The Role of Whole-Body Magnetic Resonance Imaging (WB-MRI) in Patients with Paroxysmal Nocturnal Hemoglobinuria (PNH)

In PNH thromboembolic events (TEs) represent the leading cause of morbidity and mortality. Between Dec 2013 and Jan 2016 37 PNH patients (pts) (23 PNH, 14 AA/PNH; 51% (19/37) females; median age 44 years, median D-dimer levels 0.22 mg/l) were examined with a whole-body magnetic resonance imaging (WB-MRI) scan at 1.5 T to detect TEs. Pts were treated according to German PNH guidelines, including eculizumab therapy. 64% (24/37) of the pts had no documented TEs prior to observation. Two pts had suspected TEs in their clinical history. 29% of the pts (11/37) had a known history of venous thromboses (deep venous thrombosis (DVT) (5/11), portal venous thrombosis (PVT) (4/11), vena caval thrombosis (VCT) (2/11). A myocardial infarction was reported in one pt, and two had a cerebral venous sinus thrombosis (CVST) or a thalamic infarction. Six pts (16%) had at least two prior TEs. In pts with prior TEs no progression of the existing TEs was observed. In pts on eculizumab and prior TEs as well as treatment-naïve pts silent bone and renal infarctions were detected. Furthermore, a clinically non-critical arterial occlusion was identified. WB-MRI scans present a novel, non-invasive method to assess the complete vascular status of PNH pts and allow the detection of previously undiagnosed vascular complications, affecting treatment indications and regimens.

All patients were treated according to current German PNH guidelines, including the use of the terminal complement inhibitor eculizumab (70% (26/37)), with eculizumab being initiated in 23 patients prior to WB-MRI scans. In the remaining three patients, eculizumab was initiated later on for reasons other than TEs.
The standardized protocol included a cranial T2-SWI (susceptibility weighted imaging), a cranial 3D-TOF-MRA (time of flight MR-angiography), a cranial T2-SPACE (sampling perfection with application optimized contrasts using different flip angle evolution), a whole-body arterial and venous contrast-enhanced MRA and a contrast enhanced T1-weighted whole-body VIBE (volume interpolated breathhold examination).
The whole-body contrast-enhanced arterial MR-angiography (MRA) is based on the acquisition of four overlapping 3D datasets acquired in coronal planes with a 3D fast low angle shot (FLASH) gradient recalled echo sequence with steps for the The contrast agent administered was Gadovist (Gadobutrol 1.0 mmol/ml, Bayer) with a dose of 0.2 ml/kg bodyweight and an injection rate of 1.2 ml/s for the first half of the contrast agent and an injection rate of 0.7 ml/s for the second half of the contrast agent followed by a saline flush with an injection rate of 0.7 ml/s for 20 ml of 0.9% saline solution.
The whole body MR-venography was performed immediately after the arterial MRA with a fat suppressed 3D T1-weighted gradient echo volume interpolated breath hold-examination (VIBE) sequence acquired in axial planes with 8 to 10 steps depending on the patients height (repetition time (TR) 3.66 ms; echo time (TE) 1.4 ms; matrix 166 × 512 mm; FOV 480 mm; flip angle 12°; acquired voxel size 2.08 × 0.94 × 3.50 mm, reconstructed voxel size 2.1 × 0.9 × 3.5 mm; slice thickness 3.5 mm; acquisition time per step 0:20 min). The venography was acquired from the base of the skull to the veins of the feet.
The total time of the protocol was 20:10 minutes. A DWI/ADC or FLAIR sequence of the brain was not acquired, since none of the patients examined presented with acute or subacute neurological symptoms. Instead of the FLAIR sequence a 3D SPACE sequence was used to assess white matter lesions.
All acquired MR images were then assessed by two radiologists in consensus for the clinical report and again assessed retrospectively by two radiologists (4 years in cardiovascular training and 14 years in cardiovascular training) in consensus for the presence or absence of TEs as well as infarctions as a result of a TE.
