Discussion
Pleural effusion of unknown origin remains a frequent problem, with important implications. Thoracoscopy is frequently necessary, as well for diagnostic or therapeutic reasons.
These data represent our experience of medical thoracoscopy in a period of almost 10 years, the first Belgian data to be published to my knowledge. The amount of patients is comparable to other studies.6–8
Out of 49 patients with nonspecific pleuritis 15 had less than 6 months of follow-up, and thus, we calculated a maximal and minimal sensitivity, as was done before in another study.9 Due to the little number of some diagnoses these parameters sometimes have a wide range, which makes it impractical for daily use. However, we think the sensitivity will be closer to the highest percentage, since we do not expect all patients with incomplete follow-up to have gotten a rare or missed diagnosis.
Thoracoscopy was performed under general anaesthesia, so in this discussion we tried as much as possible to compare to other studies in which thoracoscopy was performed under general anaesthesia.
Sensitivity analysis
General sensitivity
General sensitivity in our data was between 80.9% and 92.4%. This is comparable to the sensitivity of 93% in the recent meta-analysis of Wei et al, however, for local anaesthesia thoracoscopy.7 Sensitivities for video assisted thoracoscopy under general anaesthesia are, however, in the same range or a little higher (ranging from 82.3% to 95.2%).8 10–12
Negative predictive value range is 57.6%–83.1% for any diagnosis, which is comparable to 76% of McDonald et al8 but probably a bit lower than 85%–90% in Menzies et al.9 (95% CI 84% to 96% and 78% to 92%, depending on possible diagnoses in patients lost to follow-up).
Specific sensitivities
Sensitivity for malignant pleuritis was high (92.0%), but a little lower than in other studies (94.0%–95.0%).12 13 Sensitivity for mesothelioma in our study was a lot higher than in the meta-analysis of Wei et al (93.3% vs 42.0%), for unclear reasons (eg, percentage of mesothelioma was even less than in Valsecchi et al6, which study was however not included in the meta-analysis of Wei et al). It is however comparable to 90.9% in Ferrer et al.13
The malignancies diagnosed during follow-up were one mesothelioma, two gynaecological cancers, one primary renal cell cancer and one cancer of unknown primary tumour. So, in our study, there seems to be no specific primary cancer type that is missed frequently, in contradiction with the data of several other studies, in which frequently malignant mesothelioma or metastatic pulmonary carcinoma were diagnosed.9 14–16
Sensitivity was 90.9% for tuberculous pleuritis, and not 100% as in another study1 due to one missed case, which was diagnosed with a thoracoscopic lung biopsy. In a Turkish study, even after all patient got a closed pleural biopsy first, also 1 case of tuberculosis was missed on thoracoscopic pleural biopsy.17
Sensitivity for other benign conditions was very low, as has also been described elsewhere.1 9 12 18 19 This concerns most frequently patients with benign asbestos-related pleural effusion, and autoimmune disease such as systemic lupus erythematosus and rheumatoid arthritis. This is probably due to aspecific histological findings in these diseases, due to not well understood pathophysiology.20 Some diagnostic clues of these rare diseases are described by Boutin et al and by Anevlavis et al.18 20
Diagnoses made on pleural biopsy
The 44.3% of patients with malignant pleuritis is in the range of other studies (60.0%, 54.0%, 44.2%, 36.6%).8 9 12 17 Of course this also depends on previous diagnostic work up before thoracoscopy (eg, imaging, number of thoracentesis, closed pleural biopsy), but also on your patient population, and if the data were collected in TB-endemic countries.21 We have a higher percentage of nonspecific pleuritis compared with these other studies, and lower percentage of other benign conditions. We however have no empyema, haemothorax nor heart failure, which is prevalent in some other studies.9 12 Neither do we have (clinical) probable diagnoses, but only certain diagnoses, and we categorised all these uncertain diagnoses as nonspecific pleuritis.
