Introduction

Sepsis and tuberculosis kill around 11 million [1] and 1.5 million people per year [2], respectively. Tuberculosis sepsis (TBS), also known as sepsis tuberculosis gravissima, was first described by Landouzy in 1908 [3]. TBS has been mostly reported in HIV-infected patients [4,5,6,7]; however, it can also occur in immunocompetent patients [8, 9].

It has been estimated that half of TBS remains undiagnosed at the time of death [10]. An analysis of the United States’ databases of patients with sepsis showed that fifteen percent of patients did not have clinical signs and symptoms leading to suspect sepsis on admission [11]. The mortality rate was worst in this group of patients as compared to the group of patients with sepsis at presentation [11]. Prompt diagnosis and early treatment are key to the management of sepsis. Delay in antibiotics administration is associated with the worsening of sepsis severity both in sepsis in general and in TB sepsis in particular [12, 13].

TBS carries a fatal prognosis because it is overlooked by clinicians; mainly in the case of patients without HIV infection [10]. A case series reported by Kethireddy et al. suggests that most TBS patients died [13], likely because TBS is not coming into the mind of the clinician as an alternative diagnostic or etiology of organ failure. These findings highlight the critical need to improve clinicians’ awareness of TBS. This systematic review and meta-analysis aim to better understand the epidemiology, clinical features, and factors associated with the treatment outcome of TBS in HIV-negative patients.

Methods

The search strategy and inclusion criteria

The review was undertaken and reported by following the preferred reporting items for systematic review and meta-analysis (PRISMA 2020 and PRISMA-S guidelines [14, 15]. The protocol of the review was registered with PROSPERO (CRD42022296768).

An electronic search of the published literature was conducted on December 1, 2021, and updated on September 25, 2022, in PubMed, Embase, Web of Science (core collection) and Google Scholar to identify case reports or case series of tuberculosis sepsis. As suggested by PRISMA-S [15] and Bramer and collaborators [16], the first 200 results on Google Scholar were selected. We also searched the reference lists of the included case reports. The following search terms were used in PubMed:

("tuberculosis"[All Fields] OR "tuberculosis"[MeSH Terms] OR "tuberculosis"[All Fields] OR "tuberculoses"[All Fields] OR "tuberculosis s"[All Fields]) AND ("sepsis"[MeSH Terms] OR "sepsis"[All Fields]) AND ((("ieee int conf automation sci eng case"[Journal] OR "case phila"[Journal] OR "case"[All Fields]) AND "report*"[All Fields]) OR (("ieee int conf automation sci eng case"[Journal] OR "case phila"[Journal] OR "case"[All Fields]) AND "serie*"[All Fields])). The full description of the search strategy of the others databases used is reported in Supplementary File S1. Additionally, we conducted a cross-reference analysis to retrieve manuscripts that were not identified during our initial search. With the purpose of uniformly applying consensus criteria for the definition of sepsis, we restricted the case reports or series to be included to those published after the first consensus definition of sepsis by the American College of Chest Physicians and the Society of Critical Care Medicine (1991) [17, 18]. We excluded case reports (a) with unclear clinic-pathological data of the diagnosis of sepsis or lack of information on the diagnostic method and treatment outcome; (b) duplicate cases using Rayyan platform [19]; (c) TBS cases in HIV-infected patients; (d); case reports in languages other than English or French; (e) tuberculosis bloodstream infections not fulfilling sepsis criteria; (f) sepsis cases due to Mycobacterium species other than Mycobacterium tuberculosis, M. bovis and M. africanum; and (g) new-borns or infants with congenital tuberculosis not reported as tuberculosis sepsis. The titles and abstracts were initially screened independently by two reviewers (EJR and NOE). The full texts of the relevant articles were assessed for inclusion by two independent reviewers (BRA and NOE) using the Rayyan platform [19]. The agreement of both reviewers was required for inclusion and exclusion. Any disagreement was resolved by consensus. If BRA and NOE did not agree after discussion, a third investigator (YJH) was consulted. The full list of excluded cases is reported in Supplementary File S2.

