Gender differences in tuberculosis treatment outcomes: a post hoc analysis of the REMoxTB study.

BACKGROUND: In the REMoxTB study of 4-month treatment-shortening regimens containing moxifloxacin compared to the standard 6-month regimen for tuberculosis, the proportion of unfavourable outcomes for women was similar in all study arms, but men had more frequent unfavourable outcomes (bacteriologically or clinically defined failure or relapse within 18 months after randomisation) on the shortened moxifloxacin-containing regimens. The reason for this gender disparity in treatment outcome is poorly understood. METHODS: The gender differences in baseline variables were calculated, as was time to smear and culture conversion and Kaplan-Meier plots were constructed. In post hoc exploratory analyses, multivariable logistic regression modelling and an observed case analysis were used to explore factors associated with both gender and unfavourable treatment outcome. RESULTS: The per-protocol population included 472/1548 (30%) women. Women were younger and had lower rates of cavitation, smoking and weight (all p < 0.05) and higher prevalence of HIV (10% vs 6%, p = 0.001). They received higher doses (mg/kg) than men of rifampicin, isoniazid, pyrazinamide and moxifloxacin (p ≤ 0.005). There was no difference in baseline smear grading or mycobacterial growth indicator tube (MGIT) time to positivity. Women converted to negative cultures more quickly than men on Lowenstein-Jensen (HR 1.14, p = 0.008) and MGIT media (HR 1.19, p < 0.001). In men, the presence of cavitation, positive HIV status, higher age, lower BMI and 'ever smoked' were independently associated with unfavourable treatment outcome. In women, only 'ever smoked' was independently associated with unfavourable treatment outcome. Only for cavitation was there a gender difference in treatment outcomes by regimen; their outcome in the 4-month arms was significantly poorer compared to the 6-month treatment arm (p < 0.001). Women, with or without cavities, and men without cavities had a similar outcome on all treatment arms (p = 0.218, 0.224 and 0.689 respectively). For all other covariate subgroups, there were no differences in treatment effects for men or women. CONCLUSIONS: Gender differences in TB treatment responses for the shorter regimens in the REMoxTB study may be explained by poor outcomes in men with cavitation on the moxifloxacin-containing regimens. We observed that women with cavities, or without, on the 4-month moxifloxacin regimens had similar outcomes to all patients on the standard 6-month treatment. The biological reasons for this difference are poorly understood and require further exploration.

Pretreatment chest x-ray severity and its relation to bacterial burden in smear positive pulmonary tuberculosis.

BACKGROUND: Chest radiographs are used for diagnosis and severity assessment in tuberculosis (TB). The extent of disease as determined by smear grade and cavitation as a binary measure can predict 2-month smear results, but little has been done to determine whether radiological severity reflects the bacterial burden at diagnosis. METHODS: Pre-treatment chest x-rays from 1837 participants with smear-positive pulmonary TB enrolled into the REMoxTB trial (Gillespie et al., N Engl J Med 371:1577-87, 2014) were retrospectively reviewed. Two clinicians blinded to clinical details using the Ralph scoring system performed separate readings. An independent reader reviewed discrepant results for quality assessment and cavity presence. Cavitation presence was plotted against time to positivity (TTP) of sputum liquid cultures (MGIT 960). The Wilcoxon rank sum test was performed to calculate the difference in average TTP for these groups. The average lung field affected was compared to log 10 TTP by linear regression. Baseline markers of disease severity and patient characteristics were added in univariable regression analysis against radiological severity and a multivariable regression model was created to explore their relationship. RESULTS: For 1354 participants, the median TTP was 117 h (4.88 days), being 26 h longer (95% CI 16-30, p < 0.001) in patients without cavitation compared to those with cavitation. The median percentage of lung-field affected was 18.1% (IQR 11.3-28.8%). For every 10-fold increase in TTP, the area of lung field affected decreased by 11.4%. Multivariable models showed that serum albumin decreased significantly as the percentage of lung field area increased in both those with and without cavitation. In addition, BMI and logged TTP had a small but significant effect in those with cavitation and the number of severe TB symptoms in the non-cavitation group also had a small effect, whilst other factors found to be significant on univariable analysis lost this effect in the model. CONCLUSIONS: The radiological severity of disease on chest x-ray prior to treatment in smear positive pulmonary TB patients is weakly associated with the bacterial burden. When compared against other variables at diagnosis, this effect is lost in those without cavitation. Radiological severity does reflect the overall disease severity in smear positive pulmonary TB, but we suggest that clinicians should be cautious in over-interpreting the significance of radiological disease extent at diagnosis.