Follow-up of non-specific pleuritis
In our study, in 59 patients with nonspecific pleuritis 5 (8.5%) were diagnosed with malignant pleuritis afterwards. This is practically the same as in a previous study of our department with a mean follow-up of 32.9 months, which showed 5 patients (8.3%) of malignancies diagnosed afterwards,16 2 of them being mesothelioma.
This is also comparable to two other studies, in which both 5% of malignancy was diagnosed during follow-up,14 15 however, in other studies up to 25% of malignancy was found,22 23 in the latter after a mean period of 6 months. In another study in a tertiary referral hospital with high incidence of mesothelioma 12% showed to have malignant pleural effusion, all of them mesothelioma related, and mean follow-up until diagnosis was 9.8 (±4.6) months.24 In another study in which 86 patients were followed up for approximately 5 years, 3.5% patients showed to have malignancy within the first year of follow-up, all of them mesothelioma.25 Janssen et al found 14.9% malignant pleuritis, diagnosed after a mean follow-up of 8.7 months, almost 1 in three being mesothelioma.
The high percentage of mesothelioma is again contradictory to our data, in which we found multiple types of metastatic cancers, and no majority of mesothelioma.
We chose a 6-month period of follow-up, which, comparing to other studies, might have been short and it is possible we missed some malignancies. This decision was made because of the retrospective nature of our study, and for a longer period of follow-up ranges of sensitivities would have been even bigger (more loss to follow-up). However, our percentage of malignancy is almost exactly the same as in Venekamp et al, in which the follow-up was almost 33 months, and certainly comparable to some of the studies mentioned above. Most malignancies are diagnosed within 1 year after thoracoscopy, and rarely later. Number needed to detect one malignancy during the first year after thoracoscopy was calculated as 18, and afterwards 250.15 17 25
In our data persisting pleural effusion was seen in 26.5% (another 9 patients loss to follow-up), and in 22 of 34 patients with follow-up of more than 6 months the effusion regressed, (mean time of 7 months). In two other studies, persistence or reappearance ranged from 12.5% to 16.7%.14 16
Safety
Median length of stay was 4 days, which is 1 day longer than the study of McDonald et al.8 However, in their study, patients admitted more then 2 days before thoracoscopy were excluded from analysis of length of stay, which was not the case in our data set. Since the IQR is from 3 tot 14 days, the rather long median length of stay seems to have been caused by a minority of patients with unusual long length of stay.
Mortality in our study was 3.8%, with a mean age 78 years, and all of these patients were admitted urgently before diagnosis of pleural effusion. According to the standards of the British Thoracic Society and the Society of Cardiothoracic Surgeons in Great Britain and Ireland (BTS/SCTS) mortality should be less than 1.6%. As two of five deceased patients died on postoperative day 19 and 23, one could argue if this due to thoracoscopy, and it seems mortality in our study was not only due to the thoracoscopic procedure, but also due to underlying disease and general weak condition.10 In Harris et al, mortality was 6% during hospitalisation.12
Major complication rate in our study was 4.6%, which is higher than the standards of BTS/SCTS (<2%).10 However, in other studies, major complication rates ranged from 1.9% to 15.0%.5 8 12 26 Definition of major and minor complications is not always clearly stated and can be different between studies. Also, it is not always noted if (and which proportion of) patients are hospitalised electively or urgently because of symptoms. The latter could indicate a worse general condition. These two reasons make it difficult to compare complication rates.
Minor complication rate was 1.5%, which is a lot less than reported in the studies of Colt and Harris et al (5.7% and 8.0%, respectively).5 12
Strengths and limitations
Our data collection was performed retrospectively from a single tertiary centre. Given the retrospective nature, we relied on data in the patients’ files, with the inherent limitations. Follow-up was 6 months, after which we possibly missed some diagnoses. We also have some patients lost to follow-up, generating statistical uncertainty. The number of patients included is not very big, but certainly comparable to other studies. We have no probable diagnoses, only certain diagnoses, which makes our data more robust than some other studies.
Generalisability
This is a single-centre study, performed by an experienced respiratory physician during a period of almost 10 years. Although this certainly limits generalisability, given the data are comparable to studies in multiple other countries, we think these results should be rather generalisable in the same setting and region.