Background Historically, the survival of patients with von Hippel-Lindau disease (vHL) has been poorer than that of the general population. We aimed to determine whether the survival of VHL mutation carriers and their risk of vHL-related death has changed over time and how it has been affected by sex, genotype and surveillance attendance.
Methods In a retrospective cohort study, we included all known Danish vHL families with a VHL mutation. We assessed the survival and causes of death for 143 VHL mutation carriers using Cox regression models and compared vHL survival with that of 137 siblings without vHL. vHL life expectancy was compared with the general population using a relative survival model.
Results The estimated mean life expectancies for male and female patients born in 2000 were 67 and 60 years, respectively. Survival is influenced by the sex and genotype of the patient. Female patients have a significantly higher risk of vHL-related death than male patients (HR=2.25, 95% CI 1.20 to 4.20, p=0.011). Overall, 79% (53 of 67) of deaths were vHL-related, but the risk of vHL-related death has decreased over time, as has the frequency of renal cell carcinoma (RCC)-related death. Surveillance is especially beneficial for truncating mutation carriers, who have the greatest RCC and central nervous system (CNS) hemangioblastoma risk.
Conclusions vHL survival has improved over time and has become closer to that of siblings without vHL and the general population. Even though the risk of vHL-related death has decreased significantly, the main cause of death is still CNS hemangioblastomas and hence improved treatment options are essential.
- von Hippel-Lindau disease
- Clinical genetics
- Life expectancy
Statistics from Altmetric.com
von Hippel-Lindau disease (vHL) (OMIM no. 193300) is characterised by a genetic predisposition to develop cysts and tumours in multiple organs, most commonly hemangioblastomas in the central nervous system (CNS) and retina, renal cell carcinoma (RCC) and pheochromocytoma.1 ,2 vHL is autosomal-dominantly inherited and caused by germline mutations in the VHL gene.2 The gene's primary function is the regulation of a group of transcription factors, hypoxia-inducible factors (HIFs), that are normally kept in check by the VHL protein through controlled degradation.1 ,2 HIFs accumulate when VHL function is defective and facilitate overexpression of target genes involved in angiogenesis and cell proliferation.2 Manifestations are often asymptomatic at early stages but can cause severe sequelae such as blindness, neurological complications or early death if not diagnosed and treated appropriately.1 ,2 No systemic treatment exists and VHL mutation carriers are advised to undergo surveillance to ensure early diagnosis and surgical treatment.1 ,3 ,4 Historically, vHL survival has been poorer than that of the general population. The largest study of vHL survival to date reported a life expectancy of 49 years, but the estimate was made more than 25 years ago.5 Since then, a few studies have reported improved life expectancies of up to 64 years, but the estimates cannot be directly applied to the average patient with vHL, as they are based on selected vHL cohorts.6 ,7 Previous studies of vHL-related mortality have merely reported the frequencies of the vHL-related causes of death but have not assessed the risk of dying from the disease.5 ,8–13 Implementation of surveillance during recent decades is believed to have markedly improved survival9 but the effect has not been directly assessed. The reported low life expectancies and sparse information about survival and mortality can cause distress for vHL families, complicate genetic counselling and create problems with life and health insurances. The vHL phenotype is variable, in part owing to genotype-phenotype associations2 but other factors such as age, sex, and pregnancy have been suggested to influence the natural course of the disease.14–17
We aimed to determine whether the survival of VHL mutation carriers and their risk of vHL-related death have changed over time and how vHL survival compares with the survival of family members without vHL as well as to the survival of the general population. We further explore how survival is affected by sex, genotype and surveillance attendance.