Liver toxicity associated with tuberculosis chemotherapy in the REMoxTB study.

BACKGROUND: Drug-induced liver injury (DILI) is a common complication of tuberculosis treatment. We utilised data from the REMoxTB clinical trial to describe the incidence of predisposing factors and the natural history in patients with liver enzyme levels elevated in response to tuberculosis treatment. METHODS: Patients received either standard tuberculosis treatment (2EHRZ/4HR), or a 4-month regimen in which moxifloxacin replaced either ethambutol (isoniazid arm, 2MHRZ/2MHR) or isoniazid (ethambutol arm, 2EMRZ/2MR). Hepatic enzymes were measured at 0, 2, 4, 8, 12 and 17 weeks and as clinically indicated during reported adverse events. Patients included were those receiving at least one dose of drug and with two or more hepatic enzyme measurements. RESULTS: A total of 1928 patients were included (639 2EHRZ/4HR, 654 2MHRZ/2MHR and 635 2EMRZ/2MR). DILI was defined as peak alanine aminotransferase (ALT) ≥ 5 times the upper limit of normal (5 × ULN) or ALT ≥ 3 × ULN with total bilirubin > 2 × ULN. DILI was identified in 58 of the 1928 (3.0%) patients at a median time of 28 days (interquartile range IQR 14-56). Of 639 (6.4%) patients taking standard tuberculosis therapy, 41 experienced clinically significant enzyme elevations (peak ALT ≥ 3 × ULN). On standard therapy, 21.1% of patients aged >55 years developed a peak ALT/aspartate aminotransferase (AST) ≥ 3 × ULN (p = 0.01) and 15% of HIV-positive patients experienced a peak ALT/AST ≥ 3 × ULN compared to 9% of HIV-negative patients (p = 0.160). The median peak ALT/AST was higher in isoniazid-containing regimens vs no-isoniazid regimens (p < 0.05), and lower in moxifloxacin-containing arms vs no-moxifloxacin arms (p < 0.05). Patients receiving isoniazid reached a peak ALT ≥ 3 × ULN 9.5 days earlier than those on the ethambutol arm (median time of 28 days vs 18.5 days). Of the 67 Asian patients with a peak ALT/AST ≥ 3 × ULN, 57 (85.1%) were on an isoniazid-containing regimen (p = 0.008). CONCLUSIONS: Our results provide evidence of the risk of DILI in tuberculosis patients on standard treatment. Older patients on standard therapy, HIV-positive patients, Asian patients and those receiving isoniazid were at higher risk of elevated enzyme levels. Monitoring hepatic enzymes during the first 2 months of standard therapy detected approximately 75% of patients with a peak enzyme elevation ≥3 × ULN, suggesting this should be a standard of care. These results provide evidence for the potential of moxifloxacin in hepatic sparing.

Pharmacokinetics, Tolerability, and Bacteriological Response of Rifampin Administered at 600, 900, and 1,200 Milligrams Daily in Patients with Pulmonary Tuberculosis.

In a multiple-dose-ranging trial, we previously evaluated higher doses of rifampin in patients for 2 weeks. The objectives of the current study were to administer higher doses of rifampin for a longer period to compare the pharmacokinetics, safety/tolerability, and bacteriological activity of such regimens. In a double-blind, randomized, placebo-controlled, phase II clinical trial, 150 Tanzanian patients with tuberculosis (TB) were randomized to receive either 600 mg (approximately 10 mg/kg of body weight), 900 mg, or 1,200 mg rifampin combined with standard doses of isoniazid, pyrazinamide, and ethambutol administered daily for 2 months. Intensive pharmacokinetic sampling occurred in 63 patients after 6 weeks of treatment, and safety/tolerability was assessed. The bacteriological response was assessed by culture conversion in liquid and solid media. Geometric mean total exposures (area under the concentration-versus-time curve up to 24 h after the dose) were 24.6, 50.8, and 76.1 mg · h/liter in the 600-mg, 900-mg, and 1,200-mg groups, respectively, reflecting a nonlinear increase in exposure with the dose (P < 0.001). Grade 3 adverse events occurred in only 2 patients in the 600-mg arm, 4 patients in the 900-mg arm, and 5 patients in the 1,200-mg arm. No significant differences in the bacteriological response were observed. Higher daily doses of rifampin (900 and 1,200 mg) resulted in a more than proportional increase in rifampin exposure in plasma and were safe and well tolerated when combined with other first-line anti-TB drugs for 2 months, but they did not result in improved bacteriological responses in patients with pulmonary TB. These findings have warranted evaluation of even higher doses of rifampin in follow-up trials. (This study has been registered at ClinicalTrials.gov under identifier NCT00760149.).

