Ultrasound is an ideal imaging modality to detect pericardial and pleural effusion as it is more sensitive than physical examination or chest X-ray [6]. It is, therefore, suggested in the current guidelines as a standard assessment in the evaluation of patients with HF and should, therefore, also be part of a detailed echocardiographic examination in patients with CA [7]. One study reported on the predictive effect of pericardial effusion in patients with light chain amyloidosis (AL); however, the studied cohort was small (n = 31) and only analysed the effects of effusion on all-cause death rather than cardiac death [8, 9]. There has recently been evidence that pericardial effusion may also be a sign of myocardial edema following local amyloid infiltration and consecutive inflammation in patients with AL and ATTR [10, 11]. This may be especially relevant in AL due to cytotoxicity of light chain amyloid fibrils [12]. On this note, Riduoani et al. were recently able to show that cardiac magnetic resonance imaging (CMR) could detect local inflammation using T2 weighted sequences and may be useful in distinguishing between ATTR and AL [13]. It has also previously been proposed that in patients with AL, pleural effusion may not only result from right HF, but also from primary pleural infiltration causing consecutive fluid secretion [14]. Another potential mechanism could be a fluid shift resulting from a decrease in intravascular oncotic pressure due to an increased renal loss or reduced production of serum proteins. The exact mechanism of the development of pericardial and pleural effusions in CA even in patients with presumably preserved left and right ventricular (LV, RV) function, however, remains unclear.

We conducted this study to further shed light on pericardial and pleural effusions in AL and ATTR with cardiac involvement and their impact on cardiac outcome.

We consecutively included patients with AL and ATTR amyloidosis in our prospective clinical registry at the Medical University of Vienna. Diagnosis as well as follow-up were performed and documented at our dedicated CA outpatient clinic. Visits were routinely scheduled every 6 months or more frequently when appropriate, as judged by the clinician. When necessary, patients were admitted to the cardiology ward for intensified treatment. All patients gave written informed consent before study inclusion. The study protocol complies with the Declaration of Helsinki and was approved by the local Ethics Committee of the Medical University of Vienna, Austria (Ethics committee identification number: 796/2010).

In the years preceding 2016, the diagnosis of ATTR was made by endomyocardial biopsy (EMB). At least five tissue samples were taken from the LV myocardium (Bipal® biopsy foceps Cordis® Corporation, Bridgewater, NJ). Specimens were then fixed in formaldehyde, embedded in paraffin and subsequently stained with congo-red dye to detect interstitial amyloid deposits. Additionally, tissue samples were examined under polarized light to reveal green birefringence. When amyloid fibrils were found, inmmunohistochemical analysis was performed to further characterize amyloid type according to local standard established procedures (AmY-kit amyloid antibodies, Martinsried, Germany).

Following the landmark publication of Gillmore et al. in 2016, which presented a non-invasive diagnostic algorithm for the diagnosis of ATTR using serum and urine light chain analysis as well as bone scintigraphy, transthoracic echocardiography (TTE) and/or CMR, EMB was only performed when non-invasive test results, where ambiguous or unclear [15]. CMR was performed including gadolinium contrast application, T1 mapping and calculation of extracellular volume, as previously described [16, 17].

The diagnosis of AL was either made by EMB or by extramyocardial biopsy. To confirm the presence of cardiac involvement, at least one of the following features had to be present: (1) LV hypertrophy with an interventricular septum (IVS) thickness of > 12 mm and/or presence of apical sparing determined by TTE or (2) elevated cardiac biomarkers [18, 19].

Clinical outcomes were documented by follow-up at our outpatient clinic, as well as by phone calls and by screening medical records in our electronic hospital record system. The clinical endpoints were defined as cardiac death and death from any cause. Further endpoints were death from any cause and a combined endpoint of cardiac death or HF hospitalization. In the case of a clinical event, local and external records were carefully screened and cause of death was reviewed by a clinical adjudication committee of board certified cardiology specialists (D.B., R.B.). HF hospitalization was defined as an event leading to sudden dyspnea, weight gain, peripheral edema and requiring admission to hospital and/or intravenous diuretic therapy.

Baseline assessment was performed at our CA outpatient clinic and included demographic, clinical, laboratory and imaging parameters.

All TTE exams included standard imaging parameters, as well as an extended protocol performed by experienced and certified specialists on high-end machines (GE Vivid 95 and Vivid 7; GE Healthcare, Wauwatosa, WI, USA). Images were acquired and measurements were taken according to current guideline recommendations [20‐22].

