Factors Associated With Establishing a Causal Diagnosis for Children With Cardiomyopathy

Study Design

The design of the PCMR has been described in detail elsewhere.¹⁶ This report is based on the retrospective data set of the PCMR, which includes detailed diagnostic and management data on 916 children with cardiomyopathy who presented to a pediatric cardiologist between January 1, 1990, and December 31, 1995, at 1 of 38 sites in the United States (814 patients) or Canada (102 patients). Two patients were excluded from analyses, 1 with missing cause status and 1 with a known cause of unspecified type. The window for baseline registry data was the 1-month period after the initial diagnosis of cardiomyopathy.

The protocol was reviewed and approved by the institutional review boards or ethics committees at all participating PCMR sites. Written informed consent from individual patients or surrogates was not required, because there was no direct patient contact and no procurement of medical materials other than written records.

Eligibility Criteria

To be eligible for the PCMR, a patient was required to be <18 years of age at diagnosis and to meet specific echocardiographic, pathologic, or clinical criteria for cardiomyopathy.¹⁶ Of the 14 clinical exclusion criteria,¹⁶ the most common were congenital heart defect not associated with a malformation syndrome, endocrine disease known to cause myocardial damage, chronic arrhythmia, pulmonary parenchymal or vascular disease, immunologic disease, and drug use known to cause hypertrophy.

Definitions of Cardiomyopathy

Cardiomyopathy was classified as pure if it consisted of 1 functional type only (dilated cardiomyopathy [DCM], hypertrophic cardiomyopathy [HCM], or restrictive cardiomyopathy [RCM]) or mixed if it had features of >1 functional type. Although the World Health Organization classifies acute viral myocarditis as an inflammatory cardiomyopathy distinct from DCM,¹⁷ both groups were classified here as DCM because of their similar echocardiographic appearances. Similarly, in the adult cardiomyopathy literature, HCM denotes a sarcomeric defect as the specific cause.¹⁸ Here, we combined all cases that shared a hypertrophic functional type, as indicated by echocardiography, irrespective of cause.

Cardiomyopathy was defined as familial when it occurred in ≥2 family members, irrespective of functional type, cause, or inheritance pattern. Familial isolated cardiomyopathy was defined as cardiomyopathy with no systemic features occurring in ≥2 family members, a single proband with an identified genetic defect, or a metabolic disorder known to cause isolated cardiomyopathy. The cause of familial isolated cardiomyopathy was considered to be known when a specific inheritance pattern could be inferred or when the genetic defect was identified and to be unknown when the relationship to affected relatives (eg, non-first-degree relatives) precluded assignment of a conventional inheritance pattern (autosomal recessive, autosomal dominant, X-linked, or mitochondrial). When >1 inheritance pattern was theoretically possible, the more-conventional pattern was assumed. For example, if a brother and sister were affected and both parents had negative echocardiograms, then the inheritance pattern was described as autosomal recessive, although autosomal dominant with incomplete penetrance would be possible.

Specific laboratory testing included viral serologic testing or culture, chromosomal analysis, metabolic testing, and biopsy. Testing was performed at the discretion of the participating center. Metabolic blood testing was defined as having measurements for ≥1 of the following: ammonia, creatine phosphokinase, carnitine, free fatty acids, lactate, pyruvate, quantitative ketones, acylcarnitines, or amino acids. Metabolic urine testing was defined as having measurements for ≥1 of the following: acylglycines, quantitative amino acids, mucopolysaccharides or oligosaccharides, organic acids, or ketones. Biopsy referred to endomyocardial or skeletal muscle biopsy procedures.

Statistical Methods

Fisher's exact test was used to compare the rates of known versus unknown causes in patient subgroups defined by categorical variables, unless otherwise noted. The distributions of continuous variables according to causal status were compared with Student's t test or Wilcoxon's rank-sum test, as appropriate. Multivariate logistic regression was used to determine which diagnostic tests were independent predictors of a known cause, adjusted for 3 potential confounders (age, presence or absence of congestive heart failure [CHF] at cardiomyopathy diagnosis, and geographic region). The significance level was set at .05, and all tests were 2-tailed. All analyses were conducted with SAS version 8.2 (SAS Institute, Cary, NC) and S-Plus version 6.1 (Insightful Corp, Seattle, WA) software.

