The underlying etiology of infantile spasms (West syndrome): Information from the United Kingdom Infantile Spasms Study (UKISS) on contemporary causes and their classification†
This work was done at The Royal United Hospital Bath NHS Trust and the School for Health, the University of Bath.
Address correspondence to Professor John P. Osborne, Children’s Centre, Royal United Hospital, Combe Park, Bath BA1 3NG, United Kingdom. E-mail: email@example.com
Purpose: To examine the underlying etiology of infantile spasms from the United Kingdom Infantile Spasms Study (UKISS), using the pediatric adaptation of ICD 10.
Methods: Infants were enrolled in a randomized controlled trial or a parallel epidemiologic study. Etiological information included history, examination, and investigations. The infants were classified as proven etiology, if a neurologic disease was identified; as no identified etiology, if no neurologic disease was identified; and as not fully investigated, if a major piece of information was missing. Proven etiology was subclassified using the pediatric adaptation of ICD 10. The results were then examined to identify further methods of classification.
Results: Of 207 infants, 127 (61%) had proven etiology, 68 (33%) had no identified etiology, and 12 (6%) were not fully investigated. Etiologies were prenatal in 63, perinatal in 38, postnatal in 8, and 18 other. The most common etiologies were: hypoxic–ischemic encephalopathy (HIE) 21 (10%), chromosomal 16 (8%), malformations 16 (8%), stroke 16 (8%), tuberous sclerosis complex (TSC) 15 (7%), and periventricular leukomalacia or hemorrhage 11 (5%). The remaining 32 etiologies were all individually uncommon. Response to treatment is given for individual etiologies.
Discussion: Our method of classification allows the reporting of results by individual diseases, disease groups, or categories and is structured and clear. It avoids the use of poorly defined terms such as symptomatic and cryptogenic. It can adapt to new neurologic diseases, such as gene defects, and can be used for comparison of different groups of infants, thereby aiding meta-analysis.
Infantile spasms is an age-dependent epilepsy that most frequently presents in the first year of life with ictal episodes consisting of spasms that usually occur in clusters (Dulac et al., 1994; Schwartzkroin & Rho, 2002). There is a characteristic chaotic and high-voltage interictal electroencephalography (EEG) pattern, which, when typical, is called hypsarrhythmia. At presentation, some infants have already been diagnosed with a neurologic disorder that affects the brain and that carries, independently of the spasms, a risk of developmental delay. Other infants will be investigated as a result of the diagnosis of infantile spasms and will be found to have a neurologic disorder. Both groups of infants have a worse developmental outcome than the large minority of infants who have no such disorder identified despite investigation (Koo et al., 1993).
One neurologic disorder, the tuberous sclerosis complex (TSC), frequently presents because of infantile spasms, perhaps in up to 50% of infants with TSC (Webb et al.,1996). However, TSC is the underlying etiology in only about 5% of infants with infantile spasms and many independently rare disorders together make up the large number of the diagnoses made (Koo et al., 1993). Why some infants with these diagnoses develop infantile spasms, and others with the same disorder do not, has never been fully understood. Labeling the neurologic disorder found as the “cause” of the infantile spasms may, therefore, be at best inaccurate and at worst totally misleading, suggesting that we understand why the disorder has arisen in that particular child when the neurologic disorder has only predisposed the child to infantile spasms. Why do only a few infants with Down syndrome develop infantile spasms? Yet Down syndrome is accepted (Eisermann et al., 2003) as a “cause,” and frequently such infants are not offered a brain scan. As a result, we propose to adopt a terminology that distinguishes the underlying etiology from the cause. Using this scheme, we use the term proven etiology to refer to any identified underlying neurologic disorder, and regard cause as a more specific term that may be a complex and less well-understood sequence or combination of events.
Perhaps because of the many different diagnoses that can be made in these infants and the developmental outcomes associated with them, classification into diagnostic groups has been common. The most frequent nomenclature has been symptomatic, cryptogenic, and idiopathic, but unfortunately there is no clear definition of these terms (Lux & Osborne, 2005).
Symptomatic is often used to indicate that a prior disorder exists (but sometimes that it ought to exist, perhaps because of developmental delay).
Cryptogenic is often used to mean that there must be an etiology, but that one has not been found (but by some that no disorder has been found but developmental delay or some other nonspecific neurologic abnormality is present). The derivation of cryptogenic suggests that it should be used to mean that there is a hidden disorder: the term would, therefore, most logically be used to identify, retrospectively, infants who at presentation had no identified cause but in whom a proven etiology was later found after further investigation; however, it is rarely, if ever, used in this way.