Data extraction and quality assessment

The following data were extracted from the original studies: first author; year of publication; country of origin; study population and participant demographics and baseline characteristics; clinical features, outcomes, and times of measurement. BRA and NOE independently extracted data using the items pre-defined on the excel sheet. The quality of included studies was assessed with the Joanna Briggs Institute Critical Appraisal Checklist for Case Reports [20], which consists of eight yes/no/unclear questions which led to the overall appraisal: ‘Include’, ‘Exclude’ or ‘Seek further info’.

Summary measures and statistical analysis

Descriptive statistics of publication characteristics and patient demographic variables were performed. Case report data were grouped by type of patients (adult or infant). The patients’ sociodemographic data were presented separately for adults and infants. The means of age, time to diagnosis of tuberculosis sepsis, time to initiation of an empiric anti-tuberculosis treatment and corresponding standard deviation (SD) were described for each category of patients. To better describe factors associated with death; the clinical features, the predisposing co-morbidities, and initial diagnosis were ranged into three categories maximum. Fisher’s exact test was used to identify factors significantly associated with death. Cox’s regression logistic was not performed as planned in the protocol due to the small number of cases.

Results

The initial database search identified 2337 articles, of which 1416 were screened using the abstract and title. A total of 137 full articles were assessed for eligibility and 25 articles (28 individual case reports) were included in the final analysis (Fig. 1). All case reports included in this review met the ‘Include’ overall appraisal using the Joanna Briggs Institute tool and were classified as Low risk of bias (see Supplementary File S3).

Fig. 1
figure 1

Flowchart of the study selection process

Full size image

Patient demographics, co-morbidities, predisposing factors, and clinical manifestation

Among the total of 28 individual cases [8, 9, 21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43], four of them were infants (Table1). The mean age of the infants was 23 days (SD = 9). The mean age of adults was 44 years (SD = 18). A total of 15/28 (54%) cases were women. For 50% (14/28) of the patients, no pre-disposing factors were reported. BCG vaccination status was only reported in two cases. The main chief complaints were ‘weakness’ and ‘dyspnoea’, Table 2. A total of 61% (17/28) of patients reported cough, and all of the patients had a fever, respiratory distress, and hypotension. Leukopenia was the most common laboratory abnormality reported.

Table 1 Overview of included studies and descriptive data
Full size table
Table 2 Clinical features and underlying conditions in 25 cases of tuberculosis sepsis
Full size table

Tuberculosis sepsis diagnosis and patient management and treatment outcomes

For 75% (3/4) of infants, the initial diagnosis at admission was sepsis, and patients were managed in the intensive care unit; while only 46% (11/24) of adults were diagnosed with sepsis on admission. The empiric anti-tuberculosis was initiated in 53% (15/28) of the patients, and the mean time lapse from presentation to treatment initiation was six days (SD = 9; Table 3). Most (18/28; 64%) patients died within 30 days of the presentation, and in 39% (7/18), the TBS diagnosis was confirmed post-mortem. The diagnosis of tuberculosis was confirmed by either blood culture, extrapulmonary liquid culture, or sputum microscopic, TB molecular diagnosis/culture or pathology of infected organ biopsy (Table 1). The probability of survival of patients diagnosed with tuberculosis sepsis was 68% on day 10; 41% on day 20; and 33% on day 30 after admission, Fig. 2. A higher proportion of death occurred in patients with other diagnoses than sepsis at admission. The mean time (day) of starting empiric anti-tuberculosis treatment since the presentation was 10 (SD = 4) in patients who died, while the treatment was imitated earlier in the patients who survived (mean time = 5, SD = 3).

