Investigation and outcomes in patients with nonspecific pleuritis: results from the International Collaborative Effusion database

Discussion

We describe the first multicentre, international dataset for NSP, through which we attempt to answer several research questions proposed by the ICE database group.

We demonstrated that NSP was idiopathic in 44% of cases, a rate comparable to existing evidence (mean incidence 48%) [2–8, 10, 11, 13, 14, 17–19]. However, “idiopathic” NSP is difficult to define, as there are >60 known causes for pleural effusions and, therefore, only after exclusion of most or all of these aetiologies should the label be attributed. To date, a minimum standard panel of investigations to further investigate the aetiology of NSP has yet to be agreed upon, so this label is prone to inter-operator and inter-site variability.

The subsequent diagnosis of a pleural malignancy has been the focus for most NSP studies [3–5, 7, 8, 10, 11, 13–15, 17–21]. Our dataset demonstrated an incidence of eventual pleural malignancy following an NSP result of 6% (95% CI 3–11%). This is slightly lower than the rate described in the existing literature: 8% (95% CI 7–10%) [3–5, 7, 8, 10, 11, 13–15, 17–21]. All eventual malignant diagnoses were lung cancer or MPM, with the latter more frequently observed. All cases save two were diagnosed histologically. The remaining two cases were diagnosed clinically based on suggestive radiological features and MDT consensus. This highlights an issue around the definitions for what constitutes a malignant effusion as opposed to the less well defined “paramalignant” effusion [13]. The precise definition of malignant pleural disease requires histological or cytological confirmation, but in clinical practice this is not always possible nor appropriate. Often a clinical diagnosis is made based on many factors, including radiological evidence. While a “paramalignant” effusion has been described as an inflammatory pleuritis in the presence of an active malignancy, not explained by other aetiologies, but not due to malignant infiltration of the pleura [13], there clearly exists a degree of overlap between these entities and there are difficulties in differentiating the conditions in clinical practice, which bears consideration and again highlights a need for standardised definitions.

With regards to risk stratification for those who will progress to an eventual malignant diagnosis, at any stage, asbestos exposure was the only baseline characteristic indicating a signal towards association, but not meeting statistical significance. CT appearances of a mass lesion, either within the lung or pleura, were the only imaging features associated with a false-negative biopsy result, which is entirely intuitive. There was a trend towards identifying pleural thickening on ultrasound and an eventual malignant diagnosis, although this was not statistically significant. This finding is entirely consistent with previous work that has shown this feature is highly specific for malignant pleural disease [25]. As a readily performed point-of-care test, this can be of value in the risk stratification of patients with NSP and further reinforces the need for standardisation in both performing thoracic ultrasound, and training therewithin, both key ERS directives [26]. In contrast to other studies that demonstrated recurrence of effusion as a predictor for an underlying malignant process [11], we detected no such associations, with repeat thoracocentesis being required equally often in both groups. We suggest that the need for repeat thoracocentesis, and by extension, effusion recurrence in NSP, depends entirely upon the prevalence of the different aetiologies driving the condition, rather than a specific feature of malignancy alone. All the false-negative biopsies arose in the LAT group, but interpretation is difficult given how few patients with a surgical biopsy were included in this dataset. Although it is intuitive to think a difficult or incomplete thoracoscopic examination may provide an explanation for false-negative biopsies [10], most thoracoscopists would advocate an alternative biopsy modality if they were faced with a technically challenging procedure and were not reassured they had satisfactorily inspected the pleural cavity, when there is a high pre-test probability for pleural malignancy. The TARGET study aimed to test this hypothesis, by directing patients with an initial nondiagnostic biopsy and ongoing clinical suspicion to PET-CT versus usual care [27].

