Estimated Frequency of Genetic and Nongenetic Causes of Congenital Idiopathic Cerebral Palsy in West Sweden


Correspondence with the author at: H. Costeff, P. O. Box 5026, 47140 Ramat Hasharon, Israel. Fax: 9723-549-4543. E-mail:


Mathematical analysis of prenatal and perinatal risk factors was performed on the first 681 published cases of idiopathic congenital cerebral palsy (born 1959–1970) in the west Swedish population-based cerebral palsy (CP) study. Analysis indicates that an estimated 40% of etiologically undiagnosed cases of CP in the community (48% of those born at term and 24% of those born prematurely) are genetically caused. These proportions of genetic causation are no less in CP than in idiopathic mental retardation. Genetic causes account for 60% of maturely born hemiplegics, 45% of maturely born spastic diplegics, 32% of premature spastic diplegics and virtually all cases of pure ataxia. About 23% of CP cases in the community have suffered nongenetic brain damage in accordance with the two-stage model. The residue of 37% is characterized by a single risk factor, usually perinatal.


Cerebral palsy (CP) is a prime example of a recognized medical condition comprising different clinical syndromes and caused by different and often unknown etiologies. What unites the syndromes and the etiologies in a single diagnosis of cerebral palsy is the need for common medical and paramedical modes of rehabilitation. Clinicians accustomed to dealing with syndromes, etiologies and their interactions in other areas of child neurology must necessarily feel the need of firmer epidemiologic data relating the various CP syndromes to specific etiologies.

The medical literature of the past 50 years reveals only one database of cases of cerebral palsy in which epidemiologic data and clinical neurologic diagnosis are detailed in numbers large enough to permit analysis of the relationship between them. This database has been compiled by Hagberg and colleagues in the Goeteborg area of west Sweden, on an ongoing basis since 1959. It has been the subject of eight published reports that analyzed various clinical and epidemiologic aspects. Four of these reports (Hagberg & Olow 1975; Hagberg & Hagberg, 1984; Hagberg & Hagberg, 1996; Hagberg et al. 2001) presented an overall picture of the relationship of various risk factors to the clinical syndrome presented. However, no quantitative hypotheses were derived from this material. Questions such as, “What proportion of cases of congenital hemiplegia in term births are caused by traumatic delivery/genetic pathology?” remain unasked and unanswered, despite the availability of a highly informative method of analysis of medical histories, that proved its value in the study of mental retardation. This method (Costeff et al. 1983) had not yet been published when the first overall survey of this case material appeared in print (Hagberg & Olow, 1975) and it has not since been applied to these data. It seems worthwhile to apply this analysis to the west Swedish database, even at this late date, for two reasons:

  • 1There is reason to expect that its conclusions and hypotheses will be more accurate, more precise and certainly more testable than those that have been proposed on the basis of less rigorous methods.
  • 2This analytic method is worth demonstrating in its own right, as it may be useful in other conditions characterized by a common phenotype and multiple etiologies, such as infantile autism, congenital cleft palate, and attention deficit disorder.

These two considerations may justify the renewed analysis of clinical material collected in 1959-70 and published in 1984, especially in the absence of any other comparably sized database. This article presents the medical history analysis of the first 681 cases of cerebral palsy published in the ongoing study of idiopathic congenital CP in west Sweden. It will describe the method of analysis in greater detail than is usual for a clinical publication, in the hope that readers may find it useful in other appropriate situations.

Patient Data and Analytic Methods

The 681 cases of CP included 457 born at term (>259  days) and 224 prematures. The 457 term cases included 189 hemiplegics, 136 diplegics, 54 dyskinetics, 46 ataxics and 26 tetraplegics. The 224 prematures included 55 hemiplegics, 138 diplegics and 25 dyskinetics. Nine of the prematures and 81 of the term deliveries had obvious etiologic diagnoses. The remaining 376 term births and 215 premature births were without any specific diagnosis other than a summary of the characteristics of the CP syndrome (e.g. spastic diplegia, dyskinesia). Together with a control group of 215 healthy children born at term they constitute the database to be analyzed. The relevant published data from the west Swedish study are presented in Tables 1a and 1b. The risk factors serving as the object of study are listed in Table 1c.

Table 1a.  Risk factors among 376 cases of CP born at term
CP categoryHemiplegiaTetraplegiaSpastic diplegiaAtaxic diplegiaAtaxiaDyskinesiaControl
  1. *Category in table 1a includes only 13-15 cases.