Table 3 Summary of initial diagnosis and length of time between admission and confirmation or suspicion of tuberculosis sepsis diagnosis
Full size table
Fig. 2
figure 2

Survival probability patients diagnosed with tuberculosis sepsis

Full size image

Discussion

We performed a systematic review and meta-analysis of published cases of tuberculosis sepsis in HIV-negative patients from 1991, up to September 25, 2022. The major findings were that the half of patients did not report known underline immunocompromised co-morbidity, 61% (17/28) reported cough at admission, 64% (18/28) died within 30 days since presentation and the TBS diagnosis was confirmed only at post-mortem for 39% (7/18) of the patients who died. A higher proportion of death occurred in adult patients with delayed initiation of anti-tuberculosis treatment.

The major challenge of tuberculosis sepsis is the delay of diagnosis or difficulty of its recognition by clinicians. Our study confirms that there are no specific signs or symptoms of TBS and that patients present with the common sign of sepsis. Recent research and advocacy improve the awareness of clinicians on tuberculosis bacteraemia and/or sepsis in HIV patients [44,45,46,47]. However, our review shows that tuberculosis sepsis can occur in patients without HIV infection or known co-morbidities. Therefore, TBS should be investigated in any tuberculosis patient presenting sepsis signs irrespective of HIV status. In highly TB- endemic settings, we recommend broadening the investigation of the etiology of sepsis to Mycobacterium tuberculosis. The culture of samples is currently the gold standard for tuberculosis diagnosis. However, culture takes two to six weeks to be reported. Therefore, rapid molecular diagnostic such as GeneXpert is suggested in the purpose to allow as soon as possible anti-tuberculosis treatment. However, the pulmonary manifestations were not common in all cases of TBS and the M. tuberculosis is not identifiable all-time in the sputum. The sensibility of GeneXpert in extrapulmonary samples ranged from 55.2 to 69.9% [48,49,50]. Barr et al. suggested that a combination of sputum GeneXpert, blood culture and urine lipoarabinomannan, could improve the diagnostic yield of TB in critically ill adult patients [14]. There is a need to improve the diagnostic tools for disseminated tuberculosis which increase the risk to develop TBS. A risk score derived from a model containing independent predictors has been suggested; however, it was derived from patients with HIV infection only and needs to be validated in other settings [7]. The clinical feature of TBS does not seem to be different according to the HIV infection status of the patients. Therefore, the clinician should keep in mind the alternative diagnosis of TBS in HIV-negative patients with tuberculosis signs associated with organ failure manifestations. With BCG vaccination protecting from TB meningitis and sepsis at least at a younger age, it would be important if future reports would include this information routinely.

Strengths and weaknesses

Using a systematic search strategy in four widely used databases, we increased the chance to identify all case reports of tuberculosis sepsis. Despite having applied refined selection inclusion criteria, there is the possibility of missing some important case reports. Publication bias is another weakness of case report review since only rare and untypical observations are more likely to be published. The publication of case reports also depends on the research experience and ability of the clinician in charge of the case seeming to be unusual. The BCG vaccination status was reported only in one case report. The proportion of TBS cases without known co-morbidity is impressive. Since this is a summary of case reports published, we cannot ensure that an in-depth investigation (including further laboratory assessments) will confirm or refute the absence of co-morbidities in these patients. Therefore, the interpretation of our findings should be done with caution. The role of radiology imaging in the diagnosis of not bacteriological confirmed is well known. We did not extract such information from the articles included. Despite these limitations, our review is of clinical practice and research implications interest because findings from this review contribute to improving the awareness of clinicians on the clinical feature of tuberculosis sepsis and showed that it could occur in infant and adult patients irrespective of HIV infection status.

Conclusions

TBS is reported as a case report because of its rare incidence in many settings. It is not usually suspected in the first place, inducing a risk of delayed diagnosis. Most case reports had death as an outcome and the mortality rate is more frequent in groups of patients where the diagnosis of TBS was not suspected at admission. In an endemic setting, TBS should be envisaged in patients with tuberculosis likely symptoms presenting sepsis signs as well. Empiric anti-tuberculosis treatment should be initiated as soon as possible.