Blood and urine tests and biopsies were analyzed statistically in 2 stages, that is, initially for all patients with a definitive testing status and then after certain subgroups of patients, whose diagnoses generally were established through other means, were excluded. For the latter analyses, patients with neuromuscular disease (n = 68) were excluded, because typically musculoskeletal disease is identified before cardiomyopathy develops and test results obtained before cardiomyopathy was diagnosed were not collected in the PCMR. Patients with familial isolated cardiomyopathy (n = 74) or a malformation syndrome (n = 27) also were excluded, because typically these diagnoses are made through echocardiographic evaluation of other family members and physical examination, respectively.

Sample Characteristics

The patient sample was 58.2% male, 66.9% white, 17.0% black, 11.3% Hispanic, and 4.9% other race or ethnicity. The median age at diagnosis was 2.3 years (interquartile range, 0.4–11.3 years). Family histories of cardiomyopathy, sudden death, or genetic syndrome were present in 23.7%, 11.6%, and 6.1% of cases, respectively. Patients were classified as having been diagnosed in Canada (11.1%) or in 1 of 7 geographic regions of the United States (88.9%), with the largest contributors in the Southwest (29.0% of US patients; Arizona, New Mexico, Nevada, Oklahoma, and Texas), Northeast (20.2%; Connecticut, Massachusetts, Maine, New Hampshire, Vermont, and Rhode Island), and Atlantic (17.1%; District of Columbia, Maryland, New Jersey, New York, and Pennsylvania) regions. More than one half (53.8%) of the children were diagnosed as having DCM, one third (34.2%) HCM, 3.2% RCM, and 8.9% other or mixed type of cardiomyopathy.

Only one third (33.3%, or 305) of the 916 patients with cardiomyopathy had a known cause at diagnosis (Table 1). The known causes for the 112 patients with HCM were nearly evenly distributed among familial isolated cardiomyopathy, inborn errors of metabolism (most commonly Pompe disease), neuromuscular disorders (most commonly Friedreich's ataxia), and malformation syndromes (most commonly Noonan syndrome). In contrast, myocarditis (51.6%) was among the most common known causes for the 161 patients with pure DCM, followed by neuromuscular disorders (25.5%; most commonly Duchenne muscular dystrophy) and familial isolated cardiomyopathy (15.5%). All 6 of the 29 RCM cases with known cause were familial isolated RCM. Approximately one half (46.2%) of the 26 known causes in the mixed or other cardiomyopathy group (81 patients) were familial isolated cardiomyopathy, and one fourth each were myocarditis (23.1%) and inborn errors of metabolism (26.9%).

Associations Between Causal Status and Patient Characteristics

For patients with HCM, gender was associated with having a known cause, but age at diagnosis of cardiomyopathy was not (Table 2). Girls were more likely to have a known cause than were boys (48.2% vs 29.3%; P = .001). Among those with known causes, a disproportionate number of the girls (36.5%) had familial isolated HCM, compared with the boys (20.0%). Patients with HCM who had a known cause were also smaller than those without a known cause, as evidenced by their lower mean height-for-age z score (−0.98 vs −0.34; P = .02) and lower mean weight-for-age z score (−0.78 vs 0.03; P < .001). The HCM subgroup with growth z scores below the sample average consisted mainly of children with mitochondrial and other metabolic disorders, Noonan syndrome, or Friedreich's ataxia.

For patients with DCM, there were no gender differences according to causal status, but patients with a known cause were significantly older at the time of cardiomyopathy diagnosis than were those with an unknown cause (median age: 5.3 years vs 1.3 years; P < .001), largely as a result of patients with Duchenne and Becker muscular dystrophy (median age at diagnoses: 14.7 years and 14.2 years, respectively).