All acquired cranial MR images were also assessed by two blinded radiologists (4 years in neuroradiological training and 5 years in neuroradiological training) retrospectively for cerebral changes associated with TEs.
As suggested by Barcellini et al. 16 , the agerelated white matter changes (ARWMC) scale as classified by Wahlund et al. 17 , was used to assess and quantify chronic ischemic small vessel disease defined as white matter abnormalities with T2 hyperintense signal with a distinction between deep white matter (WM) and basal ganglia (deep grey matter). The presence or absence of diffuse T2 signal hyperintensities of the periventricular WM which were consistent with chronic vascular degeneration or leukoaraiosis was also assessed. Focal alterations consistent with old ischemic strokes were also assessed and define as malacic fluid-filled cavities presenting with a hyperintense signal on T2-weighted images. The absence or presence of active or previous bleedings or microhemorrages as well as atrophy was also evaluated.
The score for WM lesions was calculated as suggested for each side separately for the frontal, parietal-occipital, and temporal lobe, and also separately for the infratentorial compartment. WM lesions as well as basal ganglia lesions were scored as following: 0 = no lesions, 1 = focal lesion, 2 = initial confluence of lesions, 3 = broad involvement of the entire region.
For leukoaraiosis a more general score was used with 0 = absence of lesions, 1 = mild to moderate, 2 = severe. The circle of Willis and its branches was also evaluated for abnormalities, e.g. stenosis, aneurysms, and Moya-Moya-like alterations, as well as the cerebral venous sinus for the presence or absence of thrombosis or abnormalities e.g. aplasia.

Results
In PNH patients with a medical (10/37) or even possible medical history (2/37) of TEs, we could not confirm a progression of the existing thrombotic events at the time of MRI examination. In 11 out of these 12 patients, eculizumab treatment was initiated prior to the WB-MRI examination (median eculizumab treatment duration 15.1 months; range 2.9-68.5 months) either due to thrombotic events (3/11) and/or other PNH-associated complications, including a significant transfusion requirement for red blood cells, recurrent abdominal pain crisis, or recurrent and severe PNH-associated hemolytic crisis. In patients on chronic eculizumab therapy and documented prior thromboses or TEs, in addition to patients with suspicion of thrombotic events in their medical history, the reported thrombotic events could be confirmed by WB-MRI in four patients (patient 2, 15, 16, and 22). Beyond this, unknown TEs prior to the examination date, including non-symptomatic bilateral bone infarctions of the vascular bed of the lower extremities (femoral and proximal tibial bone) could be detected in one patient (patient 26 - Fig. 1A) by the WB-MRI in addition to prior undiscovered bilateral renal infarctions in another patient (patient 15) with a history of multiple thromboembolic complications in the past. Moreover, in one female patient (patient 33), in whom eculizumab treatment was initiated due to a severe transfusion requirement for red blood cells, an asymptomatic bone infarction of the right proximal femur was additionally diagnosed by the MRI examination. In one male patient, in whom eculizumab was initiated due to recurrent abdominal pain crises, a previously not documented, clinical non-critical occlusion of the right anterior tibial artery was identified.
In comparison to PNH patients, in whom eculizumab treatment was not indicated by the time of the MRI examination according to current German PNH guidelines, a previously undiagnosed left renal infarction Only one patient was known to have a cerebral infarction in the history. In 22 patients white matter abnormalities relative to chronic ischemic small vessel disease could be detected, with 14 patients presenting with periventricular and deep white matter lesions, three patients with only periventricular white matter lesions, and five patients presenting with only deep with matter lesions. No intracranial arterial abnormalities were detected and no venous abnormalities distinguishable from anatomical variants were present, whereas, in five patients microbleeds or cerebral hemorrhages were detected (e.g. patient 18 or 29 - Fig. 1C,D).
