214 Chapter 9 Sub-phenotype E was associated with community-acquired SAB, younger age, injection drug use, liver disease and endocarditis. In the SAFO trial, a 4-class model was considered best. Injection drug users were excluded from this study, and as expected, sub-phenotype E was not identified in this cohort. The four classes identified in the LCA using the SAFO data were similar to sub-phenotypes A-D in the Edinburgh and ARREST cohort. Sub-phenotypes A was associated with the highest 84-day mortality, and sub-phenotypes B and E with the lowest 84-day mortality. In a post-hoc analysis of the effect if adjunctive rifampicin, stratified by SAB subphenotype, patients assigned to sub-phenotype B and randomised to rifampicin had a higher 84-day mortality compared to patients randomised to placebo (odds ratio (OR) 18.8, 95% confidence interval 1.1-334.4, p=0.04). Patients in sub-phenotype C randomised to rifampicin were less likely to have microbiological failure (OR 0.17, 95% CI 0.04-0.80, p=0.02. In conclusion, the clinical heterogeneity of patients with SAB can be rationalised to identify clinically relevant subgroups, who have differential outcomes and potentially differential treatment responses. Utilising sub-phenotypes for patient stratification in clinical trials may provide information on subgroup specific treatment effects, and improve clinical outcomes through a personalised medicine approach. In the third part of this thesis, the aim was to facilitate causal inference. While mortality is a relevant clinical endpoint, the sample size needed to detect a treatment effect can be considerable. In the case of outbreaks of viral respiratory disease, the demand for fast results is even more critical compared to non-outbreak scenarios. The development and evaluation of S/F94 in Chapter 7 showed that this new outcome measure had a strong relation with the definitive outcome (mortality), but required a smaller sample size, also compared to other commonly used outcome measures in COVID-19 research. S/F94 as an outcome measure could improve causal inference by providing a better assessment of the definitive outcome compared to other commonly used outcome measures, such as the WHO ordinal scale. By improving the assessment of the outcome of interest, the required sample size to identify the treatment effect also decreases. There are many advantages if randomised controlled trials can be performed with a smaller sample size. First, results are available faster, as there is a shorter recruitment and data collection period. This is especially relevant during pandemic outbreaks. Second, the faster availability of results means the information can be used to inform medical decision and public health policy sooner, and benefit more patients. Moreover, there are fewer costs. Smaller trials require fewer resources, like staff and funding. This also makes conducting an RCT more accessible to organisations with limited budgets. Finally, a smaller sample size means that fewer patients are exposed to the potential risks associated with new treatments or interventions and the lifestyle regulations and additional measurements that are part of many clinical trials.
RkJQdWJsaXNoZXIy MTk4NDMw