In conducting this study, the authors aimed to combine information regarding the most common causes of pleural effusion and thoracic radiography to assist veterinarians with decision making in the instance of a respiratory emergency. The present study is the first to examine the positive and negative predictive values of radiographic signs in cats with pleural effusions. In our population of cats with pleural effusion, over half were caused by congestive heart failure, and the other half comprised of neoplasia, pyothorax, idiopathic chylous effusion, other causes. This is consistent with recent studies2,3,16 and in contrast to previous retrospective studies17–19 which identified that pleural effusions were more likely to be caused by neoplasia, pyothorax or FIP.
Radiographic assessment of the thorax with pleural effusion can be extremely challenging, due to obscuring of structures, including the cardiac silhouette and pulmonary parenchyma. The varied intra-observer and inter-observer measurements in assessing specific radiographic traits suggest radiographic assessment is difficult even for experienced radiologists. Of the radiographic variables assessed, we found that, by recognising an enlarged cardiac silhouette and presence of bilateral effusion on thoracic radiographs, clinicians should be strongly suspicious of CHF. A caveat to finding is that a small proportion of cats in congestive heart failure do not present with radiographic cardiomegaly (due to concentric hypertrophy being the most common cause of congestive heart failure 10,20). Therefore, echocardiography or other tests such as NT-proBNP are still recommended as an initial diagnostic test when pleural effusion is identified. Cardiac biomarkers such as NT-proBNP can assist in differentiating cardiac versus non-cardiac disease through measurement of plasma21 or pleural effusion levels22. NT-proBNP is useful when comprehensive echocardiography is not available, or when there may be concurrent disease (i.e., cardiac disease and neoplasia) or for monitoring of cardiac disease progression.
There was minimal added benefit in assessing the pulmonary vasculature, as radiographic assessment of these structures was poorly reliable without significant improvement on PPV and NPV than cardiomegaly alone. The poor reliability is likely due to differences in subjective assessment of vessel size and also the difficulty of assessing pulmonary vessels in the presence of pleural effusion.
There were some discrepancies in assessing bilateral versus unilateral effusions by the two radiologists. On review of the 17 cases with discrepancies in fluid distribution, majority of the effusions (13/17) had trivial or a small volume of effusions in both hemithoraces. A trivial or small volume of effusion can be difficult to assess on radiographs and may be open to differences in subjective interpretation. The remainder of the 5/17 cases had unequal distributions of effusion, in one hemithorax there was a moderate to large volume of effusion, whereas the other hemithorax was had a trivial effusion. This slight difference in opinion and assessment of unilateral versus bilateral effusion likely the reason for the large disparity in positive predictive values of fluid distribution for pyothorax among the two radiologists (radiologist 1 PPV 75%, radiologist 2 PPV 20%). The difference in interpretation of bilateral and unilateral effusion suggests that the assessment of fluid distribution when there is a small or trivial effusion may not be a reliable criterion. In retrospect, the addition of a third observer may be beneficial to reduce these discrepancies.
In the literature, there are varying reports1–4, 7, on whether neoplastic, chylous and pyothorax effusions cause unilateral or bilateral effusions. It is postulated that the mediastinum is fenestrated in cats 4; however, an excessive fibrin may seal the mediastinum (i.e. rupture of abscess, unilateral pleuritis causing adhesions etc). Previous reports 4,23 describe chylothorax and pyothorax to have unilateral effusions due to the highly inflammatory effusion. Other studies by Barrs and Beatty 24 and others 19,25−27, identified that most cats with pyothorax (70-90%) had bilateral effusions. It is not reported in detail how ‘unilateral’ and ‘bilateral’ effusions are interpreted in these studies. More importantly, our results suggest that cats with CHF or idiopathic chylothorax did not present with unilateral effusion. This finding may be important for ruling out these two diseases and prompt clinicians to perform thoracocentesis and fluid analysis or advanced imaging, as there is a higher likelihood of pyothorax or neoplastic disease when a unilateral effusion is identified.