When is an outbreak not an outbreak? Fit, divergent strains of Mycobacterium tuberculosis display independent evolution of drug resistance in a large London outbreak.

OBJECTIVES: To study the evolutionary relationship of Mycobacterium tuberculosis isolates from 13 patients in a large outbreak of isoniazid-resistant tuberculosis in London. METHODS: Genotypic and phenotypic susceptibility tests were performed. Molecular genotyping using restriction fragment length polymorphisms and mycobacterial interspersed repetitive units was carried out. Additionally, the generation times of 13 strains of M. tuberculosis from the outbreak were measured to determine relative fitness. RESULTS: Genotypic and phenotypic susceptibility testing demonstrated variations between isolates. Polymorphisms causing isoniazid resistance varied within clusters of isolates that were indistinguishable by standard genotyping. The measurement of in vitro generation times demonstrated that the fitness of the resistant strains was not significantly different from either wild-type or susceptible isolates in the outbreak, indicating that apparently no fitness cost was associated with the acquisition of drug resistance. CONCLUSIONS: It appears that this outbreak comprised a heterogeneous collection of closely related strains, which appear to exhibit more variation than would usually be associated with a point source outbreak. These strains appear to have evolved by acquisition of additional antimicrobial resistance mutations while remaining competitive. The acquired resistance and retained competitiveness may be partly responsible for the difficulty in controlling the outbreak.

Clinical perspectives in integrating whole-genome sequencing into the investigation of healthcare and public health outbreaks - hype or help?

Outbreaks pose a significant risk to patient safety as well as being costly and time consuming to investigate. The implementation of targeted infection prevention and control measures relies on infection prevention and control teams having access to rapid results that detect resistance accurately, and typing results that give clinically useful information on the relatedness of isolates. At present, determining whether transmission has occurred can be a major challenge. Conventional typing results do not always have sufficient granularity or robustness to define strains unequivocally, and sufficient epidemiological data are not always available to establish links between patients and the environment. Whole-genome sequencing (WGS) has emerged as the ultimate genotyping tool, but has not yet fully crossed the divide between research method and routine clinical diagnostic microbiological technique. A clinical WGS service was officially established in 2014 as part of the Scottish Healthcare Associated Infection Prevention Institute to confirm or refute outbreaks in hospital settings from across Scotland. This article describes the authors' experiences with the aim of providing new insights into practical application of the use of WGS to investigate healthcare and public health outbreaks. Solutions to overcome barriers to implementation of this technology in a clinical environment are proposed.

A partially randomised trial of pretomanid, moxifloxacin and pyrazinamide for pulmonary TB.

BACKGROUND: Treatment for TB is lengthy and toxic, and new regimens are needed.METHODS: Participants with pulmonary drug-susceptible TB (DS-TB) were randomised to receive: 200 mg pretomanid (Pa, PMD) daily, 400 mg moxifloxacin (M) and 1500 mg pyrazinamide (Z) for 6 months (6Pa200MZ) or 4 months (4Pa200MZ); 100 mg pretomanid daily for 4 months in the same combination (4Pa100MZ); or standard DS-TB treatment for 6 months. The primary outcome was treatment failure or relapse at 12 months post-randomisation. The non-inferiority margin for between-group differences was 12.0%. Recruitment was paused following three deaths and not resumed.RESULTS: Respectively 4/47 (8.5%), 11/57 (19.3%), 14/52 (26.9%) and 1/53 (1.9%) DS-TB outcomes were unfavourable in patients on 6Pa200MZ, 4Pa200MZ, 4Pa100MZ and controls. There was a 6.6% (95% CI -2.2% to 15.4%) difference per protocol and 9.9% (95%CI -4.1% to 23.9%) modified intention-to-treat difference in unfavourable responses between the control and 6Pa200MZ arms. Grade 3+ adverse events affected 68/203 (33.5%) receiving experimental regimens, and 19/68 (27.9%) on control. Ten of 203 (4.9%) participants on experimental arms and 2/68 (2.9%) controls died.CONCLUSION: PaMZ regimens did not achieve non-inferiority in this under-powered trial. An ongoing evaluation of PMD remains a priority.

