Kassirian S, et al.
Pleural fluid cytology to diagnose malignancy is one of the cornerstones of pleural diagnostics. Which is why it rather surprising that this is the first meta-analysis on the topic. Published in Thorax the pooled results from 36 studies (6057 patients) were analysed.
The overall sensitivity of fluid cytology for malignant effusions was 58%. However there was considerable heterogenicity between studies, much explained by the variation in primary cancers within each cohort. Those studies with a greater proportion of mesothelioma tended to record lower overall sensitivity. The malignancies with the highest sensitivities were ovarian (85%), lung adenocarcinoma (84%) and breast (65%).
Six studies had also looked at the value of repeat sampling after an initial negative result with sensitivities of 22% for the 1st repeat and 14% for the 2nd. It should be noted that the retrospective nature of these studies means that the samples were probably repeated in patients where more invasive diagnostics were inappropriate.
Kidd AC, et al.
A very interesting study from the Glasgow group looking to address the difficulties with monitoring mesothelioma through artificial intelligence (AI). Trials, and to a lesser extent clinical practice, have relied on modifiedRECIST to assess response to treatment which has numerous flaws as laid out by this article.
Using a training set of 123 CTs from patients with mesothelioma a fully automated Convolutional Neural Network (CNN) was trained through carefully human labelled scans. The performance of the CNN was subsequently tested on an independent dataset of 60 scans to prevent overfitting.
The human assessed tumour volumes and AI assessed were strongly correlated in the validation set (r=0.85), and not well correlated with mRECIST. The AI struggled in some areas such as fissural or contralateral tumour or adjacent atelectasis. Importantly, there was a trend towards better outcome prediction than mRECIST.
The authors state that the next step in this important journey is further calibration in larger datasets before AI can be rolled out in clinical trials or clinical care.
van Kooten JP, et al
Another pleural paper in Thorax this month. This epidemiological study from the Netherlands has assessed the diagnoses of mesothelioma (pleural and peritoneal) from 1993 (when asbestos was banned) to 2018. This is courtesy of the impressive Dutch cancer registry.
They have shown that mesothelioma cases peaked in 2010 and have been falling since. This is notable as the peak was predicted to be 2020 in the Netherlands which might be related to incomplete reduction in asbestos use in the 1970s use prior to its outright ban.
Unsurprisingly the authors note that despite advances in bother chemotherapeutics and immunotherapies the “prognosis is still dismal”.
Janowak CF, et al
Whilst the evidence base for combined intrapleural fibrinolytics for pleural infection has grown considerably in the last two decades, there remains little on their role in haemothorax.
tPA activates plasminogen to plasmin thereby degrading the thrombin within the haemothorax. There are several series and trials documenting its effective use. The utility of the addition of DNase is less well explored and has a weaker biochemical logic.
In this retrospective single centre cohort study the authors compare the outcomes of patients who recived tPA alone (n=28) vs those who received tPA and DNase (n=22) over a 6.5yr period. There was no clear indication as to why practice varied within the centre.
From 1140 patients admitted with a traumatic haemothorax, 638 (56%) had a chest tube inserted, of which just 50 had fibrinolytics (28 tPA vs 22 tPA/DNase). There were no significant differences in outcomes between the two groups and no evidence that DNase is beneficial in the setting of haemothorax.