Presence and size of pericardial effusion was assessed in a subcostal four-chamber view. When imaging quality was suboptimal from a subcostal approach, an apical four-chamber view was used to determine the presence of pericardial effusion. Assessment of pleural effusion was performed bilaterally in every patient. Additionally, the presence of effusion was confirmed by chest X-ray or CMR in patients who underwent CMR for the diagnosis of CA.

Speckle tracking imaging was performed after image acquisition on a modern offline clinical workstation equipped with dedicated software (EchoPAC; GE Healthcare, Wauwatosa, WI, USA). The global longitudinal strain of the LV (LV-GLS) was measured by tracking the myocardium in an optimized apical three- four- and two-chamber view. The global longitudinal strain of the RV (RV-GLS) was measured in the free lateral RV wall in an optimized apical four-chamber view. All obtained TTE parameters are shown in Table 1.

Pericardial effusion size was measured at the location of maximum fluid accumulation. Pericardial effusions were defined as hemodynamically significant according to current guidelines [4], i.e., presence of swinging heart, early diastolic collapse of the RV, abnormal ventricular septal motion, exaggerated respiratory variability (> 25%) in mitral inflow velocity, or when the patient was clinically compromised (cardiogenic shock).

Continuous data are shown as median values and inter-quartile range (IQR). Categorical variables are expressed as numbers and percentages. Analysis of variance (ANOVA), Student’s T-test and Chi square test were used as appropriate to determine statistically significant differences in baseline variables. Univariable Cox regression analysis was applied to test baseline variables for effects on the primary endpoint of cardiac death or HF hospitalization. Variables which significantly predicted outcome were then entered in a stepwise forward multivariable Cox regression model within respective categories to adjust for potential confounding effects. Kaplan–Meier plots with respective log-rank tests were computed to assess the time-dependent discriminative power of effusion type on endpoints. A p-value < 0.05 was considered as statistically significant. All analyses were performed using SPSS 24 (IBM Corp. NY, USA).

In total, 169 patients with cardiac AL or ATTR amyloidosis were included in our clinical amyloidosis registry between March 2012 and January 2019. However, only 143 patients had sufficient echocardiographic image quality of both ventricles to be included in this study. Of these, 85 patients (59.4%) presented with ATTR and 58 (40.6%) with AL. Diagnosis was confirmed by EMB in 19 patients with AL (13.3%) and 31 patients with ATTR (21.7%). In 37 patients, the diagnosis of CA was made using a combination of non-cardiac biopsy and imaging (AL = 33, ATTR = 4). All other were diagnosed by imaging alone. CMR was performed in 113 patients (AL = 47, ATTR = 66).

Detailed baseline characteristics of the study population according to presence of pericardial and/or pleural effusion are shown in Table 1.

Patients with effusions (n = 78) presented with more severe clinical symptoms as described by New York Heart Association (NYHA) class (p = 0.018) and were more commonly already on loop diuretics (p < 0.001) and/or aldosterone antagonists (p = 0.002) at baseline. Left ventricular global longitudinal strain (LV-GLS) as well as RV free wall strain (RV-GLS) were more severely impaired in patients with pericardial or pleural effusions with p-values of 0.007 and 0.004, respectively.

Notably, patients with effusions not only showed higher levels of cardiac biomarkers (median NT-pro BNP 4268.5 pg/ml (IQR 2048.0–8866.0 pg/mL) versus 1635.0 pg/ml (IQR 827.0–3549.0 pg/mL), p < 0.001 and median troponin T 68.0 ng/L (IQR43.5–118.0 ng/L) versus 28.0 ng/L (IQR 22.0–56.0 ng/L), p < 0.001), but also higher levels of gamma glutamyltransferase (GGT) of 79.0 U/L (IQR 46.0–140.0 U/L) versus 42.0 U/L [(IQR26.0–126.0 U/L), p = 0.033] and lower levels of serum albumin of 38.7 g/L (IQR 33.0–42.4 g/L) versus 43.0 g/L [(IQR 38.3–45.2 g/L), p = 0.002]. Furthermore, median serum lactate dehydrogenase was higher in patients presenting with pericardial and/or pleural effusions [(241.0 U/L (IQR 206.0–292.5 U/L) versus 224.0 U/L (IQR 190.0–249.0 U/L), p = 0.017].

At baseline, 35 patients (24.5%) presented with isolated pleural effusion, 24 patients (16.8%) with isolated pericardial effusion and 19 patients (13.3%) with both pleural and pericardial effusion. Pleural effusions detected by echocardiography could only be confirmed by chest X-ray in 76.2% of all patients. Table 2 shows a detailed description of types of effusions and their distribution in patients with AL versus ATTR. In general, pericardial effusions were small with a medium diameter of 7.1 mm (IQR 5.2–9.7 mm) and none were hemodynamically compromising.

