Improving early diagnosis of tuberculous meningitis in children

Thesis (PhD)--Stellenbosch University, 2015.ENGLISH SUMMARY: Due to the sub-optimal performance of definite diagnostic tests, the early identification of paediatric TBM suspects relies on a thorough assessment of all the evidence derived from a careful history, clinical examination and relevant investigations. Rapid diagnosis is needed for early treatment initiation but microbiological confirmation is difficult at stage 1 disease because a lumbar puncture is generally only done once signs of meningitis have developed (stage 2 and 3 TBM). With this thesis, I aim to investigate mechanisms of improving the early and/or more accurate diagnosis of childhood TBM. PART I In chapter 2 I provide an update on the diagnosis and management of TBM in children, based on local experience, which can be transposed to similar settings. Highlights include firstly that short (6 months) intensified therapy in children with drug susceptible TBM is safe and effective, with a good outcome and low mortality. Secondly, home-based TBM treatment after initial in-hospital stabilization is feasible in carefully selected patients under close supervision. Thirdly, treatment of tuberculous hydrocephalus depends on the level of cerebrospinal fluid (CSF) obstruction. Communicating hydrocephalus can be successfully treated with medical therapy with normalization of intracranial pressure occurring within days in the majority children and non-communicating hydrocephalus requires neurosurgical intervention. Fourthly, thalidomide is the local drug of choice in children who develop life-threatening TB mass lesions (IRIS) despite corticosteroids. PART II In clinical practice, TBM diagnosis is most often based on a combination of clinical, laboratory and radiological findings. A uniform research case definition utilizing these criteria was proposed by an international panel of experts as a means of improving diagnostic standardization in order to answer critical research questions, categorizing patients as definite, probable, or possible TBM. Part 2 of my thesis focuses on the diagnostic utility of the uniform research case definition criteria for TBM. In chapter 3.1 I retrospectively evaluate the diagnostic performance of probable and possible TBM criteria in children with culture-confirmed TBM and culture-confirmed bacterial meningitis. The proposed uniform research case definition provided excellent diagnostic accuracy compared to microbiologically-confirmed TBM, when a ‘probable’ TBM score was used. When a ‘possible’ TBM score was used, not a single TBM case would have been missed, but clinical utility was minimal given the low specificity achieved. In order to strengthen my findings I prospectively assessed the diagnostic accuracy of the uniform TBM research case definition (see chapter 3.2). Excellent diagnostic accuracy was obtained for a diagnosis of TBM when compared to bacterial and viral meningitis controls. The high specificity of a probable TBM score justifies its use as an alternative reference standard to microbiological confirmation in future studies. In both studies poor sensitivity was obtained when a probable TBM score was used to diagnose early (stage 1) TBM, emphasizing a very high clinical index of suspicion of TBM in young children with recent TB exposure and persistent non-specific signs. CSF findings are essential to early diagnosis of TBM. Cut-off values for CSF glucose in TBM lack evidence. A CSF protein cut-off of >1g/L (100mg/dL) differentiated between cases of TBM and other forms of meningitis. My study on the diagnostic value of cerebrospinal fluid chemistry results in childhood TBM found that the optimal lower limit of CSF glucose concentration as a diagnostic aid for TBM was 2.2 mmol/L (see chapter 3.3). Absolute CSF glucose differentiated non-TBM from TBM cases