How mechanistic in silico modelling can improve our understanding of TB disease and treatment.

TB is one of the top 10 causes of death worldwide and the leading cause of death from a single infectious agent. Decreasing the length of time for TB treatment is an important step towards the goal of reducing mortality. Mechanistic in silico modelling can provide us with the tools to explore gaps in our knowledge, with the opportunity to model the complicated within-host dynamics of the infection, and simulate new treatment strategies. Significant insight has been gained using this form of modelling when applied to other diseases - much can be learned in infection research from these advances.

Audit of laboratory diagnostic methods for syphilis in England and Wales.

OBJECTIVES: The number of cases of infectious syphilis is increasing rapidly across England and Wales. Concern has been expressed about diagnostic delay and its potential impact on patient care. A standard operating procedure for the serological diagnosis of syphilis has recently been developed by the Health Protection Agency. This study aimed to audit clinical and laboratory practice in England and Wales against this standard. METHODS: All microbiology departments, genitourinary medicine (GUM) clinics and antenatal clinics in England and Wales were invited to complete a web-based questionnaire. RESULTS: The overall response rate was 76%. Practices varied between laboratories. The proportion of microbiology departments performing enzyme immunoassay (EIA), Treponema pallidum particle agglutination assay/T pallidum haemagglutination assay, rapid plasma reagin/Venereal Disease Reference Laboratory and EIA IgM were 94%, 34%, 41% and 10%, respectively. Of these, 57% only perform a single screening assay. The turnaround time for negative results was less than 1 week for 84% of microbiology departments. For positive samples, turnaround times varied from less than 1 week to 6-8 weeks, with 19% of GUM clinics reporting turnaround times of over 3 weeks. Notably, 26% of GUM clinics and 6% of antenatal clinics reported that delays in turnaround time had adversely affected patient management in the past year. CONCLUSION: This study suggests that there is significant room to improve laboratory turnaround times for the diagnosis of syphilis in England and Wales, and such improvements would be a positive step in limiting the spread of infection and of congenital syphilis.

Think TB! Is the diagnosis of pulmonary tuberculosis delayed by the use of antibiotics?

SETTING: Effective tuberculosis (TB) control requires prompt diagnosis of infectious cases through early suspicion of pulmonary TB in all subjects with suspected respiratory infection. OBJECTIVE: To test our hypothesis that prior antibiotic treatment for presumed bacterial infection leads to a delay in diagnosing TB in a European country with low TB incidence. DESIGN: Adults with culture-confirmed pulmonary TB at a single metropolitan centre were assessed for the impact of any previous antibiotic treatment on symptoms and the time to starting specific anti-tuberculosis treatment. RESULTS: Of 83 patients, 42 (51%) received antibiotics prior to TB diagnosis, with symptomatic improvement reported in 20 of the 42 (48%) patients. This was unrelated to specific drug class. Although the median time to diagnosis in subjects receiving antibiotics was prolonged (P=0.001), this was not predicted by treatment response. In 94% of cases, the initial chest radiograph was suggestive of TB infection. CONCLUSION: Patients receiving antibiotics prior to TB confirmation experience a process-related delay in starting treatment. To minimise the risk of ongoing TB transmission, we propose that clinicians should include TB in their differential diagnosis and initiate simple, TB-focused investigations early on in the diagnostic process.

Clinical application of a rapid lung-orientated immunoassay in individuals with possible tuberculosis.

BACKGROUND: Immunological ex vivo assays to diagnose tuberculosis (TB) have great potential but have largely been blood-based and poorly evaluated in active TB. Lung sampling enables combined microbiological and immunological testing and uses higher frequency antigen-specific responses than in blood. METHODS: A prospective evaluation was undertaken of a flow cytometric assay measuring the percentage of interferon-gamma synthetic CD4+ lymphocytes following stimulation with purified protein derivative of Mycobacterium tuberculosis (PPD) in bronchoalveolar lavage fluid from 250 sputum smear-negative individuals with possible TB. A positive assay was defined as >1.5%. RESULTS: Of those who underwent lavage and were diagnosed with active TB, 95% (106/111) had a positive immunoassay (95% CI 89% to 98%). In 139 individuals deemed not to have active TB, 105 (76%) were immunoassay negative (95% CI 68% to 82%). Of the remaining 24% (34 cases) with a positive immunoassay, a substantial proportion had evidence of untreated TB; in two of these active TB was subsequently diagnosed. Assay performance was unaffected by HIV status, disease site or BCG vaccination. In culture-positive pulmonary cases, response to PPD was more sensitive than nucleic acid amplification testing (94% vs 73%). The use of early secretory antigen target-6 (ESAT-6) responses in 71 subjects was no better than PPD, and 19% of those with culture-confirmed TB and a positive PPD immunoassay had no detectable response to ESAT-6. CONCLUSIONS: These findings suggest that lung-orientated immunological investigation is a potentially powerful tool in diagnosing individuals with sputum smear-negative active TB, regardless of HIV serostatus.

