Comorbidity among HHT patients and their controls in a 20 years follow-up period

The vast majority of affected individuals suffer from epistaxis and about half experience episodes beginning in childhood [13]. The severity and occurrence of bleeding episodes vary widely, as does the need for treatment and medical counselling [14].

Gastrointestinal bleedings are common among HHT patients (about 25%) and severe bleedings are more frequent in elderly patients. In some cases, the telangiectatic lesions can be identified by endoscopy and treated with laser therapy, however the success rate is often limited due to large numbers of widely scattered telangiectatic lesions [15].

A substantial part of HHT patients are affected by arteriovenous malformations (AVMs), which are found in multiple internal organs. Hepatic AVMs (HAVMs) are in screening studies observed in more than half of HHT patients, and most frequently in HHT2 patients [16, 17]. Nevertheless, severe symptomatic liver-disease is rare and is often related to high-output cardiac failure and portal hypertension [18]. Cerebral AVMs (CAVMs) are diagnosed in 11–23% of patients with HHT, most commonly among HHT1 patients [19, 20], and are associated with haemorrhage, migraine and seizures [21, 22]. About 35% of the HHT patients have pulmonary AVMs (PAVMs), occurring considerably more frequently in HHT1 patients [12, 16, 17, 23]. PAVMs can lead to dyspnoea [24], but often the patients present with asymptomatic hypoxemia [24, 25]. Even in the absence of symptoms, recognition and treatment of PAVMs is recommended [26, 27], as they compromise the filter function of the lungs, and therefore can be the cause of ischemic strokes and cerebral abscesses [23, 26‐29]. PAVM haemorrhage leading to haemoptysis or haemothorax is a rare, but potentially life-threatening complication that also can be evaded by embolisation of PAVMs [30].

An association between HHT and severe extra-cerebral infections including septicaemia, arthritis, osteomyelitis, spondylodiscitis and endocarditis have previously been reported [31‐35]. Prolonged epistaxis and nasal colonisation of Staphylococcus aureus may contribute to this association [32‐35]. Moreover, elderly fragile patients affected by comorbidities are suggested to be more prone to these infections [35]. Several abnormalities of the adaptive immune system are associated with HHT [36]. Concurrently, iron deficiency, which HHT patients often experience in relation to frequent bleedings, is correlated with a reduced oxidative burst and impaired phagocytic activity [37]. It is not yet clear though, if these alterations impair the immune response.

The frequent bleeding episodes among HHT patients lead to an increased erythropoiesis, which can cause elevated blood viscosity. The combination of increased blood viscosity and iron deficiency, which is associated with pulmonary embolism/deep venous thromboses, suggests that HHT patients could be at increased risk of thromboembolic events [38]. Furthermore, paradoxical emboli passing through untreated PAVMs to the coronary arteries can lead to cardiac ischemia [39] and have in case reports been suggested as the cause of myocardial infarction [40, 41].

At the first visit at the HHT Centre, history taking and clinical examination focusing on HHT manifestations and neurological symptoms was performed and the patients were offered screening for PAVM. The screening included contrast echocardiography and if positive either a pulmonary angiography or a chest CT- scan was offered. If a PAVM with feeding arteries larger than 2 mm was identified, the patient was offered embolisation therapy [44]. Screening of CAVM and/or HAVM is not implemented as routine in Denmark. Screening for CAVM was offered if the patient had experienced neurological symptoms or deficits. In a few cases screening was performed in patients with concerns regarding CAVM. In all HHT families genetic counselling and mutation diagnostics were offered.

At birth or upon immigration a unique civil registration number is assigned to all Danish residents. This number is used throughout in the private and public sector, including the Danish healthcare sector, for the registration of data related to a given individual. As part of the present study all relevant data on cases and controls were extracted from The Danish National Patient Register (Lands Patient Registeret, LPR). LPR is a nationwide register that since 1977 has registered information on all Danish residents hospitalised in somatic wards and from 1995 all hospitalisations and outpatient visits have been registered [45]. Data of relevance for the present study are specified in the section “Definition and grouping of relevant data” and a more detailed description can be found in Additional file 1: Appendix 1, online only. Our LPR data analysis included only hospitalisations with a date of discharge within the period from January 1st 1995 through December 31st 2014 or with an active contact as of January 1st 2015.

To validate and supplement the information acquired from LPR, all accessible medical data on patients and controls from the period January 1st 1995 through December 31st 2014 were obtained from the electronic administration-systems employed in the County of Funen during this period. The first data collection was performed in the period September–November 2014. For patients alive or recently deceased at the time of the first data-collection, a second review was performed in March 2015 to ensure inclusion of all data from the follow-up period. All hospital admissions and outpatient visits were individually reviewed to verify the consistency between the ICD-10 classification and the patient’s clinical presentation at each event, although only data regarding hospitalisation were employed for analysis, as stated above.