with sensitivity of 68% and specificity of 96%, excluding its use as a ‘rule-out’ test. Simultaneous determination of serum and CSF glucose was seldom performed but my findings suggest that the CSF:serum glucose ratio may further improve diagnostic sensitivity.. CSF protein cut-off of >1g/L as well as CSF macroscopic appearance, cell counts and the presence of lymphocyte predominance are required to assist the distinction between TBM and bacterial meningitis. Previous studies suggest that chest X-ray findings consistent with active pulmonary TB are observed in 70% to 84% of children with TBM. In my study (chapter 3.4) only 46% of cases with TBM had chest radiograph findings highly suggestive of pulmonary TB. A need to treat calculation showed that only 1 in 4.39 children ≤3 years of age with TBM are likely to have ‘certain TB’ on chest X-ray. PART III Microbiological confirmation of TBM remains the gold-standard of diagnosis, but is challenging in young children due to the paucibacillary nature of disease, low CSF volumes available for diagnostic analysis and sub-optimal sensitivity of direct microscopy for acid-fast bacilli and M.tuberculosis culture on CSF. Several new commercially available NAA tests have been developed for the rapid diagnosis of TB. In part III of the thesis my meta-analysis of newer commercial NAA tests found a summary sensitivity of 64% and specificity of 98%. Summary sensitivity of commercial NAA tests remains suboptimal and is unlikely to greatly enhance early accurate diagnosis. However, the excellent specificity suggests that commercial NAA tests may be regarded as definitive in the correct clinical setting. In 2013, the WHO recommended Xpert MTB/RIF® as the preferential initial investigation in all adults and pediatric TBM suspects. My sub-analysis of 5 studies reporting Xpert MTB/RIF® on CSF, found summary sensitivity of 70% and specificity of 97%. In chapter 4.2 I aim to assess the utility of MTBDRplus® and Xpert MTB/RIF® to diagnose TBM in a clinical setting, alone and/or in combination. The main finding was the incremental increase in diagnostic accuracy achieved with combined use of these commercial NAA tests performed on CSF. Although both NAA tests were superior to liquid culture, sensitivity remained low compared to a rigorous predefined clinical case definition. The MTBDRplus® assay performed with sensitivity of 33% (98% specificity), Xpert MTB/RIF® was 26% sensitive (100% specificity) and combining positive results from both these tests provided a sensitivity of 49% (98% specificity) against a TBM case definition. This is insufficient to serve as a rule-out test and provides limited clinical guidance. However, microbiological confirmation provided by a positive test prevents unnecessary treatment delay and potential life-threatening consequences. The additional advantage of a positive NAA test is that of early detection of mycobacterial resistance. A major limitation of this study was the failure to improve diagnosis of stage 1 childhood TBM, mainly because it was hospital-based. Good surveillance at primary healthcare level, identifying children with poor weight-gain (or weight loss) and persistent non-remitting cough for longer than 2 weeks, could improve the detection of both childhood TB and early TBM. IMCI is potentially a valuable tool to increase awareness of TBM among healthcare workers, and in detecting early TB and TBM, as it is practiced at the healthcare level of first contact. Household contact tracing and prophylaxis with isoniazid therapy, as well as more general measures such as improving nutrition, housing and poverty relief, are valuable measures in preventing TBM in young children.AFRIKAANSE OPSOMMING: Die vroeë identifisering van kinders met vermoedelike TBM berus op 'n deeglike geskiedenis, kliniese ondersoek en toepaslike spesiale ondersoeke weens die lae