In a retrospective cohort study, we identified all known Danish vHL families with a confirmed pathogenic VHL germline mutation. Follow-up ended on 1 January 2016. We included 143 patients with vHL (67 deceased and 76 living; 74 women and 69 men) and 137 siblings without vHL (48 deceased and 89 living; 64 women and 73 men) from 34 unrelated families. The individuals were born between 1841 and 2010. We excluded individuals of 5 years or younger at the time of death/end of follow-up, as we could not determine their vHL status in the earlier generations due to the negligible incidence of symptomatic vHL manifestations before this age.5 ,10 ,18 Of the included patients, 89% (127 of 143) had at least one manifestation, the rest were asymptomatic mutation carriers. We included the patients' entire lifetimes until death/end of follow-up, median included time was 44 years/person (range 6–89 years), in total 6 075 person-years. Family pedigrees were drawn using information from the Danish Civil registration system as well as records of historic population censuses and church records. Individuals were classified as patients with vHL if they (1) had a VHL germline mutation (n=89) and/or (2) fulfilled the clinical vHL criteria (n=48)3 or (3) were obligate mutation carriers (no diagnosed vHL manifestations but had an affected parent/siblings and affected children/descendants, n=6). The vHL status of family members who had not themselves been genetically tested was assessed from their detailed medical histories. These were obtained through interviews with living family members and review of old medical records and interviews conducted with now deceased individuals in 1930 and 1952 (information about 21 members of two families),19 ,20 and from 1987 to 1993 (information about 65 members of 13 families). Further, we evaluated medical records, autopsy reports, data from Danish national health registries (the Patient Register, the Cancer Register, and the Death Register) and Death certificates. Family members without vHL had either (1) a negative VHL mutation analysis (n=59) or (2) no reported vHL manifestations and no affected descendants (n=78). We excluded individuals with no or unspecific information about their vHL status and/or date of birth or death. We obtained data about survival and death rates in the general Danish population from 1900 to 2014 (data from 2015 was not available).21 Table 1 describes the included individuals' characteristics.
The study was approved by the Danish Data Protection Agency (Journal no. 2009-41-3994) and the Danish Committees on Biomedical Research Ethics (H-2-2010-012).
We assessed how vHL survival had developed over time using Kaplan-Meier curves in groups of patients born in different periods: (1) between 1840 and 1900, (2) between 1900 and 1955 and (3) between 1956 and 2010. The three groups were chosen to reflect periods with different practices of adulthood vHL surveillance. Patients born before 1900 had no surveillance, 17% had regular surveillance among those born between 1900 and 1955, while the majority of patients born after 1955 (88%) attended regular surveillance (see online supplementary table S1). We then compared the three groups' survival times in a Cox regression model with robust SEs allowing for variation between families.
vHL patients: Frequency of surveillance attendance
To assess whether the improved survival among the younger patients was primarily due to a general improved life expectancy, we compared vHL survival with the survival of their non-vHL siblings for individuals born after 1900. Survival times were illustrated with Kaplan-Meier curves and compared using the Cox-regression model with birth year, sex and vHL status as covariates. Furthermore, we compared sex-specific and age-specific death rates for Danish birth cohorts from 1900 to 201421 with the survival times of patients with vHL using a relative survival model.22 Thus, the sex-specific and age-specific death rate of patients with vHL from a given birth cohort (RvHL) was assumed to be the sum of the death rate observed for an individual in the general population of the same sex and born in the same year (RDK), and a rate (RExcess) describing the excess mortality in patients, that is: RvHL=RDK+RExcess. In this model, we explored whether the excess death rate depended on birth year and sex. To illustrate the development in life expectancy over time, we used the relative survival model to determine the mean remaining lifetime at age 10 in different years. We chose the age of 10 to minimise the possible bias of childhood mortality from other causes. The mean remaining lifetimes for patients with vHL and individuals from the general population were plotted for different birth years. The calculation of mean remaining lifetime was based on the period life table method, where age-specific death rates in a given year was estimated from the model and then the mean lifetime for that year was calculated by assuming that the birth cohort will have these death rates throughout their lives.