The effect of oxidative stress on the mutation rate of Mycobacterium tuberculosis with impaired catalase/peroxidase function.

OBJECTIVES: To determine the effect of oxidative stress on isoniazid-resistant Mycobacterium tuberculosis deficient in catalase/peroxidase activity to varying degrees through mutation in katG. METHODS: The mutation rate was determined for a set of isogenic strains with different katG alleles giving different catalase and/or peroxidase activities following exposure to the oxidizing agent, hydrogen peroxide. Mutants were selected on rifampicin, and the location and nature of the mutation were identified by sequencing the rpoB gene. RESULTS: No evidence was found to suggest that strains that had impaired catalase/peroxidase activity were hypermutable, and the presence of excess hydrogen peroxide had no effect on the mutation rate. An unusual pattern of mutations in rpoB was observed in catalase-deficient strains with only 3 of 66 having mutations within the rifampicin resistance-determining region. CONCLUSIONS: The mutation rate of M. tuberculosis in response to oxidative stress is not increased in strains with significant deficits in catalase and peroxidase activity. Our data suggest that isoniazid-resistant strains compensate for their reduced ability to detoxify oxidative stress effectively. Interestingly, mutations were found in unusual locations at positions similar to those found in clinical isoniazid-resistant strains.

Mycobacterium tuberculosis DNA repair in response to subinhibitory concentrations of ciprofloxacin.

OBJECTIVES: To investigate how the SOS response, an error-prone DNA repair pathway, is expressed following subinhibitory quinolone treatment of Mycobacterium tuberculosis. METHODS: Genome-wide expression profiling followed by quantitative RT (qRT)-PCR was used to study the effect of ciprofloxacin on M. tuberculosis gene expression. RESULTS: Microarray analysis showed that 16/110 genes involved in DNA protection, repair and recombination were up-regulated. There appeared to be a lack of downstream genes involved in the SOS response. qRT-PCR detected an induction of lexA and recA after 4 h and of dnaE2 after 24 h of subinhibitory treatment. CONCLUSIONS: The pattern of gene expression observed following subinhibitory quinolone treatment differed from that induced after other DNA-damaging agents (e.g. mitomycin C). The expression of the DnaE2 polymerase response was significantly delayed following subinhibitory quinolone exposure.

Pseudomonas aeruginosa intensive care unit outbreak: winnowing of transmissions with molecular and genomic typing.

BACKGROUND: Pseudomonas aeruginosa healthcare outbreaks can be time consuming and difficult to investigate. Guidance does not specify which typing technique is most practical for decision-making. AIM: To explore the usefulness of whole-genome sequencing (WGS) in the investigation of a P. aeruginosa outbreak, describing how it compares with pulsed-field gel electrophoresis (PFGE) and variable number tandem repeat (VNTR) analysis. METHODS: Six patient isolates and six environmental samples from an intensive care unit (ICU) positive for P. aeruginosa over two years underwent VNTR, PFGE and WGS. FINDINGS: VNTR and PFGE were required to fully determine the potential source of infection and rule out others. WGS results unambiguously distinguished linked isolates, giving greater assurance of the transmission route between wash-hand basin water and two patients, supporting the control measures employed. CONCLUSION: WGS provided detailed information without the need for further typing. When allied to epidemiological information, WGS can be used to understand outbreak situations rapidly and with certainty. Implementation of WGS in real-time would be a major advance in day-to-day practice. It could become a standard of care as it becomes more widespread due to its reproducibility and lower costs.

A paradigm for the molecular identification of Mycobacterium species in a routine diagnostic laboratory.

