Cardiovascular malformations and other cardiovascular abnormalities in neurofibromatosis 1

INTRODUCTION

Neurofibromatosis 1 (NF1) is a distinctive autosomal dominant disorder (MIM 162200) with multisystem involvement [Rubenstein and Korf, 1990; Huson and Hughes, 1994; Friedman et al., 1999]. The type and frequency of cardiovascular abnormalities is unclear, although some individuals with NF1 have a vasculopathy consisting of peripheral arterial stenoses and ectasias [Rubenstein and Korf, 1990; Riccardi, 1992; Huson and Hughes, 1994; Friedman, 1999].

The type and incidence of cardiovascular malformations (CVMs) in NF1 have not been well-defined [Lin and Garver, 1988; Friedman and Birch, 1997; Friedman, 1999]. The frequency of CVMs in 8 large series of NF1 patients ranged from 0.4–6.4% (Table I). Often, it was unclear whether the diagnosis of a CVM was well established in these patients, whether the diagnosis of NF1 was certain, or whether a related dysmorphic syndrome (Watson, NF1-Noonan) was present. Despite the vagaries of reporting, there is a predominance of pulmonic stenosis, often valvar, in these series. Likewise, smaller series and individual case reports frequently noted pulmonic stenosis. The single “NF” patient reported with a complex CVM [Stoll et al., 1995] did not have NF1 substantiated and may have had a multiple congenital anomaly syndrome related to parental consanguinity.

Table I. Literature Review of Cardiovascular Malformations in NF1 Patients*

Author	Year	No. of NF1 patients	No. of CVM patients	%	Age	Gender	Comment
Large series:
Crowe et al.	1956	223	1 ASD	0.4%	37	M	The single patient was clinically diagnosed.
Kaufman et al.	1972	NS (19 families)	6 total:				None had features of Noonan or Watson
			5 PSV		10	M	syndrome.
					53	M	Unclear if leaflets abnormal; annulus definitely small.
					4	M	May also have had thickening of aortic valve leaflets.
					3	F	Brother (1 mo): COA, ASD. No CLS to diagnose NF1.
							Sister (8): Clinically diagnosed VSD. 3 CLS.
							Brother (7): Clinically diagnosed VSD, CLS, MR.
					18	M
			1 SVAS		6	M	Paternal uncle with NF1 and PVS.
Neiman et al.	1974	78	5 total:	6.4%
			Family				May represent NF1-Noonan syndrome.
			1 PSV		2	M	Brother. Also, MR, VPI.
			1 COA, BAV		4½	F	Sister. Also, MR, “Turner phenotype”.
			1 PSV		>18	F	Mother. Also, MR, “Turner phenotype”.
			Unrelated cases
			1 VSD			M	Location of VSD NS.
			1 ASD		6	M	Location of ASD NS, probably secundum.
			(1 CHB)		5	F	Exclude since there is no underlying CVM.
Holt	1978	NS	3 NS	NS	NS	NS	Supplements preceding series of Crowe et al., and Neiman et al.
Carey et al.	1979	131 (60 families)	2 NS	1.5%	NS	NS
Schorry et al.	1989	78	2 NS	2.6%	NS	NS
Colley et al.	1996	453	9 PS	2.0%	NS	NS
		(235 families)	(3 PS)	(0.7%)	NS	NS	(After excluding 4 patients with NF1-Noonan syndrome and 2 patients with Watson syndrome).
Tonsgard et al.	1997	35	3 total:	8.6%			From general NF1 cohort, 70 had features of deletions.
		(406 total)	2 PPS		10, 14	NS	35 (26 families) were studied, 4 had deletions.
			1 VSD, NOS		11	NS
Small series or individual case reports:
Zoethout et al.	1964	1	1 ASV	NS	39	M	1 NF patient in a series of 126 patients with AS.
Rosenquist et al.	1970	NS	1 RVOT membrane	NS	9	F	Right ventricular hypertrophy probably secondary to RVOT, not a primary cardiomyopathy.
Pernot et al.	1971	NS	3 PSV	NS	9    2 1	F F F	All 3 had CLS, none had neurofibromas. Unclear if NF1, NF1-Noonan syndrome, Watson. Fourth patient described as having lentiginosis.
Wille et al.	1980						See Table III for description. Unknown if “aortic obstruction” was valvar (ie., a CVM), or subaortic hypertrophy (ie., HCM).
Shimada et al.	1981	1	1 ECD	NS	24	M
Leppig et al.	1994	5	1 ASD, NOS		5	M	All 5 patients had dysmorphic features, MR, and NF1 gene deletions.
Wu et al.	1995	4	1 PS murmur		16	M	All 4 patients had dysmorphic features, MR, and NF1 gene deletions.
Stoll et al.	1995	1	1 dTGA, TriAtr, PAS, ASD, VSD		birth	F	NF1 not substantiated. CLS, no NFs. Multiple anomalies (microcephaly, dysmorphic face, scoliosis, short digits), pheochromocytoma. Parents first cousins.
Fischberg et al.	1996	1	1 Vascular ring	NS	birth	M	Esophageal compression due to both vascular ring and disseminated neurofibromatosis (including atrial septum and auricular appendages).