We assessed the effect of birth year on survival (hazard of death) in a Cox regression model with robust SEs allowing for variation between families, where birth year was included as a continuous covariate. In this and all the following Cox regression analyses, we excluded individuals born before 1900, as identification of patients with vHL before this time may be biased towards including the most severe cases. We further explored whether vHL survival is affected by sex, genotype and surveillance attendance in the same Cox regression model. Each family's genotype was characterised as either a missense mutation (62 mutation carriers born after 1900 from 15 families) or a truncating mutation (nonsense, frame-shift, splice-site mutations and exon deletions) (70 mutation carriers born after 1900 from 19 families; online supplementary table S2). We defined surveillance attendance as periods of time, where a patient attended examinations performed with the specific purpose to diagnose CNS and/or abdominal lesions for at least three consecutive years. Based on each patient's medical records, we recorded the dates of initiation and end of each surveillance period.
Genotypes of the included vHL families
We added sex, genotype and surveillance attendance as covariates to the Cox regression model estimating hazard of death (from all causes) as well as to a similar Cox model estimating the cause-specific hazard of vHL-related death. To allow for improved survival in later cohorts, birth year was included as an additional covariate. While sex and genotype are constant throughout life, we assessed the effect of surveillance attendance using an age-dependent covariate that indicated any reduced risk that each patient's periods of surveillance entailed. We also explored whether a patient's genotype influenced the effects of sex or surveillance attendance by separately evaluating interaction between genotype and sex as well as genotype and surveillance attendance.
We evaluated each deceased patient's cause of death based on information from the Death register/certificate. In addition, autopsy reports and/or medical records from around the time of death were also assessed in the majority of cases (57%, 38 of 67 deaths). Each death was classified as being either vHL-related (n=53), vHL-unrelated (n=10) or due to unknown causes (n=4, all patients born before 1900). Deaths due to surgical complications or postoperative sequelae were classified as vHL-related, as they would not have occurred if the individual had not had vHL.
Analyses were conducted in SAS (V.9.4). Statistical significance was taken at 5%.
vHL survival has improved over time as illustrated by the higher survival frequencies for patients born later than 1955 compared with patients born before that time period (online supplementary figure 1). Accordingly, the Cox model demonstrated that birth year had a highly significant effect on survival; the later the patient was born, the lower was the hazard of death (HR for each year later a patient is born after 1900=0.98, 95% CI 0.97 to 0.98, p<0.001).
Age-related survival of vHL patients born in different periods
Compared with the general population, the excess vHL mortality has decreased over time; the later a patient was born, the closer his or her survival was to that of the general population (table 2 and figure 1). Overall, the excess mortality of patients with vHL was decreased by 2.93% with each later birth year (95% CI 1.64% to 4.21%, p<0.001). This is illustrated in the estimated mean lifetimes for female and male patients with vHL, compared with that of women/men born in the same year in the general population (table 2 and figure 1). Female patients with vHL had an excess mortality compared with male patients (HR=3.60, 95% CI 1.27 to 10.23, p=0.016).
Patients with vHL also had a poorer survival than their non-vHL siblings (figure 2, see online supplementary table S3). For male patients, the death rates were increased by a factor of 2.25 (95% CI 1.02 to 4.96, p=0.045) compared with their male siblings, while the effect was stronger for women, whose death rates were increased by a factor of 8.09 (95% CI 4.88 to 13.4, p<0.001) compared with their female siblings.
vHL patients compared to their siblings without vHL: Effect of birth year, sex, genotype, and vHL status on the hazard of death
Overall, 79% (53 of 67) of deaths among patients with vHL were vHL-related, most commonly CNS hemangioblastomas (51%, 27 of 53) and RCC (36%, 19 of 53), table 3. We included three suicides in the analysis, as an exclusion of suicides was assessed to bias the results; two of the suicides would most likely not have occurred if the patients had not had vHL. Even though all of the deceased patients born after 1955 had died due to vHL-related causes, the risk of dying from vHL was lower the later a patient was born. For each birth year the patient was born later than 1900, the risk of vHL-related death was decreased by 2.28% (95% CI 1.57% to 2.66%, p<0.001).
