Predictors of Mortality in Patients With Tuberculous

To the Editor—We read with great interest the article by van Laarhoven et al, and we commend the authors for reporting one of the largest prospectively followed patient cohorts with tuberculous meningitis [1]. The main findings of the study were that progressively increasing blood leukocytosis (>11 000 leukocytes/mm³), decreased ratio of glucose level in cerebrospinal fluid to that in blood (<0.15), CSF neutrophil percentage >20%, and fever (temperature, >38^oC) are predictors of mortality up to 365 days after diagnosis. We have certain observations regarding the findings of this study.

The median Glasgow coma scale (GCS) score reported by the authors was 13 (interquartile range, 11–15), which defines a mild encephalopathy. It would be interesting to know the causes of mortality in patients with mild encephalopathy. With interventions like CSF diversion, the GCS score may rapidly change during the first week of disease. Therefore, a GCS score based on a severity assessment 7–14 days after disease onset is suggested to be a better indicator of prognosis [2]. In the reported cohort, the median time to death was 4–6 days. An early death, within 1 week after disease onset, is probably related to severe encephalopathy, hydrocephalus, and infarcts. However, these details are not evident from the report. Details of causes responsible for early and late deaths, stratified by severity of encephalopathy, will help readers be better prepared for these complications of tuberculous meningitis. In fact, it would be vital for the treating physicians to know the cause of mortality (and timing of deterioration) in various risk groups. For example, patients with CSF neutrophilic leukocytosis may be at a risk of paradoxical worsening, so there may be role for a higher dose or prolonged course of corticosteroids.

The 2 most important determinants of mortality reported in previous studies—hydrocephalus and infarcts—were not evaluated in the Cox regression model summarized in Table 2 of the article by van Laarhoven et al [1]. It may be worthwhile to analyze these factors, as well. The other predictors of mortality, like systemic organ dysfunction, complications of CSF drainage, deep venous thrombosis, pulmonary thromboembolism, and healthcare-associated infections, also contribute to mortality.

One of the main reasons provided by the authors for higher mortality is an advanced stage of disease at presentation; however, only 89 of 608 patients were in stage 3. Presentation of stage-specific mortality data by the authors would help clarify this reasoning. Given the large cohort and a high mortality rate, a more detailed description of causes of mortality will further enhance the value of this report [3].

The British Medical Research Council scoring is based on GCS scoring. We wonder whether it therefore prudent to consider GCS and British Medical Research Council scoring together in the same model. We also would like to emphasize that tuberculous meningitis is frequently complicated by dysnatremia, and details of dysnatremia in the reported cohort would further enhance the information that readers get from this report.

Contrary to most guidelines, the patients were treated with 6-month regimens, so it would be interesting to know the rates of disease relapse and treatment failure after they completed the 6-month intensive regimen. Rates of hepatotoxicity and treatment interruption are also high in patients with tuberculous meningitis; a mention of these complications may help the reader.

Although we understand that a focus on brevity while reporting large cohorts may force authors to cut down on several details, additional details on causes of mortality are likely to enhance the clinical application of the results. Last, taking another look at the predictors by addressing some or all the above and previously reported predictors, may help us view the results of this large and relevant prospective study more thoroughly.

Note

Potential conflicts of interest. Both authors: No reported conflicts of interest. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References

Author notes