In PNH patients, in whom previously unveiled thrombotic complications were identified, PNH clone size of GPI-deficient granulocytes (FLAER) was larger than 50% at the date of the MRI examination. Taking the formerly not diagnosed thrombotic findings into account, the overall number needed to treat in the setting of PNH by WB-MRI scans was 8.0 (95% CI, 4.1-27.8). Independently of the observed findings, there was no evidence of pulmonary embolism in the studied PNH patients at observation time. The complete patient characteristics at the time of examination, including the observed findings, are summarized in Table 1.

Discussion
In PNH patients, thromboembolism represents the most common cause of morbidity and mortality 18 . In these patients thromboses may occur in typical as well as atypical locations, affecting in particular the venous system, especially hepatic (Budd-Chiari syndrome) and cerebral venous vessels 1,19 . Beyond that, the occurrence of visceral and intraabdominal venous thromboses with involvement of mesenteric, portal, and splenic veins, in addition to vena cava and deep venous thromboses, are found to be more frequent in the setting of PNH in comparison to the general population 8,20 . Apart from venous thromboses, arterial thromboses are also more frequently found in these patients, leading to the occlusion of cerebral, coronary, visceral, and retinal vessels with subsequent infarctions 9,21-23 .
As depicted by multiple studies in the past and summarized by Hill et al. in 2013, the presence of TEs and their associated complications is of high prognostic value and affiliated with a poor survival (relative risk at 8 years, 10.2) 4 . Likewise, a survival rate of 40% at four years was computed by Socié et al., 1996, in the context of thrombosis at initial presentation prior to the eculizumab-era, in addition to an up to five-to 15.4-fold overall increased risk of death as shown by Nishimura et al. (2004) in patients with PNH in the United States and Japan 12,24 . It is not surprising, that the rate of silent thrombosis, even in patients with recent disease onset, seems to be underestimated, as proven by the work of Hill et al., 2012, providing evidence of previous subclinical pulmonary thromboses in addition to subclinical myocardial damage in the presence of PNH, respectively 14 . Therefore, silent thrombosis and TEs, if not detected promptly, may subsequently result in permanent organ dysfunction, significantly restricting quality of life.   The current retrospective monocentric study performed at the Department of Hematology at the University Hospital of Essen between Dec 2013 and Jan 2016 was designed to evaluate the rates of subclinical and silent venous and arterial thromboembolic complications in otherwise clinical asymptomatic PNH patients and also to follow up on known TEs by the use of WB-MRI scans. To the best of our knowledge, this is the first report in PNH patients for whom the complete vascular status at a given point in time was subject to analysis. As described in the patient characteristics, 37 patients with either classic PNH or a detectable PNH-clone were assessed with a WB-MRA. Out of these 37 patients, 62% of the patients (23/37) were on current eculizumab treatment according to current German PNH guidelines by the time of examination. Twelve out of these 23 patients (52%) had a medical history of prior thrombotic events. In three of these twelve patients (25%), thrombotic events preceded the diagnosis of PNH. In one of these three patients, a diagnosed portal venous thrombosis in addition to a late abort led to the diagnosis of PNH (patient 22). In the second patient a possible medical history of pulmonary embolism prior to the diagnosis of PNH was evident (patient 36). The third patient had a possible history of a caval and right femoral venous thrombosis prior to the diagnosis of PNH (patient 2), which was confirmed by duplex sonography and the acquired MR images. In three of the remaining 14 patients, eculizumab treatment was initiated for other reasons than thrombotic events, while in the rest of the patients, eculizumab therapy was not indicated at the time of examination as they remained asymptomatic or were under current immunosuppressive therapy due to bone marrow failure at the time of examination.
Assuming the initial hypothesis of silent thrombosis in the setting of PNH, the observed data are consistent with other studies, supporting the evidence of previously not-recognized thrombotic events 14 . Our findings demonstrate the presence of additional silent renal infarctions of unknown age in five percent of the patients (2/37). Out of these patients, one patient had a history of multiple thrombotic events and therefore was already under chronic eculizumab treatment (patient 15), whereas in the second patient, no prior thrombotic complications had been documented in the past, so far being treatment-naïve. In contrast to previous studies there was no evidence for silent major TEs (e.g. PE) at the time of analysis 14 .