sensitiwiteit van definitiewe mikrobiologiese diagnose huidiglik. Vroeë diagnose is noodsaaklik vir spoedige inisiasie van behandeling. Mikrobiologiese bevestiging is selfs meer problematies in stadium 1 TBM omdat lumbale punksie gewoonlik slegs gedoen word in kinders met kliniese tekens van meningisme (stadium 2 en 3 TBM). Met hierdie tesis, beoog ek om meganismes te ondersoek wat vroeë en/of meer akkurate diagnose van pediatriese TBM kan verbeter. DEEL I In hoofstuk 2 word die plaaslike diagnose en behandeling van kinder TBM beskryf wat in soortgelyke omgewings toegepas kan word. Hoogtepunte sluit in, eerstens, dat kort (6 maande) intensiewe terapie in kinders met middel-sensitiewe TBM veilig en doeltreffend is, met goeie uitkoms en ‘n lae sterftesyfer. Tweedens, tuis-gebaseerde TBM behandeling na aanvanklike binnepasient stabilisering is haalbaar in noukeurig geselekteerde pasiënte onder streng toesig. Derdens, die behandeling van tuberkuleuse hidrokefalus word bepaal deur die vlak van serebrospinale vog (SSV) obstruksie. Kommunikerende hidrokefalus kan suksesvol met mediese terapie behandel word, met normalisering van intrakraniale druk wat plaasvind binne dae in die meerderheid van kinders. Nie-kommunikerende hidrokefalus vereis neurochirurgiese intervensie. Vierdens, thalidomied is die plaaslike middel van keuse in kinders wat lewensgevaarlike TB massa letsels (IRIS) ontwikkel, ondanks kortikosteroïede. DEEL II In kliniese praktyk, word TBM dikwels gediagnoseer op grond van 'n kombinasie van kliniese, laboratorium en radiologiese bevindings. 'n Eenvormige gevalsdefinisie, gebaseer op die bogenoemde kriteria is voorgestel deur 'n internasionale paneel van kundiges om die diagnostiese standaard vir navorsingsdoeleindes te verbeter. Vermoedelike gevalle word gekategoriseer as definitiewe, waarskynlike, of moontlike TBM. Deel 2 van my tesis fokus op die diagnostiese nut van hierdie voorgestelde eenvormige gevalsdefinisie kriteria vir TBM. In hoofstuk 3.1 evalueer ek retrospektief die diagnostiese akkuraatheid van waarskynlike en moontlike TBM kriteria in kinders met kultuur-bevestigde TBM en kultuur-bevestigde bakteriële meningitis. Die voorgestelde gevalsdefinisie het uitstekend vergelyk met mikrobiologies-bevestigde TBM wanneer 'n 'waarskynlike' TBM telling gebruik is. Wanneer 'n ‘moontlike’ TBM telling gebruik is, soos bepaal deur die gevalsdefinisie, was daar nie 'n enkele geval van TBM gemis nie, maar kliniese nut was minimaal gegewe die lae spesifisiteit. Om my bevindinge te versterk het ek ook die diagnostiese akkuraatheid van die voorgestelde TBM gevalsdefinisie prospektiewelik ondersoek (sien hoofstuk 3.2). Die gevalsdefinisie het diagnosties uitstekend vergelyk met bakteriële en virale meningitis kontroles. Die hoë spesifisiteit van 'n ‘waarskynlike’ TBM telling soos verkry met behulp van die gevalsdefinisie regverdig die gebruik daarvan as 'n alternatiewe verwysingsstandaard tot mikrobiologiese bevestiging in toekomstige studies. In beide studies (retrospektief en prospektief) was swak sensitiwiteit verkry wanneer 'n ‘waarskynlike’ TBM telling gebruik was vir vroeë (stadium 1) TBM diagnose. Hierdie bevinding bevestig weereens die belang van ‘n hoe indeks van suspisie van TBM in jong kinders met onlangse blootstelling aan TB en persisterende non-spesifieke kliniese simptome en tekens. SSV bevindinge is noodsaaklik vir die vroeë diagnose van TBM, maar die bewyse vir afsnywaardes vir SSV glukose in TBM is gebrekkig. ‘n SSV proteïen afsnywaarde van >1g/L (100mg/dL) is wel bewys om te onderskei tussen gevalle van TBM en ander vorme van meningitis. My studie wat die diagnostiese waarde van SSV chemie uitslae in kinder TBM beskryf het bevind dat die optimale ondergrens van SSV glukose konsentrasie as 'n diagnostiese hulpmiddel vir TBM 2.2 mmol/L was (sien hoofstuk 3.3). Absolute SSV glukose konsentrasie onderskei