The presence of a radiographic mediastinal mass had 83% and 100% positive predictive value for neoplasia with a low sensitivity and high specificity. Not all neoplastic disease was caused by mediastinal masses leading to a low sensitivity; however, when a mediastinal mass was radiographically diagnosed, there is a very high probability of neoplastic disease. There was one false positive case, where one cat assessed to have a radiographic mediastinal mass, had a diagnosis of pyothorax. This radiographic appearance may be due to loculated fluid within the mediastinum, causing an appearance of a mediastinal mass secondary to the exudative nature of the fluid. This case was reviewed and no post mortem or thoracic ultrasound was available to confirm this speculation. No cats with CHF and idiopathic chylothorax had radiographic mediastinal mass(es). Limitations of this finding are associated with the low sample size (only a total of 5 cats with confirmed mediastinal masses had radiographs available for assessment).
Radiographic pulmonary nodules only predicted neoplastic disease 50% of the time and masses from 50–67% of the time. Of the cats diagnosed with neoplastic disease, a small portion did not have sonographic or radiographic pulmonary nodules or masses, these included two cats with heart base and myocardial neoplasms, and three cats with neoplastic pleural effusions without a primary mass identified (supplementary table I). The variable neoplastic processes that were present in the population examined could be one reason for the low predictive value pulmonary nodules and masses. In addition. Some cats with congestive heart failure and chylothorax were interpreted to have radiographic nodules. On review of the congestive heart failure cases, it is speculated that cardiogenic oedema infiltrates and peri-bronchial infiltrates may mimic nodules or masses (Figure 6 [Insert Figure 6]). Several cats with pyothorax and chylothorax had radiographic nodules and this may be due to a combination of atelectasis and retraction of the lung margins causing regions of patchy consolidation. Unfortunately, these cats did not have a CT, post mortem or biopsy to confirm these speculations. No chylous and pyothorax cases had radiographic pulmonary masses.
In previous studies, cats with chylothorax have been described to have rounded pleural contours due to chronic fibrosing reaction of visceral pleura27,28 (Figure 7 [Insert Figure 7]), exerting a constrictive and restrictive effect on lung lobes. It was found that pleural abnormalities were not useful as a predictor, and only had a 30% positive predictive value for chylous effusion. This suggests that chylothorax may not always cause a constrictive and fibrosing effect to the pleura; however, the sample size examined was small and these results may not generalise to other populations. An additional finding from this cohort is that idiopathic chylous effusion never presented with unilateral effusion or presence of a mediastinal mass.
There are many limitations to this study, particularly pertaining to the retrospective access to cases. Not every patient had radiographs recorded which may create bias from missing data. Although we have accounted for multiple diseases in a single patient, it is possible that some cats had more than one disease causing pleural effusion that we were unaware of, with a potential for misclassification bias. In addition, not every patient had histopathology or cytology performed in the neoplasia group; although this was a small proportion of patients (n=4), this may reflect some inaccuracies in classification. Despite collecting data across 10 years, the sample sizes in the non-CHF groups were small, which may cause the lack of precision in the predictive values, sensitivity and specificity estimates. In particular, the FIP group had very small sample size; the data was collected prior to the widespread availability of Remdesivir and many owners with suspected FIP did not pursue further diagnostics. In addition, not all FIP cases with further diagnostics had pleural effusion and therefore not included in this study. A larger scale study examining more cases (pyothorax, neoplasia, idiopathic chylothorax, feline infectious peritonitis) would allow for more accurate results, therefore although this study is novel, it is preliminary.
All cases were obtained in practices in metropolitan Western Australia and the practices were within a 10km radius from the referral hospital. When interpreting the data, clinicians should be aware of the effects of disease prevalence on test results, which may affect the accuracy of the results, as the PPV and NPV results assume the prevalence of the disease aetiologies are similar in all populations. Several radiographic parameters were interpreted differently with some performed a priori as there are some radiographic parameters that could not be objectively measured (such as interstitial and alveolar patterns and small volumes of pleural effusion), leading to inconsistent application of measurement techniques. Although, inter-observer agreement was moderate to substantial for majority of the radiographic signs, several radiographic signs only had slight to fair agreement (distribution of pleural effusion, pulmonary parenchymal abnormalities) between the radiologists, which partly reflects the challenges of interpreting thoracic structures in the presence of pleural effusion, even among experienced operators.