vHL survival depends on sex and genotype (table 4 and see online supplementary tables S4 and S5). Female patients had significantly higher death rates than male. There was no effect of genotype when male and female patients were considered as one group. The effects of sex and genotype were similar whether we considered risk of death from any cause or just vHL-related death. In both models, we found a significant interaction between sex and genotype (p=0.006 (risk of death from any cause) and p<0.001 (risk of vHL-related death)), meaning that the effect of a patient's genotype on survival depended on the patient's sex. Thus, female patients with vHL with a missense mutation had a significantly poorer survival than male missense mutation carriers (HR=4.75, 95% CI 1.66 to 13.61, p=0.004). Meanwhile, there was no significant sex difference for truncating mutation carriers (HR 1.15, 95% CI 0.51 to 2.62, p=0.73). The death rates of female missense and truncating mutation carriers were not significantly different (HR=1.28, 95% CI 0.78 to 2.11, p=0.32). In contrast, male missense mutation carriers had significantly lower death rates than those with truncating mutations (HR=0.26, 95% CI 0.08 to 0.79, p=0.0179). The effects were even larger if only the risk of vHL-related death was considered. Table 4 shows the effects of the covariates in Cox models where all four covariates are included and online supplementary tables S4 and S5 show the separate effects and interactions of the covariates in different Cox models.
Overall, about half (79 of 143) of the patients with vHL had attended surveillance; the majority were born after 1955 (see online supplementary table S1). We assessed the effect of attending surveillance and found no significant differences in risk of vHL-related death. The results were similar when we considered overall survival (table 4). Nevertheless, we found a significant interaction between surveillance and genotype (p=0.0001). Truncating mutation carriers who attended surveillance had significantly lower death rates than those who did not (HR=0.22, 95% CI 0.07 to 0.69, p=0.0095). In contrast, there were no significant differences in risk of vHL-related death for missense mutation carriers who attended surveillance and those who did not (HR=1.63 (95% CI 0.52 to 5.15), p=0.41; see online supplementary tables S4 and S5). For four patients with vHL (one born before and three born after 1955), we had no information about surveillance attendance. The effect of surveillance changed only slightly if they were included as surveillance attendees (data not shown).
Effect of birth year, sex, genotype, and surveillance on the hazard of death (from any cause)
Effect of birth year, sex, genotype, and surveillance on the cause-specific hazard of vHL-related death
vHL survival has improved significantly and is getting closer to that of non-vHL siblings as well as the general population. The estimated mean life expectancy of a male patient with vHL born in 2000 was 67 years, corresponding to about 10 years less than that of a man in the general population born the same year. Meanwhile, a female patient with vHL born in 2000 has an estimated mean life expectancy of almost 60 years, about 22 years less than a woman in the general population. Median life expectancy has previously been evaluated in an unselected vHL cohort (n=152) to be 49 years.5 The cohort consisted solely of patients with a clinical diagnosis of vHL,5 which could have biased the estimate towards inclusion of the most severely affected patients. At the time of the estimate, the effect of recently initiated surveillance programmes was probably not yet evident. Presently, the vHL diagnosis is often made through pre-symptomatic genetic testing3 ,4 ,24 and we included VHL mutation carriers regardless of phenotype. We showed that vHL survival is highly dependent on a person's birth year; the younger he or she is, the longer is the life expectancy. In line with our results, another study found a significantly higher life expectancy for patients with vHL diagnosed after 1990 (n=40, 59.4 years, 95% CI 48.1 to 70.8) compared with those diagnosed before (n=16, 43.1 years (95% CI 39.2 to 47.0), p=0.02).9 They attributed the improved survival to surveillance, but did not directly assess surveillance effect.
