MGIT or MODS (Moore et al. 2006), the literature on the diagnostic performance of Middlebrook 7H9 in itself is scarce.

Alternative diagnostic techniques, such as polymerase chain reaction, are prohibitive in terms of costs in most settings, and their diagnostic performance in smear-negative tuberculosis is controversial (Sarmiento et al. 2003). The recent Xpert MTB/RIF assay (Boehme et al. 2010) may become more affordable for less developed countries after its endorsement by WHO (2011b). However, technical requirements make it unsuitable as a tool for lower levels of care and, above all, its sensitivity for the diagnosis of SNPT is just around 0.75 (Chang et al. 2012). Our algorithm was not designed for primary health care centres either, but we demonstrated that it can validly guide the management of SNPT suspects at levels where SNPT suspects are generally referred to: intermediate or reference hospitals with perhaps limited resources but radiographic and culture capacity.

The algorithm needs further adaptation before being applied at programmatic level or in a population with lower TB incidence, but our results are already of value for refining current SNPT diagnostic guidelines. The sequential use of sputum concentration and liquid culture, particularly Middlebrook 7H9, should be encouraged, and an antibiotic course should be disregarded as diagnostic tool. The role and place of CPRs definitely needs more operational research. Finally, research on the diagnosis of SNPT has focused on single tests. However, diagnostic test results should not be considered in isolation but contribute to the overall evidence on which a diagnosis is made (National Institute for Health & Clinical Excellence 2011). The thoughtful combination of diagnostic tools, the conceptual basis for algorithm design, should be the focus of future research rather than the quest for a single test.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank Yeny Bravo and Rosa Reinoso, who participated in the patient recruitment and Tullia Battaglioli for her comments on the manuscript. The work was funded by the Damian Foundation. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Aber VR, Allen BW, Mitchison DA et al. (1980) Quality control in tuberculosis bacteriology. 1. Laboratory studies on isolated positive cultures and the efficiency of direct smear examination. Tubercle 61, 123133.
  • Ait-Khaled N, Alarcon E, Armengol R et al. (2010) Management of Tuberculosis: A Guide to the Essentials of Good Practice. International Union Against Tuberculosis and Lung Disease, Paris.
  • Apers L, Wijarajah C, Mutsvangwa J et al. (2004) Accuracy of routine diagnosis of pulmonary tuberculosis in an area of high HIV prevalence. The International Journal of Tuberculosis and Lung Disease 8, 945951.
  • Bah B, Massari V, Sow O et al. (2002) Useful clues to the presence of smear-negative pulmonary tuberculosis in a West African city. The International Journal of Tuberculosis and Lung Disease 6, 592598.
  • Blumberg HM, Burman WJ, Chaisson RE et al. (2003) American Thoracic Society/Centers for Disease Control and Prevention/Infectious Diseases Society of America: treatment of tuberculosis. American Journal of Respiratory and Critical Care Medicine 167, 603662.
  • Boehme CC, Nabeta P, Hillemann D et al. (2010) Rapid molecular detection of tuberculosis and rifampin resistance. The New England Journal of Medicine 363, 10051015.
  • Bonnet M, Ramsay A, Githui W et al. (2008) Bleach sedimentation: an opportunity to optimize smear microscopy for tuberculosis diagnosis in settings of high prevalence of HIV. Clinical Infectious Diseases 46, 17101716.
  • Bossuyt PM, Reitsma JB, Bruns DE et al. (2003) Towards complete and accurate reporting of studies of diagnostic accuracy: the STARD Initiative. Annals of Internal Medicine 138, 4044.
  • Cattamanchi A, Davis JL, Pai M et al. (2010) Does bleach processing increase the accuracy of sputum smear microscopy for diagnosing pulmonary tuberculosis? Journal of Clinical Microbiology 48, 24332439.
  • Chang K, Lu W, Wang J et al. (2012) Rapid and effective diagnosis of tuberculosis and rifampicin resistance with Xpert MTB/RIF assay: a meta-analysis. Journal of Infection 64, 580588.
  • Collecting group data (2010) Delphi technique. University of Wisconsin Extension. http://www.uwex.edu/ces/pdande/resources/pdf/Tipsheet4.pdf. (accessed 5 February 2013).
  • Dujardin B, Van den Ende J, Van GA, Unger JP & Van der Stuyft P (1994) Likelihood ratios: a real improvement for clinical decision making? European Journal of Epidemiology 10, 2936.
  • Hong Kong Chest Service (1981) A study of the characteristics and course of sputum smear-negative pulmonary tuberculosis. Tubercle 62, 155167.