  2. —Category includes only 1-3 cases.

  3. All others comprise at least 20 subjects

Number of subjects16515*88203454215
Risk factors
Table 1b.  Risk factors among 215 subjects with CP, born prematurely
CP categoryHemiplegiaTetraplegiaSpastic diplegiaAtaxic diplegiaAtaxiaDyskinesiaControl
  1. *Category in table 1b includes only 13-15 cases.

  2. —Category includes only 1-3 cases.

  3. All others comprise at least 20 subjects

Number of subjects54111913*32514
Risk factors
Table 1c.  Prenatal and perinatal risk factors routinely sought in the Goeteborg study
Toxemic signsExtra-uterine asphyxia
Two or more previous abortionsHypoxia
Significant maternal illnessHyperbilirubinemia
Small for gestational ageCerebral hemorrhage
Placental infarctionPlacental abruption
Muliple pregnancy 

Analysis was conducted to estimate the proportions of genetic causation among the various_clinical CP syndromes. These proportions are group parameters, and they do not refer to the individual “genetic” subject, whose etiologic diagnosis would be better described as “probably genetic,” in view of the lack of positive evidence of genetic causation.

Analysis Consisted of the Following Operations

1.1 Let (u) = frequency of risk factors in pregnancy, and (v)= frequency of non-neuropathic risk factors in delivery. Then the likelihood that a subject with normal pregnancy and delivery will have no recorded risk factors is (1 −u) * (1 −v). This likelihood will be designated (w).

In a collection of cases of CP, causes other than pathologic prenatal or perinatal events will be given by (g = p/w), where (p) = proportion of CP subjects with no recorded prenatal or perinatal risk factors and (w) = proportion of non-CP subjects with no recorded risk factors.

Three samples of non-CP subjects were available for estimating the value of (w):

  • 1215 term newborns serving as normal controls to a Swedish study of dyskinetic cerebral palsy (Kyllerman & Hagberg, 1983)
  • 214 normal premature newborns, from the same Swedish study.
  • 330 Israeli individuals suffering from severe genetically caused mental retardation. All had IQ < 50, at least one additional retarded sibling and closely consanguineous [F ≥ 1/16] parents (Costeff et al. 1983).

Table 1d summarizes the estimates of (w). In the 30 subjects with genetically caused mental retardation there were 7 non-neuropathic recorded risk factors, yielding an estimate of (w= 0.77), with 95% limits of confidence 0.51 - 1.0 The estimates based on the 215 term newborns (w= 0.75) and on the 14 prematures (w= 0.71) were not significantly different from this or from each other. Given this similarity, the estimate based on the 215 term newborns was preferred for further calculations, because of its smaller variance.

Table 1d.  Estimated mean and variance of (w)
95% Limits
of confidence
Swedish term newborns2150.7500.0008720.69–0.81
Swedish prematures 140.7140.02040.45–1.0
Israeli genetic retardates 300.7730.01730.51–1.0

In the simplest hypothesized model the varieties of cerebral palsy are exclusively caused either by pathologic prenatal events, or by pathologic perinatal events, or by genetic pathology similar to the causes of retardation in the 30 cases. The fraction attributed to genetic causes is (g = p/w), as described above.

A more complex model postulates the existence of covert nongenetic causes of CP which are not recorded as risk factors, and which are nonetheless neuropathic events to the fetus. The fraction (g = p/w) will estimate the sum of these causes together with the genetic causes. The random distribution of these nongenetic causes, if they exist, makes possible their independent estimation:

1.1a. The different CP syndromes have widely different values of (g = q + j) where (q) represents the frequency of covert causes of CP, and (j) represents the frequency of genetic causes. Each value of (g) can be taken as an upper bound of estimate of (q) for all the CP syndromes. The lowest estimated value of (g = q + j), (usually found in dyskinesia), is the upper bound estimate of (q). If it is subtracted from the various estimates of (g), the remainder will be the lower bound estimate of the respective values of (g).

1.1b. Calculation of (g) is based on estimates of (w). If (w) is calculated for a normal control group it will reflect both covert nongenetic causes (if they exist) and genetic causes. If it is calculated on the basis of a group of subjects with chronic neural handicap - not CP - with familial prevalence and/or inbreeding it will reflect only genetic causes. The difference between (g = p/w) and (j=p/w) is an unbiased estimate of (q), which can be subtracted from the various estimates of (g) to give estimated proportions of genetic causation corrected for covert nongenetic causes if they exist.