Idiopathic is most commonly used to indicate that no disorder likely to have predisposed to spasms has been found, but other suggestions have included that there is a genetic disorder (Vigevano et al., 1993), but it has not been identified, so this is a guess, or that specific features suggest a good prognosis for developmental outcome (Dulac et al., 1993)—a suggestion that applies only to a small proportion of those with no disorder found after investigation and one that has not stood the test of time.
Developmental delay at presentation with infantile spasms might be due to the epileptic encephalopathy (with which infantile spasms is associated) having been present prior to presentation. This might happen because the clinical manifestations had not been recognized or perhaps because the hypsarrhythmia was present prior to the onset of clinical manifestations. We do not classify infants as having a proven etiology solely because of developmental delay, believing that any underlying neurological etiology explaining the developmental delay independently of the infantile spasms should be properly identified through history, examination, and investigation.
A recent report of the International League Against Epilepsy (ILAE) Commission on Classification and Terminology also suggests that the terms idiopathic, symptomatic, and cryptogenic should be replaced (Berg et al., 2010). It has suggested broad etiologic categories: genetic, structural–metabolic, and unknown.
We present our results on the underlying etiology of infantile spasms from our clinical trial (Lux et al., 2004), the United Kingdom Infantile Spasms Study (UKISS), and from a parallel study of infants whose treatment was not randomized, by using terms identified in a parallel research project (Lux & Osborne, 2005), that led to the West Delphi Consensus Statement, and to see if this improved clarity. West Delphi undertook an e-mail–based Delphi consensus elicitation process on definitions and outcomes in infantile spasms and West syndrome. West Delphi recommended an adaptation of the Paediatric version of International Classification of Diseases 10th edition (ICD 10) (Royal College of Paediatrics and Child Health, 5-11, Theobald’s Road, London, WC1 X 8SH, United Kingdom) (Crawshaw, 1995). This permits precision in classification, and subsequently meta-analyses of smaller better defined groups of infants with specific groups of etiologies and of individual etiologies, where the numbers of such infants allow. This may allow us to identify infants with a better, or worse, prognosis than for the majority—in whom less, or more (respectively), aggressive treatment than usual may prove to be possible. West Delphi also suggested reserving the term symptomatic for cases with an identified underlying disorder, and classifying cases with neurologic symptoms, signs, or developmental delay, but no proven cause or etiology on investigation, as cryptogenic. However, ICD 10 allows for the classification of infants with signs and symptoms not otherwise specified, so those who on examination had neurologic symptoms or signs (for example, microcephaly), but who on investigation had no etiology identified, can be classified as etiology proven (not otherwise specified). We have not, therefore, used the terms symptomatic or cryptogenic.
We report here the current most common etiologies found in UKISS using the West Delphi classification, and we also report the most useful investigations for identifying proven etiology.
One hundred seven infants were enrolled into a randomized controlled trial comparing hormonal treatment (either oral prednisolone or intramuscular tetracosactide depot) to vigabatrin. Another 100 infants, who for one reason or another were not enrolled into the trial, were also followed up using the same report forms. The full details of the trial have been published (Lux et al., 2004, 2005; Darke et al., 2010). Those not in the trial had their treatment determined by the local clinician, who was able to follow the trial protocol, if appropriate, but who was not required to do so. Responders were defined as those with no witnessed spasms on days 13 and 14 from the commencement of treatment.
Patients were followed until age 14 months. Using only those results that were available to us from these reports, etiology was classified using the results of history (including the pregnancy history), the examination, and investigations. We required a brain scan and advised other investigations and required them to be normal (if the results were known) before classification by JPO and AL as no identified etiology where no evidence of a neurologic disorder had been found after investigation. In particular, we asked for the results of skin examination including using ultraviolet (UV) light, metabolic screen (not further specified), and fundoscopy. We expected those with findings suggestive of a specific disorder to be further investigated, such as chromosome analysis where dysmorphic features were present, but such further investigations were at the discretion of the local investigator. Those for which an underlying etiology was found were classified as proven etiology. The classification was not fully investigated where a major piece of information was missing; usually this related to the absence of a cranial scan [either computed tomography (CT) or magnetic resonance imaging (MRI) scan]. Those with proven etiology were subclassified in three parallel ways. First, using the adaptation of the ICD 10 Paediatric international classification of disease (Lux & Osborne, 2005). This is suitable for neurologic diagnoses likely to be found in infants with this disorder that was approved by the West Delphi consensus. Second, we classified the infants into subsets, by apparent time of acquisition of the etiology, that would allow easy comparison of the following main groupings: prenatal, perinatal, postnatal, and other (where the time of acquisition of the etiology is not known or not clear)—see Table 1. Thirdly, we examined the results for other possible useful groupings. The benefit, or lack of benefit, of such groupings will only be determined with time: such groupings will allow readers to see if any publication has an unusual proportion of any group, and it may also allow the detection of groups with a different prognosis. The classification was no identified etiology, where a brain scan was reported as showing cerebral atrophy only during treatment with steroids and this was not confirmed by subsequent poor brain growth or on a subsequent brain scan while off treatment with steroids and no other identified cause was found. The study was approved by the South West Multi-centre Ethics Committee and signed informed consent was obtained from the parents of all infants studied.