We found that survival is influenced by sex and genotype. Female patients with vHL had a mortality hazard that was twice as high as male patients. Both the effects of sex and surveillance on survival significantly depend on genotype. The sex differences are especially evident for VHL missense mutation carriers, where the female death rates were almost five times as high as the male. The influence of sex on vHL natural history has previously been explored.8 ,9 ,15 ,25 Consistent with our results, one study found a poorer life expectancy for female patients compared with male (median life expectancy: men: 59.4 years (n=35)/women: 48.4 years (n=33)).9 The result was not significant (p=0.953), perhaps due to a small sample size.9 In contrast, we recently found a tendency for men to develop new vHL manifestations at higher rates than women.15 Another group found male patients to have a significant higher overall CNS tumour burden and higher rates of tumour growth.8 Hormonal effects may play a role in vHL tumorigenesis, although the mechanisms are still unclear. The difference between male and female patients' risk of new tumour development has been found to be largest after the patients' mid-20s, coinciding with the women's fertile period.15 We and others have shown that pregnancy does not increase vHL tumour development or growth and may even be associated with lower tumour risks.15 ,16 In contrast, another study found pregnancy to induce cerebellar hemangioblastoma growth, but the number of examined tumours was small.17
The reason why male patients with vHL live longer despite an apparently more severe phenotype is unclear, but sex-specific somatic genetic alterations or epigenetic changes may lead to different patterns of tumour progression in men and women, which could give rise to more aggressive tumours in women. Recently, different sex-specific patterns of mRNA expression profiles and somatic mutations in genes that may modify VHL-associated tumour phenotypes have been identified in sporadic RCC.26 Accordingly, sex-specific patterns of somatic gene inactivation and activity in cell signalling pathways have been found in CNS gliomas.27 ,28 Sex differences could be important in the future treatment approaches, both surgical and pharmacological, and needs to be further evaluated. Any differences in the quality of life or long-term sequelae of male and female patients with vHL remain to be evaluated.
Even though the risk of dying due to vHL has decreased, the main cause of death is still vHL-related illness with CNS hemangioblastoma accounting for 40% (27 of 67) of deaths. Overall, nearly a third of the deceased patients with vHL (19 of 67) died due to RCC. However, the frequency of RCC-related death decreased from 42% (18 of 50 deaths) among patients born between 1900 and 1955 to 17% (1 of 6 deaths) among patients born after 1955. The patient groups are not fully comparable, as RCC risk is age-related and the median age of the living patients born after 1955 was just 35 years (range: 6–59 years).5 Nevertheless, RCC is most commonly asymptomatic before metastasising and the notable decrease in RCC-related deaths must largely be attributed to surveillance and subsequent early treatment, which reduces risk of metastatic disease.29 Other studies have reported that 13%–47% of patients with vHL die from RCC and 25%–62% die from CNS hemangioblastoma (see online supplementary table S6).5 ,6 ,8–13 However, a shift seems to have occurred from RCC and CNS hemangioblastomas, being almost equally frequent killers, towards CNS tumours as the predominant cause of death (see online supplementary table S6). The fact that all deaths among patients born after 1955 were vHL-related, and almost all were caused by CNS hemangioblastomas, is in part due to the low age of many patients in this group, but it also underlines the importance of the ongoing development of improved treatment options, both pharmacological, and surgical.2 Particularly, CNS hemangioblastoma treatment entails substantial risks of sequelae and perioperative/postoperative death.25 Many of the deaths classified as being CNS-related in our cohort (59%, 16 of 27) occurred during CNS surgery or due to long-term postoperative sequelae.