The aim of this study was to improve the identification of Mycobacterium species in the context of a UK teaching hospital. Real-time PCR assays were established to enable the rapid differentiation between Mycobacterium tuberculosis (MTB) complex and Mycobacterium species other than tuberculosis (MOTT), followed by 16S rRNA gene sequencing for the speciation of MOTT. Real-time PCR assays gave comparable results to those from the reference laboratory. The implementation of these PCR assays using an improved bead extraction method has enhanced the mycobacterial diagnostic service at the Royal Free Hospital by providing a rapid means of differentiating between MTB complex and MOTT, and would be simple to implement in similar laboratories. Sequence analysis successfully identified a range of Mycobacterium spp. representative of those encountered in the clinical setting of the authors, including Mycobacterium avium complex, Mycobacterium fortuitum group, Mycobacterium chelonae-Mycobacterium abscessus group, Mycobacterium xenopi and Mycobacterium gordonae. It provides a useful tool for the identification of MOTT when clinically indicated.

Defining dormancy in mycobacterial disease.

Tuberculosis remains a threat to global health and recent attempts to shorten therapy have not succeeded mainly due to cases of clinical relapse. This has focussed attention on the importance of "dormancy" in tuberculosis. There are a number of different definitions of the term and a similar multiplicity of different in vitro and in vivo models. The danger with this is the implicit assumption of equivalence between the terms and models, which will make even more difficult to unravel this complex conundrum. In this review we summarise the main models and definitions and their impact on susceptibility of Mycobacterium tuberculosis. We also suggest a potential nomenclature for debate. Dormancy researchers agree that factors underpinning this phenomenon are complex and nuanced. If we are to make progress we must agree the terms to be used and be consistent in using them.

Optimising molecular diagnostic capacity for effective control of tuberculosis in high-burden settings.

The World Health Organization's 2035 vision is to reduce tuberculosis (TB) associated mortality by 95%. While low-burden, well-equipped industrialised economies can expect to see this goal achieved, it is challenging in the low- and middle-income countries that bear the highest burden of TB. Inadequate diagnosis leads to inappropriate treatment and poor clinical outcomes. The roll-out of the Xpert(®) MTB/RIF assay has demonstrated that molecular diagnostics can produce rapid diagnosis and treatment initiation. Strong molecular services are still limited to regional or national centres. The delay in implementation is due partly to resources, and partly to the suggestion that such techniques are too challenging for widespread implementation. We have successfully implemented a molecular tool for rapid monitoring of patient treatment response to anti-tuberculosis treatment in three high TB burden countries in Africa. We discuss here the challenges facing TB diagnosis and treatment monitoring, and draw from our experience in establishing molecular treatment monitoring platforms to provide practical insights into successful optimisation of molecular diagnostic capacity in resource-constrained, high TB burden settings. We recommend a holistic health system-wide approach for molecular diagnostic capacity development, addressing human resource training, institutional capacity development, streamlined procurement systems, and engagement with the public, policy makers and implementers of TB control programmes.

Co-trimoxazole as prophylaxis against opportunistic infections in HIV-infected Zambian children (CHAP): a double-blind randomised placebo-controlled trial.

BACKGROUND: No trials of co-trimoxazole (trimethoprim-sulfamethoxazole) prophylaxis for HIV-infected adults or children have been done in areas with high levels of bacterial resistance to this antibiotic. We aimed to assess the efficacy of daily co-trimoxazole in such an area. METHODS: We did a double-blind randomised placebo-controlled trial in children aged 1-14 years with clinical features of HIV infection in Zambia. Primary outcomes were mortality and adverse events possibly related to treatment. Analysis was by intention to treat. FINDINGS: In October, 2003, the data and safety monitoring committee recommended early stopping of the trial. 541 children had been randomly assigned; seven were subsequently identified as HIV negative and excluded. After median follow-up of 19 months, 74 (28%) children in the co-trimoxazole group and 112 (42%) in the placebo group had died (hazard ratio [HR] 0.57 [95% CI 0.43-0.77], p=0.0002). This benefit applied in children followed up beyond 12 months (n=320, HR 0.48 [0.27-0.84], test for heterogeneity p=0.60) and across all ages (test for heterogeneity p=0.82) and baseline CD4 counts (test for heterogeneity p=0.36). 16 (6%) children in the co-trimoxazole group had grade 3 or 4 adverse events compared with 18 (7%) in the placebo group. These events included rash (one placebo), and a neutrophil count on one occasion less than 0.5x10(9)/L (16 [6%] co-trimoxazole vs seven [3%] placebo, p=0.06). Pneumocystis carinii was identified by immunofluorescence in only one (placebo) of 73 nasopharyngeal aspirates from children with pneumonia. INTERPRETATION: Our results suggest that children of all ages with clinical features of HIV infection should receive co-trimoxazole prophylaxis in resource-poor settings, irrespective of local resistance to this drug.