Alert sign and symptoms for the early diagnosis of pulmonary tuberculosis: analysis of patients followed by a tertiary pediatric hospital

Tuberculosis (TB) is one of the top 10 causes of death worldwide [1]. It typically affects lungs (80% are pulmonary TB) but can also affect other sites (extrapulmonary TB 15–20%). In order to contain epidemic and achieve the END TB strategy (TB elimination), defined by WHO [2] new strategies are needed, such as screening programs, contact tracing [3‐5], early detection of suspected cases and preventive therapy to reduce disease progression in infected patients (treating latent TB) [6, 7]. Childhood disease is often acquired following close contact with an infectious adult case, and in 30–40% involves household contacts [8, 9]. The risk of progression is higher (40–50%) in children < 5 years especially in the first 12 months after infection [10, 11], and severe TB disease, including TB meningitis, is highest in infants (< 2 years) and usually occurs within few months of primary infection [9, 12]. Adult risk of progression is lower, only 5% of patients infected develop active disease. Diagnosis of TB in children is often challenging because it’s paucisyntomatic at onset; gold standard in diagnosis remains culture of bacteria, but as it can take up to 8 weeks, and considering the paucibacillary nature of childhood lung disease, microbiological confirmation is possible in 35–40% of children [13]. In these cases diagnosis is based on the triad: clinical findings, risk factors and radiographic pattern (Table 1) [14‐18].

The most frequently reported symptoms in literature are unremitting cough (≥ 2 weeks duration) and persistent fever (> 2–3 weeks) [4, 19‐22]. Considering that most studies have been conducted in high prevalence countries [19, 23‐26], we analyzed children’s signs/symptoms, risk factors and anamnestic data of TB at onset in a low prevalence setting. So far several attempts have been made to validate clinical scores for the early diagnosis of pulmonary TB [24, 27‐29], but none of them have been internationally validated and in most cases without a control group [30]. The aim of the study was to describe signs and symptoms of retrospective cases of patients with pulmonary TB and to evaluate alerts that could intercept earlier suspected TB cases in emergency setting [31, 32].

The study was conducted at the Bambino Gesù Pediatric IRCCS in Rome, a tertiary urban pediatric hospital. The study was approved by the Ethics Committee of our Children’s Hospital according to the Declaration of Helsinki.

TB diagnosis in children is often delayed because clinical presentation at onset is non-specific and there isn’t always a history of contact/exposure to a TB case. Furthermore, diagnostic methods available often don’t give immediate confirmation [33]. The two cohorts analyzed (retrospective and prospective) allowed us to extrapolate signs/symptoms at onset and anamnestic data (including risk factors), in order to evaluate whether there’s a different clinical presentation and/or different risk factors in low a prevalence country compared to what is described in literature. In the prospective study a questionnaire was administered to parents, assessing clinical symptoms and anamnestic data. So far most studies have been conducted in high TB incidence countries [19, 23‐26, 34]. In these countries, the correlation between respiratory symptoms and TB may be different than in low endemic countries. In addition, risk factors may also have a different weight, for example contact with a TB case may be a less discriminating factor as frequently present in anamnesis, meanwhile in low-endemic countries it assumes a greater weight.

Contact is one of the most significantly correlated risk factor, described in 53.5% of patients with TB in the retrospective study and in 50% of cases in the prospective study, while only 7% among non-TB patients (prospective study). The OR (adjusted 19.8) confirms the correlation between contact and TB, as in other studies where contact is a frequently described risk factor (53.2–65.2% of patients) [20, 29, 35]. Therefore this data will certainly have a relevant weight in a clinical score. Unfortunately, a child is often just a sentinel case of an adult infection [36].

Another important risk factor is having parents from countries at risk, as highlighted in our retrospective series in 60.2% of cases and confirmed in other studies [20, 37, 38]. In the prospective study parents didn’t declare on the questionnaire their citizenship/native country, so we don’t have this data (question to add to the questionnaire proposed).