1.2 The correlation between cerebral palsy, which is an expression of existing brain damage, and single prenatal, perinatal and neonatal risk factors for present or future development of brain damage, is notoriously poor (Leviton, 1987). Attempts to improve the clinical-etiologic correlation have sometimes led to a hypothesized two-stage model of brain damage in which a mildly morbid pregnancy (such as placental insufficiency or any other factor compromising fetal growth) renders the fetal brain vulnerable to a degree of perinatal stress (such as anoxia), which would have been harmless in the absence of the earlier insult. This approach treats all pairs of risk factors - one factor in pregnancy and one in delivery - as possibly significant pathogens. How significant they are can be shown by a comparison of their frequency in cases of CP with their frequency among normal healthy children, such as those in our control group. This latter frequency of paired non-neuropathic risk factors among healthy children (S) is the product of the likelihood (u) of a non-neuropathic risk factor of pregnancy and the likelihood (v) of a non-neuropathic risk factor of the perinatal course (S = u*v). The overall frequency (T) of cases of CP with pairs of risk factors is the sum of (S), the non-neuropathic pairs and (D), the pairs responsible for brain damage.

Thus:(T = S+D), and (D = T−S). The likelihood of a given pair being pathogenic is therefore (D/T).

Parameters based on our control group are:
(u) = 0.22; (v) = 0.05; (u*v) = 0.011.

(T) is the principal variable in estimating the frequency of pathogenic pairs of risk factors in a given group of cases of CP. It is directly observed (e.g. 13% among hemiplegics born at term, and 15% among prematurely born spastic diplegics).

With the aid of the above formulae and parameters, this study estimates the proportions of the various idiopathic CP syndromes caused by genetic defects. It also estimates the role played in cerebral palsy by the two-stage model of brain damage. The raw data serving as substrate of this study are summarized in Tables 1a and 1b.


The results of the analysis are summarized in Tables 2a and 2b. Among the 376 CP subjects born at term, the causative role of genetic pathology was surprisingly large. The estimated overall contribution of genetic causes to cerebral palsy in the community was 40%, if we assume that all intrauterine neuropathic events entail recorded risk factors. Almost half (48%) of the mature cases of CP had a genetic basis in this analysis. Table 2a shows that genetic causes were responsible for 60% of the hemiplegics and 45% of the cases of spastic diplegia born at term, and virtually all cases of pure ataxia. As seen in the table, genetic causes were somewhat less prominent among the 215 prematures, but even among them they accounted for 24% The most striking evidence of hereditary causation of cerebral palsy in preterm babies was among the cases of spastic diplegia, 32% of which were genetically caused on the basis of mathematical analysis. Table 2b summarizes the statistical calculations. These indicate that the estimates of the genetic fractions of the various CP syndromes are robust. If we assume the possible existence of intrauterine covert neuropathic events, analysis estimates that about 1.3% of CP or less is caused by such events. This suggests that such covert events make no significant contribution to the causes of CP.

Table 2a.  Estimated frequency of genetic and nongenetic etiologies of congenital cerebral palsy, by clinical category
 CP categoryHemiplegiaTetra-plegiaSpastic diplegiaAtaxic diplegiaAtaxiaDyskinesiaTotal CP
  1. *Category in table 2a includes only 13-15 cases.

  2. — Category includes only 1-3 cases.

  3. All others comprise at least 20 subjects

  4. “Double” means that at least two identified risk factors, one prenatal and one perinatal, are present.

  5. “Single” means that only one type of risk factor (prenatal or perinatal) was found.

  6. “Single” may indicate that the cause of CP remains unknown, but may also refer to a diagnostically decisive episode of anoxia or trauma in a previously healthy newborn.