nie-TBM van TBM gevalle met sensitiwiteit van 68% en spesifisiteit van 96%, wat die gebruik daarvan as 'n eliminasie toets uitsluit. Gelyktydige bepaling van serum en SSV glukose was selde uitgevoer, maar ek het bevind dat die SSV tot serum glucose verhouding diagnostiese sensitiwiteit kan verbeter. ‘n SSV proteïen afsnywaarde van >1g/L asook ander kenmerke van SSV soos makroskopiese voorkoms, seltelling en die seltipe (oorwegend limfosiete) is van waarde in die onderskeiding van TBM en bakteriële meningitis. Vorige studies het bevind dat 70% tot 84% van kinders met TBM borskas X-straal bevindinge het van aktiewe long TB. In my studie (hoofstuk 3.4), het slegs 46%, van kinders met TBM borskas X-straal bevindinge hoogs suggestief van long TB getoon. 'n Behoefte tot behandeling berekening (‘need to treat’) toon dat slegs 1 in 4.39 kinders ≤3 jaar oud met TBM waarskynlik 'definitiewe long TB' op borskas X-straal sal hê. DEEL III Mikrobiologiese bevestiging van TBM bly die goud-standaard van diagnose, maar stel hoë eise in jong kinders as gevolg van die lae basillêre tellings wat kenmerkend van TBM is, lae SSV volumes beskikbaar vir diagnostiese analise en sub-optimale sensitiwiteit van direkte mikroskopie vir suur-vaste basille en M.tuberculosis kultuur op SSV. Verskeie nuwe kommersieël beskikbare nukleiensuur amplifikasie (NSA) toetse is ontwikkel vir die spoedige diagnose van TB. In deel III van die tesis, het ‘n meta-analise van nuwe kommersiële NSA toetse ‘n opsommings sensitiwiteit van 64% en spesifisiteit van 98% bevind. Sensitiwiteit van kommersiële NSA toetse bly dus suboptimaal en daarom ‘n onwaarskynlike modaliteit vir verbetering van vroeë akkurate diagnose van TBM; die uitstekende spesifisiteit dui egter daarop dat ‘n positiewe kommersiële NSA toets as definitief van TBM beskou kan word in die korrekte kliniese omstandighede. In 2013 het die WGO Xpert MTB/RIF® as die voorkeur aanvanklike ondersoek aanbeveel in alle volwassenes en kinders met verdagte TBM. Vyf studies wat Xpert MTB/RIF® op SSV beskryf het opsommings sensitiwiteit van 70% en spesifisiteit van 97% bevind. In hoofstuk 4.2 stel ek my ten doel om die nut van MTBDRplus® en Xpert MTB/RIF®, onafhanklik en/of in kombinasie, vir die diagnose van TBM in 'n kliniese omgewing te evalueer. Die belangrikste bevinding van die studie was dat die gebruik van beide bogenoemde kommersiële NSA toetse op SSV die diagnostiese akkuraatheid verhoog het. Beide NSA toetse het beter sensitiwiteit getoon as SSV kultuur, maar was minder sensitief as ‘n streng gedefinieërde kliniese gevalsdefinisie. Die onderskeie sensitiwiteit en spesifisiteit was 33% en 98% vir die MTBDRplus® toets, 26% en 100% vir die Xpert MTB/RIF® en 26% en 100% wanneer die gekombineërde toetse vergelyk is met die TBM gevalsdefinisie. Dit is dus onvoldoende as 'n eliminasie toets en bied beperkte kliniese leiding, maar 'n positiewe toets bied vinnige mikrobiologiese bevestiging wat onnodige behandelingsvertraging en potensiële lewensgevaarlike nagevolge voorkom. ‘n Addisionele voordeel van ‘n positiewe NSA toets is die vroeë herkenning van mikrobakteriele weerstandigheid. 'n Beperking van hierdie tesis is die feit dat diagnose van stadium 1 TBM in kinders nie verbeter is nie. ‘n Hoë indeks van suspisie, veral in kinders met swak gewigstoename (of gewigsverlies ) en kinders wat vir langer as 2 weke hoes, kan die diagnose van kinder TB en vroeë TBM verbeter. IMCI, wat op primêre sorg beoefen word, is potensieël 'n waardevolle hulpmiddel om gesondheidswerkers meer bewus te maak van TBM, en dus die diagnose van vroeë TB en TBM te verbeter. Die opsporing van huishoudelike TB kontakte sowel as profilakse met isoniasied, en meer algemene maatreëls soos die verbetering van behuising en die verligting van armoede, moet toenemend as 'n voorkomende strategie teen TBM aangewend word.National Research Foundatio