We applied up-to-date knowledge of HHT to the ICD10 classification, in order to categorise hospitalisations into “potentially HHT-related diagnoses” groups: infectious conditions, thromboembolic events, other vascular conditions and bleedings, including subgroups. (Table 1). All other hospitalisations were categorised as “non-HHT related conditions”. In the manually collected data, only HHT-related conditions were analysed. Here we also included all outpatient contacts involving treatment for epistaxis as these patients are often not hospitalised, and we evaluated and defined the group “Other HHT-related contacts”, which included other outpatient visits regarding HHT. In the manually collected data, secondary discharge diagnoses were included if the condition alone would have led to hospital admission. For elaborate description and classification of diagnoses, please see Additional file 1: Appendix 1, online only. To avoid bias due to individual patients with several hospitalisations of the same category, we decided to perform the analysis based only on the first registered event of a given category of contact for each individual patient.

Chi-square statistics and Fisher’s exact test were used as appropriate for the analysis of contingency tables.

Permission to extract and analyse data was obtained from the Danish Data Protection Agency (Jnr13/42552) and the Danish National Board of Health (FSEID-00001118).

A significantly increased incidence of infections in joints and bones was identified among cases in both LPR and the clinical dataset (p-values < 0.001 respectively, Table 2). In the LPR we identified six cases and one control (five cases and no controls in the clinical dataset, Table 3). None of the cases with infections in joints and bones had in the follow up period been hospitalised due to epistaxis, and four out of six had been examined for PAVMs. Three of these patients were diagnosed with PAVMs and two of them where treated accordingly, the remaining 2 patients declined screening (Table 3). We also found a significantly increased overall incidence of bacterial infections in HHT patients in the LPR data (P-value 0.015), but this could not be validated by the clinical data (P-value 0.060.) Based on the clinical data we saw a trend towards lower airway infections being more frequent among cases (P-value 0.051), but this was not evident from the LPR data.

In this study, thromboembolic events and other vascular conditions were not more frequent among cases than among controls. Notably, the cases did not present with an increased incidence of thromboembolisms in the CNS (Additional file 2: Table S4, online only.) Though, among the cases that were hospitalised for evaluation of PAVM, 11,1% (3 of 27) had suffered from a CNS thromboembolism. Only 3,7% of the age-and sex-matched controls for this group of cases presented with a thromboembolism in the CNS. This difference was not significant (Fischers exact test P value 0.16).

The percentage of cases with one or more bleeding-related hospitalisation was 60% in the LPR data and 64% in clinical data compared to 19 and 7% for controls, respectively (Table 2). Epistaxis was the most frequent type of bleeding at 37% (LPR data)/45% (clinical data) compared to 0.02% (LPR data)/0.9% (clinical data) among the controls (clinical data) (Table 2). In the clinical data all contacts involving treatment for epistaxis were included. The second most common type of bleeding was “Other bleedings” at 23% (LPR data)/33% (clinical data) compared to 7.3% (LPR data) and 1.8% (clinical data) among the controls (Additional file 2: Table S4, online only). These bleedings were documented bleedings from other organs than GI or epistaxis, or cases where the source of the bleeding was not identified.

The category “Possible bleedings” represented mainly hospitalisations due to anaemia caused by bleeding. This was registered in 31% (LPR data)/ 22% (clinical data) of HHT cases compared to 5.5% (LPR data)/0% (clinical data) of controls (Additional file 2: Table S4, online only). The occurrence of bleeding from the gastrointestinal tract was only documented in 6.8% (LPR data)/14% (clinical data) of cases compared to 1.8% (LPR data)/4.6% (clinical data) of the controls (Table 2). Only spontaneous bleedings were included in the study, and bleedings related to pregnancy/childbirth/abortion or other gynaecologic/obstetric conditions were not included.

According to the clinical data, 37% (27 cases) were admitted for evaluation of PAVMs in the follow up period. We know from previous data on the same case-group that 31% (23 cases) had PAVM identified on a CT scan, pulmonary angiography or during surgery [23]. Only 8% (6 cases) where registered with a PAVM diagnosis in the LPR.

The HHT patients and their controls did not differ concerning the initial incidence of non-HHT-related hospitalisations. This was only assessed in the LPR data (Additional file 2: Table S4, online only.)

It is assumed that HHT patients have an increased risk of severe bacterial infections, and, in particular, the occurrence of cerebral abscesses is increased among HHT-patients with untreated PAVMs [27‐29]. The cases included in this study did not present with cerebral abscesses during follow-up. This is expected as a large part of the cases were examined for PAVMs, followed by embolisation of treatable PAVMs. We found indications of increased incidence of infections in general based on the LPR data, but this was not validated in the clinical data. We did, however, detect a significantly increased incidence of infections in joints and bones among the cases compared to controls as previously described [31‐35]. These infections are often severe and may require weeks of intravenous antibiotic treatment, underlining the need to prevent their occurrence.