• * Not included are patients with segmental or fusiform “coarctation” of the abdominal aorta, which is considered part of the NF1 vasculopathy. AS, aortic stenosis; ASD, atrial septal defect; ASsub, sub-aortic stenosis; ASV, aortic stenosis, valvar; BAV, bicuspid aortic valve; CHB, complete heart block; CLS, café au lait spots; COA, coarctation; CVM, cardiovascular malformation; dTGA, dextro-transposition of great arteries; ECD, endocardial cushion defect; F, female; M, male; MR, mental retardation; NF, neurofibromatosis; NS, not specified, not available or not applicable; PAS, pulmonary artery stenosis; PS, pulmonic stenosis; PPS, peripheral pulmonic stenosis; RVOT, right ventricular outflow tract obstruction; TriAtr, tricuspid atresia; VPI, velopharyngeal insufficiency; VSD, ventricular septal defect.

It might be hypothesized that the only NF1 patients with cardiac abnormalities are those with the Watson syndrome or NF1-Noonan syndrome. Pulmonic stenosis is one of the cardinal clinical features of Watson syndrome [Watson, 1967; Allanson et al., 1991] and may occur in patients with the NF1-Noonan syndrome [Sharland et al., 1992], although most patients with NF1-Noonan syndrome do not have a CVM.

Although Watson syndrome is an allelic variant of NF1, the molecular basis of NF1-Noonan syndrome seems diverse. Several cases of the NF1-Noonan syndrome result from an NF1 mutation [Tassabehji et al., 1993; Kayes et al., 1994; Stern et al., 1995; Colley et al., 1996], but most cases of Noonan syndrome do not have NF1. Noonan syndrome is a distinct disorder (with a gene locus at 12q22 in some families), and valvar pulmonic stenosis is also the most common CVM in this condition [Sharland et al., 1992; Marino et al., 1999]. Further cardiac overlap with NF1 is suggested by the occurrence of hypertrophic cardiomyopathy in approximately 10–20% of patients with Noonan syndrome [Sharland et al, 1992; Marino et al., 1999].

A third group of NF1 patients who might be considered more likely to have a CVM are those with large deletions. They often have dysmorphic features and mental retardation, and 3 have been reported with a CVM [Leppig et al., 1997; Wu et al., 1995; Wu et al., 1997; Tonsgard et al., 1997; Riva et al., 2000]. All of these patients (Watson syndrome, NF1-Noonan syndrome, NF1 microdeletion syndrome) are part of the NF1 population, and their association with CVMs provides evidence that NF1mutations can cause CVMs.

Less frequent than CVM are reports of hypertrophic cardiomyopathy (HCM) in NF1 patients (Table II). Miscellaneous cardiac abnormalities that have been reported in association with NF1 (Table III) include assorted intracardiac tumors, mitral valve prolapse and aortic dilation. Strictly speaking, these latter two abnormalities may be regarded as valve and arterial dysplasia distinct from cardiac malformations.