Reported causes of death in vHL cohorts from the literature
We present the first assessment of the effect of surveillance on vHL survival. We categorised the patients' lives into periods with and without surveillance attendance and looked for any age-related excess mortality in the non-surveillance periods. Patients with truncating mutations benefited more from surveillance than those with missense mutations. The differences in surveillance benefit according to genotype is most likely due to surveillance-facilitated decreased RCC mortality in truncating mutation carriers, who are known to have a higher RCC risk compared with missense mutation carriers.30–32 Further, truncating mutation carriers have more CNS tumours and develop new tumours at higher rates than missense mutation carriers.8 ,14 Our failure to demonstrate a surveillance effect for missense mutation carriers' survival could in part be because many patients did not initiate surveillance before adulthood. The overall median age of surveillance initiation was 24 years (0–67 years) (see online supplementary table S1). Before the first national Danish surveillance guideline was published in 2005,33 surveillance referral was not always consistent. Our strict definition of surveillance limits the number of patients with surveillance attendance, as patients who attended sporadic CNS or abdominal screenings were not counted. Nevertheless, many knew about their vHL diagnosis, which could have facilitated earlier manifestation diagnoses, giving a higher survival in the non-surveillance group. As pre-symptomatic genetic diagnosis is more prevalent, patients with familial vHL initiate surveillance in childhood, which will contribute to a more pronounced survival improvement. We did not assess whether surveillance has lowered the occurrence or extent of vHL-related sequelae or improved the patients' quality of life. A recent study found that most currently living patients have only minor vHL sequelae and are able to cope with every day activities.25
The use of national health registers, interviews of family members in several generations and medical record verification (clinical information verified in 94% (134 of 143 patients)) has enabled a detailed analysis of vHL survival. In our Cox regression analyses, we excluded individuals born before 1900, as the collected data from this period may be biased towards inclusion of the most severely affected patients. The individuals born before 1900 came from merely three different families and there was an unequal distribution of patients with vHL (n=11) and non-vHL siblings (n=2). We solely included families in which a VHL germline mutation had been identified, but not all the included individuals had themselves been genetically tested. Nevertheless, we believe that the classification into groups of patients with vHL and non-vHL siblings is appropriate. The patients with vHL (n=132) and non-vHL siblings (n=135) born after 1900 are of comparable sizes, as expected for an autosomal-dominant disease. As vHL is assumed to be underdiagnosed in Denmark,18 we may have missed yet undiagnosed patients, who possibly have milder phenotype or are de novo cases.
In a national cohort study including all known Danish VHL mutation carriers, we show that vHL survival has improved over time and that the life expectancy of patients with vHL has become closer to that of their siblings without vHL and that of the general population. vHL survival is improved with birth year, male sex and a missense mutation genotype. Surveillance seems to be especially beneficial for truncating mutation carriers. Surveillance is expected to have improved vHL-related sequelae and have a stronger effect on survival if initiated in childhood.
The authors especially thank all participating patients with von Hippel-Lindau (vHL) disease and their families. They also thank all the participating hospital departments for helping them locate and copy, sometimes very old, medical records. They appreciate the help of the administrators of the national Danish Health Registers, Statens Serum Institut (SSI), for providing us with register data. They thank Michael Galanakis for his great help in organising and sorting the register data and participating in statistical discussions. They are grateful to Kai Albrechtsen, who started the first collection of information about Danish vHL families and gathered valuable medical information about the oldest Danish vHL families. They also thank Thomas Rosenberg who followed up on Kai Albrecthsen's material, collected information about many more vHL families and initiated the first official national vHL register in Denmark. They appreciate the help from Hans Ulrik Møller, who provided valuable information about the oldest generations of the two largest Danish vHL families in the form of his father's and grandfather's detailed papers from 1930 and 1952.19 ,20
Contributors MLB and AMJ had the idea for the study. MLB and MLMB initiated the study and designed the overall study structure. EB-J, MLMB and MLB together planned the statistical analyses, which were performed by EB-J. AMJ and MLMB interviewed the participating subjects, collected and evaluated their medical records and register data. MLMB recorded the events of interest and sorted the data. MLB monitored the data collection. All authors interpreted the results. MLMB drafted the first version of the manuscript; EB-J drafted the section describing the statistical methods. All authors critically revised the draft paper and approved the final version of the manuscript.
Funding The study was financially supported by the Danish Cancer Society (R20-A968-10-S2), the Lundbeck Foundation (R48-A4839), Anna and Hans Steffensen's Foundation to support Danish Cancer Research and the Danish association of patients with vHL and their relatives.
Competing interests None declared.
Ethics approval The Danish Regional Committees on Biomedical Research Ethics for the Capital region of Denmark (H-2-2010-012).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.