Only few cohort studies have investigated the occurrence of infections among HHT patients. In one of the larger studies [33], it is suggested that HHT is associated with a high frequency of severe infections, defined as any infection that required hospitalisation of the patient. Rather than using a matched control group, they compared their findings to the incidence of sepsis in the United States, and the incidence of S. aureus bacteraemia in Europe. Even though the frame of reference for these outcomes may be discussed, the authors found an interesting correlation between median duration of epistaxis, and extra-cerebral infections, leading the authors of the study and those of two case-reports [34, 35] to suggest that nasal colonisation of S. aureus and frequent, prolonged epistaxis, generates a gateway for haematogenous dissemination of bacteria to other organs. However, in our study none of our cases with infections in joints and bones had been hospitalised due to epistaxis, which speaks against severe manifestation of this phenotypic event. Unfortunately, any correlation to nasal S. aureus colonisation cannot be investigated in the current study design. Furthermore, we did not find any clear correlation between PAVMs and infections in joints and bones as three out of the five cases with these infections (LPR/clinical data) had experienced hospitalisation for evaluation of PAVM, compared to 37% out of all the HHT patients. It is also suggested that HHT patients may have an impaired immune response. This is supported by other studies, linking iron deficiency to alterations in the immune response of HHT patients.This might be part of the explanation for our findings, however further studies are needed in order to address this. Furthermore, increasing age could be a factor inflicting the occurrence of infections in HHT, but as our cases and controls are age matched this cannot explain the difference in this study.

HHT-patients with untreated PAVMs have an increased risk of ischemic strokes, which is strongly associated with the grade of right-to-left shunting [47]. Additionally, it is found that ischaemic strokes are more common in patients with iron deficiency, suggesting this to be due to enhanced platelet aggregation [48]. The same study suggests that iron deficiency might be an independent risk factor in the general population. Another study finds an association between low serum iron levels, and elevated plasma levels of coagulation factor VIII and venous thromboembolic risk [38]. Elevated plasma levels of FVIII are considered a strong risk factor for VTE in the general population [49]. However, despite being prone to an increased risk of iron deficiency, the cases in our study did not have an increased prevalence of thromboembolic events.

We did not find an increased incidence of ischemic stroke in neither the total HHT patient group nor in the subgroup of HHT patients hospitalised for evaluation of PAVM. In this study cases with PAVMs larger than 2 mm were treated with embolisation in the beginning of the follow-up period. These results imply that treatment of PAVM prevents ischemic stroke by minimising right-to-left shunting. We do not know the conventional stroke risk factors for this cohort, so we need to consider that the cases, to cope with their condition, might have adapted to a healthier lifestyle compared to controls. The careful surveillance and counselling of these patients might also contribute to why this study does not support the theory proposed in the above mentioned papers.

Bleedings are invalidating for many HHT patients, which is supported by the results of this study. Thus, we found that 64% (clinical data) were treated in the hospital due to bleedings, compared to only 7% of the controls. It is not surprising that epistaxis is the most frequent type of bleeding, and 45% have been treated in the hospital due to this manifestation. In a survey carried out on a US population [50] bleeding-related complications were found to be the most frequent manifestations in hospitalised HHT patients, as they were recorded in 63% of the hospitalised patients. These findings support the results of the current study, stating that bleeding-related manifestations, including anemia, are the most common underlying reason for hospitalisation in HHT-patients. Only 14% of cases (clinical data) were hospitalised due to gastrointestinal bleedings - this is in accordance with the findings of Brinjiki et al. [50]. In a previous study [15] we found that 33% of the current patient population had a history of hematemesis or melena before 1997. This study was based on the patient’s own experience, presented at interview and not restricted to hospitalisation in a specific period of time as in the present data set. The large number of hospitalisations due to anaemia of unknown cause, may partly represent GI bleedings were the source of bleeding was not located.

In this study, 37% of HHT patients were admitted due to PAVM or PAVM-examination according to the clinical data. As mentioned, we know from previous studies that 32% of the cases in this cohort had PAVMs – the discrepancy between this and our clinical data is caused by a few patients that were admitted for PAVM examination, but where no PAVMs were validated. The prevalence of PAVMs in this study is equivalent to the prevalence suggested in internationally published data and to those found in other studies [23, 51‐54]. The low rate of comorbidities for this case group can therefore not be explained by a PAVM prevalence deviation. A pronounced registration bias is evident from the LPR data as only 8% of the cases are here reported hospitalised due to PAVMs. This emphasises the importance of manual validation of data reported to central administrative health registers.

More than half of the cases had hospital contacts for other reasons than bleedings, infections, thromboembolic events and vascular malformations. These other HHT relevant contacts dealt with information on the HHT disease, examination prior to surgery and questions about HHT’s impact on other health issues. Frequent hospitalisations have a negative impact on quality of life, and therefore further research should investigate how HHT affects the patients’ quality of life.

The cases are not hospitalised more frequently than the controls due to non-HHT-related conditions. This finding validates our diagnoses and diagnosis groups as being potentially HHT relevant or not.