For this analysis of CVMs and other cardiac abnormalities in patients with NF1, we used the National Neurofibromatosis Foundation International Database (NNFFID), an international multicenter collaborative system for collecting demographic information, descriptions of signs and symptoms, basic measurements, and certain psychological information on individuals and families with NF [Friedman et al., 1993]. The NNFFID was established and is maintained by the National Neurofibromatosis Foundation. Twenty-four clinics throughout the world voluntarily enter data into the NNFFID. Any qualified investigator can access these data for use in appropriate studies by contacting the authors.

A preliminary analysis of data from the NNFFID suggested that CVMs, especially pulmonic stenosis, are more common in NF1 [Friedman and Birch, 1998]. We present further descriptive analysis of 97 patients with verified NF1 contributed to the NNFFID between 1991 and 1998.

PATIENTS AND METHODS

As described previously [Friedman et al., 1993], the NNFF International Database collects data on 98 items and allows for serial collection. All patients in this analysis had NF1 as defined by the NIH Consensus Conference [1988]. We tried to determine whether any patients had been previously reported. We identified the syndromic patients reported to have possible Watson or NF1-Noonan syndrome and took special note of them in the analysis. The diagnoses are those provided by the contributors. In some instances, the referring NF center was queried to provide additional information, but the NF clinic records rarely contained more information on the CVM than was reported to the database initially.

To promote insight into possible mechanisms and timing, cardiac abnormalities were categorized as CVMs, other non-CVM abnormalities, and peripheral vascular abnormalities [Lin and Pexeider, 1994]. A CVM was defined as a structural malformation involving intracardiac structures or great arteries. For this analysis, CVMs were organized by mechanistic groups as proposed by Clark [1995]. Because the term “coarctation” can be used broadly, we specified whether there was a descending thoracic juxtaductal aortic shelf seen in typical non-NF coarctation [Beekman, 1995] or a fusiform or segmental hypoplasia of the thoracic or abdominal aorta.

We recorded information about the method of CVM diagnosis, i.e., whether it was based clinically using auscultation, radiography and electrocardiography, or whether it was confirmed using echocardiography, catheterization, surgery or autopsy. If the NF clinician attributed a murmur to a specific CVM, the patient was said to have a clinically diagnosed CVM. We excluded patients with heart murmurs that were called “functional,” however, or if an echocardiogram proved that a suspicious murmur had no anatomic basis. Left superior vena cava to the coronary sinus was also excluded as a normal variant. In addition to CVMs, we tabulated cardiomyopathy, “dysplastic” valve defects (e.g., mitral valve prolapse), intracardiac tumors and electrocardiographic abnormalities.

RESULTS

At the time of this analysis, the NNFFID contained 2550 definite or possible NF1 patients (Table IV). One hundred-twenty patients were submitted for review as having NF1 and a cardiac abnormality; 104 of these were verified as having NF1 according to the NIH diagnostic criteria.

Among the 228 patients who did not meet the NIH criteria for diagnosis of NF1 were 1 patient with LEOPARD syndrome and 15 submitted as having a cardiovascular abnormality. Eleven of these 15 patients were less than 10 years old and had multiple café-au-lait spots but no other features that would establish a diagnosis of NF1 according to the NIH criteria. Five of these 11 young patients had pulmonic stenosis, specified as valvar in one.

Seven of the 104 patients who met the NIH diagnostic criteria for NF1 and were originally submitted as having a cardiovascular anomaly were excluded because they were found not to have a cardiac abnormality or only to have a minor one that could be considered a normal variant (Table IV). The remaining 97 patients with well-established NF1 and a clear diagnosis of cardiovascular disease comprised the cohort that was studied further. Two of these patients had Watson syndrome, and two had the NF1-Noonan phenotype.

Table V lists clinical characteristics of the entire cohort and those with a CVM. Overall, there was a slight male predominance and young age (<10 years). All 4 Watson and NF1-Noonan patients were 6 years old or younger. The diagnosis of most (65%) CVMs was either clinical or not stated. The diagnosis was confirmed by echocardiography alone or with another test in 30%.