When analysing parameters associated with right heart dimensions and function, we found that patients with AL presenting with pleural effusions at baseline had more impaired RV function, which was reflected by RV-GLS, TAPSE and RV-TDI. There was no difference in RV strain when comparing the basal- and mid segments with the apical segments of the RV (Fig. 1a). Patients with pleural effusions had higher amyloid burden, seen as increased interventricular wall thickness (IVS) and LV extra cellular volume (LV-ECV) resulting in impaired contractility as measured by LV-GLS. Interestingly, not only cardiac biomarkers such as serum NT-pro BNP and troponin T, but also GGT and CRP levels were higher in patients with pleural effusions (Table 3).

By contrast, AL patients presenting with pericardial effusions only showed significantly more impaired strain in the basal and mid regions of the RV. Apical RV regions also showed more impaired function, even though this difference was not statistically significant (supplementary table S1). Notably, there was no difference in cardiac biomarkers between the two groups.

In patients diagnosed with ATTR, the presence of pleural effusions was not accompanied by more impaired RV strain values or other parameters of RV function. However, when comparing the basal and mid-segments of the RV to the apical segments, we were able to detect that the more apical segments showed better RV function (RV-apical sparing, Fig. 1b). Similarly, there was no statistically significant association between the presence of pleural effusion and LV-function or degree of LV amyloid infiltration seen in CMR. Nevertheless, serum levels of cardiac biomarkers were significantly higher in patients with pleural effusions compared to those without. Notably, albumin levels were lower in patients presenting with pleural effusions (Table 4). Supplementary table S2 shows, that there is no association between RV function and pericardial effusion in patients with ATTR.

LV hypertrophy was more pronounced in patients with ATTR than those who had been diagnosed with AL with a mean IVS thickness of 20 mm versus 17 mm, respectively. LV function was better in patients with AL as indicated by LV-GLS (p = 0.010), as well as LV ejection fraction (LVEF) measured by CMR (p < 0.001). Interestingly, RV function measured using the volumetric method of RV ejection fraction by CMR was also better in patients with AL (p = 0.007), while RV function parameters reflecting RV contractility, such as RV-GLS and TAPSE did not show any differences between patients with AL and ATTR (p = 0.562 and 0.298, respectively).

After a median follow-up time of 19.0 months (IQR 9.5–32.5), 42 patients had died (29.4%). In more detail, 28 patients with AL (48.3% of the total AL population) and 14 patients with ATTR (16.5% of the total ATTR population) died. During follow-up, 26 (61.9%) could be classified as cardiac deaths. In seven patients, the exact cause of death could not be certified, due to lack of medical records or information from relatives and were, therefore, classified as all-cause death.

Kaplan–Meier analysis showed that among the total population of cardiac amyloidosis, patients with pleural effusions had worse outcomes than patients with no effusions or isolated pericardial effusions. This was irrespective of concomitant presence of pericardial effusion in addition to pleural effusion. Furthermore, the presence of isolated pericardial effusion did not have an impact on outcome. When separating the studied cohort by amyloid subtype, we saw that effusion type was predictive for outcome in patients with AL, but not in ATTR (Fig. 2). In addition, we used univariable Cox regression models to calculate hazard ratios (HR) for cardiac death for all baseline parameters in patients with AL. We found that the presence of pleural effusion predicted adverse outcome with a HR of 5.604 [(2.334–13.456), p < 0.001] alongside with NYHA class, NT-pro BNP and troponin T. Among echocardiography parameters, LV global longitudinal strain (LV-GLS), LV end diastolic diameter (indexed to body surface area), interventricular septum thickness, and RV tissue Doppler index (RV-TDI) predicted outcome in the univariable model. After adjusting for confounding factors within respective categories of baseline characteristics, only NYHA class, LV-GLS and NT-pro BNP remained predictive in the multivariable model (Table 5).

We are aware of the limitations, which are associated with the single-center design of the present study. However, this also implicates advantages regarding the consistency of diagnostic work-up, treatment and follow-up. Infiltration of pericardial or pleural tissue can of course not be certified without histological tissue characterization, which is not feasible in pre-mortem patient cohorts for obvious reasons. Even contrast CMR imaging would not be able to detect, let alone quantify amyloid deposits in the pericardium due to technical limitations. Effusion samples were not collected, because most patients did not undergo thoracocentesis, but were medically treated. Analysis of effusions could have given more insight into their underlying cause. Furthermore, we acknowledge, that diuretic treatment may have influenced the results of this study.