The level of reporting of the interventions was good. The relevance of the interventions included appeared to be appropriate as both were recommended strategies in the UK NHS setting.
The sources of effectiveness data were well reported. The study setting appeared to be generalisable to the wider UK setting. It is important to note that, as highlighted in by the authors the assumption of 100% accuracy for mediastinoscopy would not hold in clinical practice. The clinical characteristics of both the prospective and historical cohort were presented; there were some differences in baseline characteristics for age: historical cohort had three patients <30 years (4%); prospective cohort had 15 patients <30 years (19%); historical cohort median age was 53 years (range 25 to 85); prospective cohort median age was 42 years (range 17 to 79). The impact of these differences is unknown, but likely to be negligible given that the model focused on diagnosis and did not include treatment or health related quality-of-life outcomes.
The cost data were adequately reported. Costs relevant to the UK NHS perspective were included; justification for excluded costs was discussed. The authors made an assumption that treatment and treatment costs were unaffected by the method of diagnosis; embedded within that was an assumption that there were no complications associated with surgical biopsy. This was unlikely to hold true in clinical practice, and the costs and health disutility of any adverse effects may impact on the results. However, this assumption was conservative for endoscopic biopsy and the inclusion of adverse events would make the intervention more cost-effective. The price year was not stated explicitly, but the authors conducted the study from 2009 to 2011 and used NHS tariffs for 2010 to 2011.
Analysis and results:
The cost-minimisation approach stated by the authors appeared appropriate, regardless of diagnosis method, treatment for the condition would remain the same. The level of reporting of results was good. The authors discussed some key limitations of their analysis, which included some of the underlying assumptions around equality of tests, conduct of the study in a tertiary centre, potential generalisability issues outside of the UK. The impact of uncertainty was addressed, but more comprehensive methods of assessing parameter uncertainty could have been used.
There were a few limitations to the study, which were discussed by the authors. The results should be considered with these limitations in mind.