For patients with HCM and DCM, the rates of a known cause were significantly higher for patients with family histories of sudden death and genetic syndromes than for those without. For all 4 functional type groups, a family history of cardiomyopathy yielded a significantly higher rate of known causes, compared with that without such a family history, but causal status was not associated with year of diagnosis, medical insurance or type of insurance, or family history of congenital heart disease or arrhythmia.

Associations Between Causal Status and Disease Characteristics

There was no significant difference in the rates of known cause according to functional type (HCM, 35.8%; DCM, 32.7%; other or mixed cardiomyopathy, 32.1%; RCM, 20.7%). Approximately one half (51.0%) of all patients had CHF at the time of diagnosis. Only for the HCM group was the rate of known cause associated significantly with the presence of CHF at diagnosis (51.9% for patients with CHF vs 32.6% for patients without CHF; P = .01). The most common cause among patients with HCM with CHF at the diagnosis of cardiomyopathy was inborn errors of metabolism (59.3%; 16 of 27 patients).

Nearly all children underwent echocardiography (95.6%) and electrocardiography (80.3%), and smaller numbers underwent cardiac catheterization (32.2%) and Holter monitoring (21.9%) when presenting with cardiomyopathy. Although none of these cardiac assessments except Holter monitoring was associated significantly with a cause of cardiomyopathy, certain measurements were (Tables 3 and 4). Patients with HCM and a known cause had a higher mean end diastolic posterior wall thickness z score (3.26 vs 2.04; P = .008) than did those with an unknown cause. This group included predominantly patients with a variety of mitochondrial and other metabolic disorders (especially Pompe disease) and Friedreich's ataxia. A known cause was almost twice as likely in the absence of left ventricular outflow tract obstruction as in its presence (39.2% vs 22.8%; P = .012). No other echocardiographic or electrocardiographic variables were associated with a known cause for patients with HCM.

For patients with DCM, those with a known cause tended to have less-severe cardiac disease than did patients with an unknown cause, as evidenced by several variables. The patients with DCM with a known cause had lower mean end diastolic dimension z scores (3.49 vs 4.56; P < .001), lower mean end systolic dimension z scores (5.32 vs 6.40; P < .001), and higher mean fractional shortening z scores (−7.99 vs −8.99; P = .009) (Table 4). The most common causes for patients with cardiac dimensions below the sample average were myocarditis and Duchenne muscular dystrophy.

A known cause of DCM was nearly 3 times more likely in the absence of mitral valve regurgitation than in the presence of severe regurgitation (41.6% vs 14.3%; P = .010). Similar patterns were seen for tricuspid regurgitation (P = .06) and right atrial dilation (P = .02). Patients with DCM attributable to a specific cause also had a lower mean heart rate (124 beats per minute vs 136 beats per minute; P < .001), which was the only significant electrocardiographic finding for any of the functional types of cardiomyopathy.

The only significant echocardiographic finding for patients with RCM was a higher mean shortening fraction z score for those with a known cause, compared with those without a known cause (1.57 vs −1.26; P = .04). There were no significant echocardiographic findings for the group with other or mixed cardiomyopathy, perhaps because of the heterogeneity of this group.

Associations Between Causal Status and Testing for Cause of Cardiomyopathy

Within each functional type of cardiomyopathy, the overall rates of known causes were similar for patients who had undergone any causal testing and those who had not. Children with HCM underwent testing at the time of cardiomyopathy diagnosis approximately one half as often as did those with other types of cardiomyopathy. The lack of an association between a known cause and diagnostic laboratory testing might have been the result of including children whose disease cause is determined typically through other means. Therefore, analyses were conducted after the exclusion of children with a neuromuscular disorder, familial isolated cardiomyopathy, or malformation syndrome (Table 5). Patients with neuromuscular disease (n = 68) were excluded because typically musculoskeletal disease is identified before cardiomyopathy develops, and test results obtained before the diagnosis of cardiomyopathy were not collected in the PCMR. Patients with familial isolated cardiomyopathy (n = 74) or a malformation syndrome (n = 27) were also excluded, because typically these diagnoses are made through echocardiographic evaluation of other family members and physical examination, respectively.