Being born in countries at risk is infrequent in our cohorts (12.8% of retrospective cases, 14.3% of prospective). This obviously depends on the country in which the study is performed, for example in USA a higher percentage is described (31% of children with TB are born abroad) [38].

Another variable investigated is use of antibiotic therapy and whether with or no benefit. In our study almost all of our TB patients treated previously with first-line antibiotics had no benefit: in the retrospective cohort 91/92 (98.9%) referred no benefit; in the prospective study 3/3 (100%). Comparing the prospective TB and non-TB cases (Table 4), antibiotic therapy was mostly prescribed in non-TB cases (69% vs 21.4%) and this probably correlates with higher frequency in this group of fever and persistent cough, and consequently prescription of first-line therapy. In the non-TB group 77.5% of children referred poor response to antibiotic therapy, not so striking as in the TB patients, but expected because if they improved they wouldn’t have come to the ED. To date various studies identified poor response to first-line therapy as an element of suspicion [4, 20, 28, 39, 40], however a large-scale comparison between positive and negative cases is not reported, but necessary to confirm these findings.

Immunosuppression or the intake of immunosuppressive drugs is not significant in our cohorts, in which there is a low incidence of HIV.

Focusing on symptoms at onset, in low income countries patients often present to healthcare facilities in advanced stages of the disease [41], while in high income countries physicians are questioned almost immediately, so diagnosis is often made in a subclinical phase, before symptoms becoming persistent. A symptom frequently reported in literature as suggestive and associated with TB is cough, considered of strong suspicion especially if lasting > 2 weeks(15–17) and described in 44–52% of children at onset [42‐44]. Unlike what is described in international literature, in our retrospective population 49.6% of cases referred cough, but persistent only in 21.7%, and the same result emerged among TB-cases of the prospective study with persistent cough in 21.4% of patients, similarly to previous Italian studies (28.7%) [20].

Another symptom referred is fever which occurs in 46% of retrospective cases, but lasting ≥ 10 days in only 19%, while there are few cases with persistent fever at onset in the prospective cohort (7.1%). These data differs from the 35–47% frequency described in literature, both in italian studies/low prevalence countries [20, 29], and internationally [22, 45]. Comparing duration of cough and fever in the two cohorts (retrospective and prospective, Tables 3, 4 and 5), although with different proportions of patients analyzed, it seems that persistent cough and fever (lasting ≥ 10 days) is not so strongly suggestive of TB diagnosis in our Italian reality. The prospective study, comparing cases with and without TB, allowed us to evaluate the correlation between fever and persistent cough with the diagnosis of TB, and it seems to confirm this poor correlation. In the multivariate analysis the OR with a value of 0.03 for fever ≥ 10 days, and 0.05 for cough ≥ 10 days shows an almost protective effect for TB, so they seem more related to other diagnoses. These findings can have 2 different explanations: on one hand, diagnosis is often made in a subclinical phase, and this allows treatment in early stages of the disease; on the other hand, the low prevalence determines a lower probability that a child with a persistent cough and fever will have TB in our reality.

Another variable analyzed is lymphadenopathy, which is not always a symptom that lead to suspect TB, but if persistent despite first-line therapies and isolated symptom, needs to exclude Non-Tuberculous Mycobacterium infection. In clinical scores proposed for the diagnosis of TB lymphadenopathy is usually included, and frequency at onset is described (19.8–23.8%) [29, 46]. In our retrospective analysis 15.9% of cases had lymphnode involvement at onset, while in the prospective cohort lymphadenopathy was referred in 42.9% of TB cases compared to 15.5% of non-TB, with an OR of 28.6 (95%CI 2.1–391.2), so probably it’s a symptom that could have a greater weight in our population and is likely underestimated. The very large CI95% is related to the low sample size and less accurate estimate.