Fifty-five (56%) of the 97 verified NF1 patients with cardiovascular disease in this study had a CVM. Patients with CVMs represent 2.3% (54/2322) of the NF1 patients in the NNFFID (Table VI). Excluding 8 patients in whom a CVM was reported as “possible” or inadequately specified, the frequency decreases to 2.0%.

Information is not available on the expected prevalence of CVMs in people without NF1 who represent the same populations and age distributions as our patients. The Baltimore-Washington Infant Study (BWIS) is a recent large population-based study of CVMs in infants (<1 year old) [Ferencz et al., 1993]. Although these data are not strictly comparable to those of the NNFFID, we have included the BWIS population frequencies and distribution of CVMs in Table VI to provide a general frame of reference.

Most (43/54, 80%) of the CVMs in NF1 patients could be classified as “flow defects,” in which right or left heart obstruction or simple shunts are postulated to be the result of abnormal embryonic intracardiac hemodynamics [Clark, 1995]. Most (25/43) of the patients with flow defects had pulmonic stenosis, usually specified (or presumed) as valvar. Three of the 4 children with Watson syndrome or NF1-Noonan syndrome had pulmonic stenosis, and one boy from Children's Hospital in Boston with a pulmonic stenosis murmur has been reported as having a large NF1 deletion [96–665 in Wu et al., 1995, 1997; Riva et al., 2000].

Left heart obstruction (aortic stenosis, coarctation) was reported in 7 patients. Of 5 patients with “coarctation,” however 3 had sufficiently detailed medical records to determine that there was a long fusiform narrowing of the aorta. This differs from the juxtaductal shelf (infolding) usually seen with coarctation in the general population [Beekman, 1995].

In this study, there were also 2 patients with tetralogy of Fallot, a conotruncal CVM (Class I) attributed to altered mesenchymal cell migration. There were no patients with atrioventricular canal defects (Class IV), anomalous pulmonary venous return (Class V), or looping or laterality defects (Class VI). In the Baltimore-Washington Infant Study of newborns [Ferencz et al., 1993], Class II “flow” defects occurred in 70% of patients with CVMs, followed in frequency by Class I “conotruncal” (or outflow) defects (16%), and Class IV “atrioventricular canal” defects (8%). Four hundred-sixty (11%) of the 4308 children with CVMs in the Baltimore-Washington Infant Study had pulmonic stenosis. The proportion of all CVMs that were pulmonic stenosis was 4.3 times (95% confidence interval, 3.2–5.8 times) greater among the NF1 patients than among the patients in the BWIS. The proportion of all CVMs that were coarctation of the aorta was 3.2 times (95% confidence interval, 1.4–7.4 times) greater among the NF1 patients than in the BWIS.

To complete a full description of the NF1 cohort with cardiovascular disease, Table VII lists 27 patients (28% of the 97 reported with cardiovascular abnormalities) with NF1 and other cardiac abnormalities, most of whom had a murmur (unspecified 10, systolic 6). There were no patients in this group from the NNFFID with HCM. Table VIII also lists 16 NF1 patients who had a peripheral vascular abnormality. Four of these patients also had a CVM and are included on Table VI. The observed frequency of peripheral vascular abnormalities probably underestimates the true prevalence because vasculopathy is unlikely to be found in our patients unless they become symptomatic or hypertensive or both.

DISCUSSION

The participants in the NNFFID are a heterogeneous group of NF centers who contribute the data voluntarily. Because most are at tertiary level hospitals, the NF1 patients they see may be more likely to have serious medical problems (such as CVMs) than an unselected group of NF1 patients. On the other hand, very few of the clinics have any particular interest in cardiac disease, as demonstrated in many cases by the poor quality of data recorded about the CVMs. This study is the largest review of cardiac abnormalities in NF1 published to date. The overall frequency of 2% of CVMs among NF1 patients without Watson or NF1-Noonan syndrome in the NNFFID is similar to what was reported by Carey et al. [1979] and Schorry et al. [1989] (Table I). Neither study reported the type of CVM present. Colley et al. [1996] also reported a frequency of 2.0%, but after excluding the patients with Watson (2) or NF1-Noonan (4) syndrome, this figure decreases 0.7%. Perhaps more striking than the overall frequency of CVMs in the Colley et al. [1996] series was that pulmonic stenosis was the sole CVM reported.