  • Jaeschke R, Guyatt GH & Sackett DL (1994) Users’ guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The Evidence-Based Medicine Working Group. The Journal of the American Medical Association 271, 703707.
  • Kent P & Kubica G (1985) Public Health Mycobacteriology. A Guide for the Level III Laboratory. Center for the Disease Control, Atlanta, Georgia.
  • Lipsky BA, Gates J, Tenover FC & Plorde JJ (1984) Factors affecting the clinical value of microscopy for acid-fast bacilli. Reviews of Infectious Diseases 6, 214222.
  • Mello FC, Bastos LG, Soares SL et al. (2006) Predicting smear negative pulmonary tuberculosis with classification trees and logistic regression: a cross-sectional study. BMC Public Health 6, 43.
  • Moore DA, Evans CA, Gilman RH et al. (2006) Microscopic-observation drug-susceptibility assay for the diagnosis of TB. The New England Journal of Medicine 355, 15391550.
  • Muyoyeta M, Schaap JA, De HP et al. (2009) Comparison of four culture systems for Mycobacterium tuberculosis in the Zambian National Reference Laboratory. The International Journal of Tuberculosis and Lung Disease 13, 460465.
  • National Institute for Health and Clinical Excellence (2011) Clinical diagnosis and management of tuberculosis, and measures for its prevention and control. http://www.nice.org.uk/nicemedia/live/13422/53642/53642.pdf. (accessed 05 February 2013)
  • Sarmiento OL, Weigle KA, Alexander J, Weber DJ & Miller WC (2003) Assessment by meta-analysis of PCR for diagnosis of smear-negative pulmonary tuberculosis. Journal of Clinical Microbiology 41, 32333240.
  • Shinu P, Nair A, Singh V, Kumar S & Bareja R (2011) Evaluation of rapid techniques for the detection of mycobacteria in sputum with scanty bacilli or clinically evident, smear negative cases of pulmonary and extra-pulmonary tuberculosis. Memórias do Instituto Oswaldo Cruz 106, 620624.
  • Siddiqi K, Lambert ML & Walley J (2003) Clinical diagnosis of smear-negative pulmonary tuberculosis in low-income countries: the current evidence. The Lancet Infectious Diseases 3, 288296.
  • Soto A, Agapito J, Acuna-Villaorduna C et al. (2008a) Evaluation of the performance of two liquid-phase culture media for the diagnosis of pulmonary tuberculosis in a national hospital in Lima, Peru. International Journal of Infectious Diseases 13, 4045.
  • Soto A, Solari L, Agapito J et al. (2008b) Development of a clinical scoring system for the diagnosis of smear-negative pulmonary tuberculosis. Brazilan Journal of Infectious Diseases 12, 128132.
  • Soto A, Solari L, Diaz J et al. (2011a) Validation of a clinical-radiographic score to assess the probability of pulmonary tuberculosis in suspect patients with negative sputum smears. PLoS ONE 6, e18486.
  • Soto A, Solari L, Gotuzzo E et al. (2011b) Performance of an algorithm based on WHO recommendations for the diagnosis of smear-negative pulmonary tuberculosis in patients without HIV infection. Tropical Medicine and International Health 16, 424430.
  • Steingart KR, Ng V, Henry M et al. (2006) Sputum processing methods to improve the sensitivity of smear microscopy for tuberculosis: a systematic review. Lancet Infect Dis 6, 664674.
  • Van Deun A, Chuquiyauri R, Torrea G et al. (2008) Yield of fluorescence microscopy versus culture for tuberculosis at a middle-income country referral hospital. Transactions of the Royal Society of Tropical Medicine and Hygiene 102, 564569.
  • WHO (2007) Improving the Diagnosis and Treatment of Smear-Negative Pulmonary and Extrapulmonary Tuberculosis Among Adults and Adolescents. WHO/HTM/TB/2007.379. World Health Organization, Geneva.
  • WHO (2010) Treatment of Tuberculosis Guidelines (4th ed). WHO/HTM/TB/2009.420. World Health Organization, Geneva.
  • WHO (2011a) Fluorescent light-emitting diode (LED) microscopy for diagnosis of tuberculosis. Policy statement. http://whqlibdoc.who.int/publications/2011/9789241501613_eng.pdf. (accessed 05 February 2013)
  • WHO (2011b) Policy statement: automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF system. http://whqlibdoc.who.int/publications/2011/9789241501545_eng.pdf (accessed 05 February 2013)
  • WHO (2012) Global Tuberculosis Report 2012. http://www.who.int/tb/publications/global_report/en/ (accessed 05 February 2013)