The median time to evolving malignancy was 12.2 months, with a wide range (0.8–32 months). Expert opinion suggests that the follow-up period should be ≥24 months and perhaps the minimum period ≥12 months. This remains a difficult area to offer fixed recommendations for duration of follow-up. In our dataset, five out of 11 patients developed malignancy within 12 months, with three having done so within 3 months. The most delayed presentation was 32 months. Within the literature, the longest follow-up period after which time pleural malignancy was diagnosed was 64 months [21]. It would be reasonable to surmise that the early progressors are more likely to represent a diagnostic accuracy problem, while the late progressors may truly represent an early or pre-malignant phase that is not diagnosable as a malignant process at time of biopsy, or indeed that they do not have malignancy of any kind, and the finding of NSP is unrelated (particularly relevant in the presence of various comorbidities). Therefore, the early progressors may well represent a genuine false negative following a thoracoscopic or other form of biopsy, while it would not be correct to label the late progressors as such. There are no agreed definitions on what constitutes a false-negative biopsy, and whether this should take the form of an arbitrary time interval, or a more nuanced set of criteria (e.g. progressive radiological features, onset of “red flag” symptoms within a defined time period). While it would be clinically useful to be able to differentiate between those who have truly benign disease from those who are likely to progress, it would be equally useful to risk-stratify the early developers from the late, to offer a more personalised follow-up plan.

There was little to differentiate between the two groups in our dataset, although we did observe that the late evolvers were often MPM and this would be entirely consistent with the natural history of MPM, with both a long latency from initial asbestos exposure and the recognition now of “mesothelioma in situ”, indicating a pre-invasive stage of disease [28]. As the research landscape evolves, we will understand the relationship between this precursor lesion and “reactive atypical mesothelial hyperplasia”, a histological finding often encountered in NSP [29, 30]. With a general consensus of 3–6-monthly follow-up with regular cross-sectional imaging for a 12–24-month period, nine out of 11 of our cases of evolving pleural malignancy would have been captured, while two would have been missed. In planning follow-up for patients, clinicians should be open about the diagnostic uncertainties they are faced with and should settle on an appropriate follow-up schedule through a process of shared decision-making that is tailored to the patient. It would not be unreasonable to offer a more prolonged follow-up in those in whom a diagnosis of MPM is suspected as opposed to an alternative malignancy, which is likely to manifest earlier. As with other screening interventions, it may be helpful for clinicians and patients to understand the concept of “number needed to follow-up” (NNF) to diagnose one malignancy. In the largest series on NSP, by Reuter et al. [15], the NNF to identify one malignancy during the first year was demonstrated to be 18, getting less effective with increasing time (NNF of 260 between years 1 and 3).

There are a number of weaknesses to this study. As with all retrospective cohort studies, incomplete data and variation in the definitions used by participating sites for data entry does affect data integrity and results in significant inter- and intra-site heterogeneity. This is clearly evident in the assigned aetiologies ascribed to NSP and there may well have been significant differences in the pathological tests and criteria used for diagnosis of NSP between centres, which this study is not able to account for. A prospective study with standardised definitions, an agreed minimal panel of both clinical and pathological tests, alongside an independent pathology review following an NSP finding may overcome this. This may be particularly relevant in the diagnosis of MPM, where BAP-1 and p16 testing, in conjunction with other investigations, have proven to be instrumental in securing a diagnosis of MPM [28], which our study did not specifically look at. The study was underpowered for a number of the questions it was intended to explore. This was largely in part due to the coronavirus disease 2019 pandemic, which severely disrupted participating centres’ ability to contribute to the study, but may also have been impeded by recruiting participants primarily from clinical databases accessed by pulmonologists rather than pathology databases, contributing to selection bias. Additionally, the event rate of eventual pleural malignancy was lower than expected, and therefore, much of the intended hypothesis testing based on our research questions failed to meet statistical significance, although associations worthy of further research were shown.

This is the first multicentre study on NSP and its outcomes, enabled through the ERS ICE research collaborative. As well as providing additional knowledge regarding NSP, we have demonstrated that international centres with an interest in pleural disease can contribute data to better study under-researched areas and generate ideas for future areas of research.