The estimate of total CVMs in NF1 obtained from the NNFFID (approximately 20/1000) cannot be directly compared to the population incidence (approximately 5/1000) of children ascertained to age 1 year in the Baltimore-Washington Infant Study [Ferencz et al., 1993]. Data about the frequency of CVMs in older children and young adults in the general population are scant. One small regional study reported a prevalence of 12.5/1000 at age 16 years [Roy et al., 1994]. Our patients range in age from infancy to late adulthood, although most are children (46% younger than 10 years) or young adults who come from specialized NF1 centers throughout the world. Therefore, our patients cannot be assumed to be representative of either all individuals with NF1 or of any particular normal population.

It is likely that some clinically mild CVMs were undetected during the BWIS ascertainment period and recognized after age one year. Nevertheless, the prevalence of CVMs and the relative frequencies of the various defects found in the BWIS are probably the most reliable figures currently available and are similar to those provided in other population-based studies of CVMs in infants and children [Table VIII in Lin et al., 1999]. Although not strictly comparable to the NNFFID data, the BWIS provides a useful general frame of reference for our study.

Cardiovascular malformations in NF1 may have been underestimated in this study if some of the NNFFID patients are not yet old enough to permit a confident diagnosis of NF1 according to the NIH Diagnostic Criteria [DeBella et al., 2000] or if some CVMs were not detected or inadequately documented to permit inclusion in this study. There may have been an overestimate due to referral bias because NF1 patients with CVMs were more likely to be seen by specialists who would recognize NF1. Finally, because a large proportion of the reported CVMs in this series were diagnosed clinically (28%) or not stated (41%), we cannot be certain that a CVM was present. Little information was available about karyotype or NF1 microdeletions. None of our NF1 patients with CVM had DNA studies or microdeletion studies for other loci (e.g., 22q11).

The study's most striking finding is the predominance of pulmonic valve stenosis among NF1 patients. The proportion of all CVMs that were pulmonic stenosis among the NF1 patients was almost 6 times greater than among the patients in the BWIS. This preponderance of pulmonic stenosis, usually valvar, among NF1 patients is consistent with the clinical series and case reports that have previously been published (Table I) and provides strong evidence that NF1 gene mutations predispose to the development of this particular CVM. The possibility of pulmonic stenosis should be considered in any NF1 patient with a systolic murmur. Similarly, the diagnosis of NF1 should be considered in any child with pulmonic stenosis who has multiple café au lait spots.

The apparently increased proportion of NF1 patients with CVMs who have coarctation must be viewed with caution. In 3/5 NF1 patients, the coarctation was a fusiform narrowing of the descending thoracic aorta, that differs from the typical juxtaductal shelf seen in the general population [Beekman, 1995]. This observation and the apparent association between the occurrence of renal artery stenosis and coarctation in NF1 patients suggests that coarctation may often be a manifestation of NF vasculopathy rather than a true CVM. We did not observe peripheral vascular anomalies in any NF1 patients with coarctation of the aorta in this study (Table VIII).

An equally important finding is the low frequency of more complex CVMs such as conotruncal, atrioventricular, laterality and looping defects. The paucity of conotruncal CVMs in patients with NF1 is in striking contrast to the Nf1 “knockout” mouse homozygous mutant embryos [Brannan et al., 1994; Jacks et al., 1994]. These embryos die with double outlet right ventricle, a CVM attributed to abnormal migration of ectomesenchymal cells [Clark, 1995] derived from the cranial neural crest [Kirby et al., 1983]. Furthermore, these mice embryos have enlarged and abnormal endocardial cushions [Lakkis and Epstein, 1998] that may obstruct forward blood flow in some cases. This anomaly suggests that neurofibromin is required for normal cardiac valve development.