In addition to previously not-diagnosed renal infarctions, bone infarctions of unknown age of the lower extremities, affecting both the femoral and proximal tibial bones, could be detected by WB-MRI scans in three patients (8%). At this point of time, bone infarctions represent a novel embolic complication in the context of PNH. As the overall prevalence of bone infarctions is still unknown, several risk factors have been identified, with glucocorticoid therapy, chronic alcohol abuse, and dyslipidemia representing the most common causes predisposing to their development. Other risk factors include an antiphospholipid syndrome or hemoglobinopathies such as sickle-cell disease (SCD) and thalassemia 25 . However, bone infarctions are reported as rare or very rare. As the empirical treatment of PNH involves the use of steroids, despite the fact that studies concerning the role of glucocorticoid therapy in PNH are inconsistent and randomized trials are lacking, two (patient 25 and 26) out of these three patients in whom bone infarcts were found to be present, were subject to intermittent steroid treatment in their past, which might have contributed to the development of our current findings of osseous infarctions. Similar to the findings of Lafforgue et al., 2016, the distal femoral and proximal tibial bones represent the most common sites of involvement.
With the help of this high sensitive and fast whole body imaging technique, an otherwise not recognized non-critical occlusion of the right anterior tibial artery could be identified in a male patient, in whom eculizumab treatment had been initiated due to hemolytic and recurrent abdominal pain crisis (patient 6).
However, to detect myocardial scarring after myocardial infarction would have required a cardiac MRI including late gadolinium enhanced (LGE) sequences. As adding a cardiac MRI (cMRI) would have lengthened the examination time for the patient, the protocol did not include a cMRI. Therefore, the number of silent myocardical infarctions could not be assessed and may be underestimated in PNH patients.
Also the number of patients with pulmonary infarctions may also be underestimated due to the CT being the gold standard to diagnose changes in the pulmonary tissue. However, pulmonary embolisms would have been detected with a specificity of up to 87% by the MRI as shown by Hu et al. in 2016 26 .
Recently Barcellini et al., 2017, investigated for the first time brain ischemic involvement in neurologically asymptomatic PNH patients 16 . In five of our patients microbleeds or signs for cerebral hemorrhage could be detected. However, the clinical value remains unclear, as none of these patients had a known history of traumatic brain injury.
A limitation of this study is, that none of the patients had been evaluated neurologically or psychologically by a certified neurologist prior to examination. However, only one of the patients included in this study had prior reported neurological symptoms. Like Barcellini et al., 2017, reported white matter lesions could be detected in our patients. White matter abnormalities were found in 59.5% (22/37) of our patients, whereas Barcellini et al. only reported 32%. The value of this finding remains unclear. Atrophy was also detected in 20 (54%) of our patients. However, as stated before, further psychological and neurological testing should be warranted to detect if this can be correlated to neurological and psychological functions. Also acute cerebral ischemic lesions might not be found, because a DWI was not performed. However none of the patients examined reported any acute neurological symptoms.
This study further provides evidence of silent and underdiagnosed thromboembolic complications in PNH patients. This is the first description of bone infarctions as possible TEs in PNH. WB-MRI scans thus seem to be a novel and feasible method for the assessment of the whole vascular status of a PNH patient and allow the detection of previously unidentified vascular complications and follow up on known thromboembolic complications. This might directly affect the treatment regimen as eculizumab is indicated in PNH patients with TEs. The question arises as to whether WB-MRA should be considered the gold standard at the time of PNH-diagnosis to determine eculizumab-treatment. Further studies should examine if silent TEs also have an impact on the mortality and morbidity of PNH-patients and therefore should be also considered as an indication to start treatment with eculizumab. In patients with existing TEs, WB-MRI scans might be employed to document their impact prior to treatment, especially in patients with major TEs or in patients with a high disease activity and suspicion for TEs (e.g. recurrent abdominal pain or persisting dyspnea).