Weight loss, which international studies reported in 22.6–27% of cases [29, 42], was reported in 9.7% of patients in the retrospective analysis, similarly to what is described in low prevalence countries like Italy (9.9% of cases) [20]. The prospective analysis, on the other hand, highlights a higher frequency (21.4% in TB cases, 19.7% non-TB cases). However it’s not a discriminating factor for diagnosis, therefore in our screening process it won’t have a significant weight.

Another symptom usually associated with TB is asthenia, in some studies in 36.6% of patients [29, 41], while from our retrospective cohort it seems less frequent(15.5% of cases). In the prospective study instead it’s equally described both in patients with (50%) and without (60%) TB, therefore it doesn’t discriminate positive patients from negative ones. This higher frequency may be related to the questionnaire submitted to parents, which points out a less alarming symptom compared to others (such as fever, cough, lymphadenopathy).

Chest pain is another symptom typically associated with pleuritic involvement [17], mostly described in older children, as the ability to express this symptom is different [19]. Our retrospective case study with 7.1% of cases confirms the incidence described in the national studies (5%) [20]. In the prospective study 1/14 (7%) TB patients presented chest pain at onset, 11/71 (15%) non TB patients, with no statistically difference. However, in the questionnaire it was associated with cough symptom, but as these symptoms aren’t always related, it will be better to separate them (proposed questionnaire in Additional File 1).

Hemoptysis and night sweats are rarely reported symptoms in our retrospective cohort (2.6% and 1.8% of cases respectively), similarly to other italian studies (haemoptysis 2.7%, night sweats 1% cases) [20] and international ones [10, 47]. In the prospective cohort night sweats are never described in TB cases, and described in 33.8% of non-TB cases, probably linked to fever symptom. Therefore, these variables are not clinically relevant for early diagnosis.

We finally analyzed diagnostic tests performed, considering retrospective data (larger sample with 226 patients). Both Mantoux and IGRA were positive in most patients (92.9% and 95.6% respectively), while only 42.2% were culture positive. Although culture is gold standard for active TB diagnosis, it’s lack of positivity must not delay the diagnostic-therapeutic process, considering children’s higher risk of progression to disease after infection [7, 8, 29, 48‐51]. Low sensitivity and specificity of microbiological tests is related to the paucibacillary nature of pediatric TB, with only 20–50% of pulmonary TB cases being culture positive [52]. Bacterioscopic test was positive only in 18% of cases, compatible with it’s low described sensitivity (60%)(45), while PCR was positive in 37.6% of our patients, similarly to other studies reports (30.4–32%) [20, 35].

Our study highlighted that in our country, at low TB prevalence, pulmonary TB has a different clinical pattern. Persistent cough and fever are less frequently reported in our patients, while significant risk factors pointed out are contact history with TB case and lymphadenopathy. Another element of suspicion, although not statistically significant in our study because of small sample size, is poor response to antibiotic therapy in TB patients. Anamnestic data like child’s and parents’ origins are also crucial. In order to validate a clinical score for pulmonary TB in a low prevalence country like Italy, it’s necessary to review risk factors that must guide the diagnostic-therapeutic workup. In our population with low HIV and TB prevalence, the most correct inclusion criteria could be the following: Fever lasting ≥ 10 days; Cough lasting ≥ 10 days; Poor response to antibiotic therapy; Lymphadenopathy; Mantoux and/or IGRA positive; Chest X-ray suspected for TB; History of contact with TB case.

It’s necessary to increase sample involving more Centers in order to validate a clinical score that allows a promptly diagnosis, limiting the number of diagnostic tests where unnecessary. From this preliminary analysis, its feasibility emerged, albeit with some variations both in terms of the inclusion criteria, and on the questionnaire (illustrated in Additional File 1).

Finally, only the comparison between cases with and without TB on a larger sample will be able to confirm or highlight the red flags that must lead us to suspect TB in order to avoid unnecessary risks of interpersonal transmission, particularly in overcrowded places like ED waiting rooms.