Term birthGenetic61%18%45%27%100%3%48%
Preterm birthGenetic17%32%0*11%24%
Table 2b.  Estimated means, variances, standard errors and 95% limits of confidence for (g)
Clinical featuresNPVariance PGVariance g95% limits
All CP cases born at term3760.360.0006130.480.04690.35–0.61
Hemiplegics born at term1650.450.00150.610.00750.44–0.78
Spastic diplegics born at term 880.320.002470.450.01060.24–0.66
Dyskinetics born at term 540.020.0003630.030.00390–0.15
All prematures with CP2150.180.0006850.240.002170.10–0.38
Premature hemiplegics 540.130.002090.170.009380–0.36
Premature diplegics1190.240.001530.320.007590.15–0.49

The results of analysis support the validity of the two-stage model of brain damage. A combination of prenatal and perinatal risk factors was found in 23.5% both of the term deliveries and of the prematures. This combination would be expected by chance in only 1.1% of the subjects (See Appendix). Therefore the finding of any such pair of risk factors in an individual suffering from cerebral palsy in this series has a 95% likelihood of being pathogenic. This type of causation seems to account for about 39% of the nongenetic cases of cerebral palsy.

The remaining 36% of cases of cerebral palsy in the study are associated neither with demonstrable genetic causes nor with specific pairs of risk factors. In analysis, after deducting the genetic cases and those caused by specific pairs of risk factors, 58% of the remaining cases born at term and 91% of the prematures, showed single perinatal risk factors. The very frequent perinatal risk factors among prematurely born cases of cerebral palsy indicate that perinatal pathology is an important cause of CP in prematures, probably accounting for the bulk of the 36% residue not caused by genetic pathology or by specific pairs of risk factors. Among cases of cerebral palsy born at term the role of single-factor perinatal pathology remains important, although clearly less so than in prematures.


The most noteworthy finding of this quantitative analysis is the unexpectedly large role of genetic causes in cerebral palsy. In this population-based study, an estimated 40% of etiologically undiagnosed cases (i.e. 35% of all cases) were due to genetic pathology. Previous studies (Hagberg & Hagberg, 1984; Leviton, 1987) have adduced evidence that much of CP is caused by “prenatal factors.” The present findings indicate that these prenatal factors are mostly genetic and are not a function of conditions such as placental dysfunction, which would be reflected by a surplus of prenatal risk factors when compared to either a group of individuals with genetically caused mental retardation, or to a control group of the normal population. The type of genetic pathology involved is uncertain. Consanguinity analysis could possibly estimate the portion accounted for by autosomal recessive genes, but the necessary information on inbreeding coefficients is not presently available in the published material. Similarly, the lack of sibling information in the database unfortunately renders segregation analysis impossible.

It is worth noting that the frequent genetic pathology is not accompanied by equally frequent dysmorphic signs, hinting that the genetic causes are unlikely to be expressed in abnormal karyotypes. It is also worth noting that the group of genetic cases cannot be expected to respond to attempts to prevent cerebral palsy by preventing premature birth or by improvements in intensive care of prematures. Appropriate intervention in one of the two stages of perinatal hypoxic/ischemic damage may well be more productive.

An additional surprise is the finding that genetic causes are no less important in cerebral palsy than they are in mental retardation. Indeed, in this patient material they are slightly more frequent in CP (40% of all cases) than previously found by Costeff et al. (1983) in mental retardation (38% of cases), although the difference is not statistically significant. This finding challenges the long-established view of cerebral palsy as an example of exogenous brain damage, in contrast to mental retardation which is pictured as a complex biosocial phenomenon. It seems more credible to view both mental retardation and congenital CP as resulting from damage to the developing fetal brain, caused in about 40% of each by pathological genes and in about 60% by a variety of biological prenatal and perinatal disturbances. In cerebral palsy these disturbances are mainly perinatal, as was observed by Little (1843) over one and a half centuries ago.

The validity of the above conclusions depends on the accuracy of the estimate of the proportions of risk factors in the control group and the study groups. An earlier study (Herlitz & Redin, 1955) found in a normal control group the same 0.25 proportion of non-neuropathic risk factors as in the Goeteborg dataset, lending support to the credibility of the present analysis. Similarly, the wide range of estimated genetic causation of the different CP syndromes, from 3% in dyskinesia to 60% in hemiplegia and 100% in pure ataxia, seems incompatible with systematic under-reporting or over-reporting of risk factors. The finding that the estimated value of (w∼ 0.75) was found in a group of individuals with severe genetic retardation (Costeff et al. 1983), as well as in normal control groups suggests that the estimated frequency of genetic causation as defined in this series may indeed be a valid estimate for genetic causation as commonly understood.



Estimation of variance of (g):
The variance of a quotient is given by:




There is no covariance between w and P. This makes possible the simplification of the function for variance of g to: Var g=[Var P+Var w]/w4.

Var P and Var w are calculated from the binomial function (w) is taken as 0.75.