Complex developmental mechanisms in addition to neural crest migration play a role in cardiac development in NF1. The rarity of complex CVMs supports the notion that NF1 is not a multiple malformation syndrome and may better be viewed as a multiple dysplasia syndrome [Carey and Viskochil, 1999].

Cardiac hypertrophy has been reported in a small number of patients with NF1 (Table III). In 14 of the 18 reported patients, this was probably primary HCM (usually left ventricular, idiopathic hypertrophic subaortic stenosis), although extrinsic compression by a neurofibroma caused biventricular hypertrophy in at least one girl [Pung and Hirsch, 1995]. In 2 patients, HCM and NF1 were probably segregating as independent autosomal dominant traits [Wille et al., 1980]. One boy may have had the NF1-Noonan syndrome [Gerbaux et al., 1974]. Two patients had either post-mortem examination [Sachs et al., 1984] or biopsy [Schrader et al., 1986] with myocardial examination showing non-specific changes and fiber disproportion with fibrosis, respectively.

No patient in the NNFFID had HCM (Table II). Although this may reflect a genuine lack of association, it may just be due to small sample size and insufficient statistical power to detect an association. The reported prevalence of HCM varies with age. In newborns followed to age one year in the Baltimore-Washington Infant Study, the prevalence was 0.4 per 10,000 [Ferencz et al., 1993]. This is similar to a retrospective study from Finland that reported 0.3 per 10,000 [Arola et al., 1997]. In contrast, an oft quoted frequency of HCM among healthy adults (23–35 years) is 20 per 10,000 (1 per 500) [Maron et al., 1995]. Power calculations indicate that we are unlikely to have detected even a 5-fold increase in the prevalence of HCM in this study.

It remains unclear whether HCM represents a rare manifestation of the NF1 mutation or a chance occurrence. Of patients in the literature (Table II), none has had molecular analysis for either the NF1 or common HCM mutations [Coonar and McKenna, 1997]. Because neurofibromin is expressed in the myocardium of the developing heart and the cardiac muscle becomes thinned in homozygous Nf1 mutant mice [Brannan et al., 1994; Jacks et al., 1994; Lakkis and Epstein, 1998], there is biologic plausibility that myocardial disease may occur in NF1.

There is no firm genotype-phenotype correlation with particular NF1 mutations. NF1 patients with deletions of the entire gene frequently have dysmorphic features and intellectual limitations [Leppig et al., 1994; Wu et al., 1995, 1997; Tonsgard et al., 1997; Riva et al., 2000], but these clinical features are not obligatory [Rasmussen et al., 1998]. A CVM was noted in 3 reported patients with large deletions [Leppig et al, 1994; Wu et al., 1995, 1997; Patient 3 in Tonsgard et al., 1997] (Table I). Because the NIH criteria remain the “gold standard” for diagnosis, molecular analysis for NF1 mutations was not routinely done for patients in the NNFFID.

At present, all individuals with NF1 should have careful cardiac auscultation and blood pressure monitoring as part of every NF-related examination. The value of routine echocardiography for all NF1 patients has not been established. There is little disagreement that a child with NF1 and a murmur, especially if there are associated dysmorphic features, needs a cardiology evaluation. This will likely include an echocardiogram.

To obtain more accurate information about CVMs in NF1, a multi-center study is needed involving geneticists and cardiologists to ensure accurate diagnosis of both NF1 and cardiac abnormalities. Cross-sectional echocardiography could be used to evaluate patients in various age groups. The natural history of pulmonic stenosis in NF1 should be studied to compare its progression (or regression) with the general population. Pulmonic stenosis in NF1 seems to be generally mild and not a cause of serious morbidity, but this clinical impression requires confirmation. To pursue an investigation of HCM in NF1, echocardiography could also be used to monitor for obvious hypertrophy and subtler increased cardiac mass.

CONCLUSIONS

Among patients entered into the NNFFID, there was a higher than expected frequency of pulmonic stenosis and aortic coarctation. Appropriate cardiac examination and blood pressure monitoring should be part of the routine care of individuals with NF1, and those found to have abnormalities should be referred for cardiovascular assessment and treatment.