In this subgroup of patients with HCM, each class of testing (metabolic blood or urine testing, biopsy, viral serologic testing or culture, or chromosomal analysis) was associated significantly with a higher rate of known causes. The most important metabolic urine tests were measurements of ketones and amino acids, with the odds of a known cause being 4 times the odds of a known cause in the absence of such testing. Several metabolic blood tests, including carnitine, creatine phosphokinase, lactate, ammonia, and pyruvate measurements, were associated significantly with a known cause (odds ratios [OR]: 4–7). Skeletal muscle biopsy was more informative than endomyocardial biopsy, but neither was performed very often for children with HCM. Interestingly, 47.8% of children with viral serologic testing or culture had a cause established, which was significantly higher than the 8.6% known cause rate for those without viral serologic testing or culture, despite the fact that none of the patients with viral myocarditis had HCM. All of these patients with HCM were diagnosed as having various inborn errors of metabolism, which indicates that viral testing might be a proxy for causal testing in general or that their cardiomyopathy might have presented in the setting of metabolic decompensation triggered by a viral infection. Finally, chromosomal analysis was associated with known cause (OR: 6.56; P = .01).

For the DCM subgroup, metabolic urine testing, biopsy, and viral serologic or culture testing were all associated with an increased rate of known causes, but metabolic blood testing and chromosomal analysis were not. The urine results were driven by testing for ketones, which was associated with an approximately twofold increase in the rate of known causes. In contrast, plasma pyruvate testing was associated with a threefold lower rate of known cause. Endomyocardial biopsy was a significant predictor (odds ratio: 5.69; P < .001), whereas skeletal muscle biopsy was not (P = .78). The most common cause identified through endomyocardial biopsy was viral myocarditis. Viral serologic or culture testing more than doubled the rate of known causes (28.6%, compared with 12.7% known cause for those without this test).

For the RCM subgroup, there were only 22 patients with testing data after exclusions and none had a known cause, despite the fact that 16 of the patients underwent some form of testing. For the other or mixed cardiomyopathy group, no test was associated significantly with an increased rate of known cause. The test that came closest to achieving significance was skeletal muscle biopsy (66.7% vs 17.2%, with and without biopsy; P = .09). Metabolic blood testing in this small sample also suggested utility (25.7% known cause, compared with 10.0% known cause for those without such testing; P = .12).

Multivariate modeling was conducted to identify which classes of blood or urine tests or biopsies were independent predictors of a known cause for patients with HCM or DCM, after adjustment for each other as well as 3 covariates, namely, age at cardiomyopathy diagnosis, presence versus absence of CHF at cardiomyopathy diagnosis, and geographic region (Northeast versus all other regions). These covariates were chosen because, for some or all types of cardiomyopathy, they were found to be associated with both causal status and the likelihood that testing was conducted. For patients with HCM, the general finding was that, as a group, blood and urine tests were associated independently with establishing a cause (OR: 4.15; 95% confidence interval [CI]: 1.43–12.05; P = .009) but biopsy was not (OR: 1.75; P = .37). Interestingly, viral serologic testing or culture and metabolic blood and urine testing (OR: 3.54; P = .01) were each significant when examined separately in covariate-adjusted models for cause among patients with HCM, but the strongest covariate-adjusted model contained only viral serologic testing or culture (OR: 6.42; 95% CI: 2.14–19.28; P < .001). This paradox suggests that, in this sample, these 2 classes of tests were most likely conducted simultaneously. When skeletal muscle biopsy was added to this model, it was not associated significantly with causal status (OR: 2.97; P = .20). In the DCM cohort, both viral serologic testing or culture (OR: 1.81; 95% CI: 1.81–3.19; P = .04) and endomyocardial biopsy (OR: 4.84; 95% CI: 2.71–8.66; P < .001) were associated independently with establishing a cause.