Table 1. The classification of underlying etiology using the Paediatric Adaptation of ICD 10
|Prenatal group (congenital)|
N = 63; 38 responders (60%)
|Malformations (if not chromosomal)||Agenesis of the corpus callosum|| || ||1 (0)|| |
|Agyria/polygyria|| || || || |
|Cortical dysplasia (focal)|| || ||1 (1)|| |
|Schizencephaly||1 (1)|| || || |
|Heterotopia (gray matter)||1 (0)|| || || |
|Holoprosencephaly|| || ||1 (0)|| |
|Lissencephaly||1 (0)||1 (1)||2 (0)|| |
|Hydrocephalus|| || ||1 (1)|| |
|Microcephaly|| ||1 (1)|| || |
|Dandy Walker malformation|| ||1 (0)|| || |
|Optic nerve hypoplasia [includes 1 PEHO: Pred (0)]||2 (0)|| ||1 (1)|| |
|Other malformations (specific diseases)||Incontinentia pigmenti||1 (1)|| || || |
|Neurofibromatosis||1 (1)|| || || |
|Tuberous sclerosis complex||3 (2)|| ||11a (8)||Valproate 1 (0)|
|Hypomelanosis of Ito|| || || || |
|Arachnoid cysts|| || ||2 (1)|| |
|Chromosomal (please specify)||Down syndrome||5 (3)||1 (0)||5a (3)|| |
|XXY|| || ||1 (1)|| |
|22q|| || ||1 (1)|| |
|17p 13.3 microdeletion|| || ||1 (0)|| |
|1p36 del||1 (1)|| || || |
|del 1q36 1ptel|| || ||1 (0)|| |
|Cerebral artery disease or stroke||3 (3)||4 (4)||2 (1)|| |
|Porencephaly|| || ||1 (1)|| |
|Muscle eye brain disease|| || ||1 (1)|| |
|Hypoxic–ischemic encephalopathy (HIE)||1 (0)|| ||1 (0)|| |
N = 38; 19 responders (50%)
| ||Maternal factors, e.g., drug abuse||1 (1)|| ||1 (1)|| |
| ||Birth trauma including HIE due to trauma|| || || || |
| ||Intrauterine asphyxia (HIE) pyridoxine given to 1 also on vigabatrin: a nonresponder.||3 (3)||1 (1)||8+ (2)||+Pyridoxine given with vigabatrin 1 (0)|
| ||Infections: meningitis, CMV, Herpes, Toxo plasmosis, etc|| || ||1 (0)|| |
| ||Intracranial nontraumatic hemorrhage|| ||1 (1)|| || |
| ||Transient endocrine or metabolic diseases of the newborn: e.g., hypoglycemia (2)||2 (0)|| || || |
| ||Other – PVL/PVH from preterm injury||2 (2)||3 (2)||6 (4)|| |
| ||HIE uncertain cause||2a (1)||1 (0)||4 (0)|| |
| ||Stroke or infarct||1 (0)|| ||1 (1)|| |
N = 8; 4 responders (50%)
|Brain neoplasm||Malignant|| || || || |
|Benign|| || || || |
|Endocrine or metabolic||Hypoglycemia|| || || || |
|Classical phenylketonuria|| || || || |
|Organic acidurias|| || || || |
|Amino acidurias|| || || || |
|Enzyme deficiencies||Pyridoxine dependency|| || ||1 (0)|| |
|Mitochondrial disorder||1 (1)|| || || |
|Nervous system||Meningitis||2 (1)||1 (1)||1 (0)|| |
|Encephalitis||1 (0)|| ||1 (1)|| |
|Cerebral abscess|| || || || |
|Porencephaly|| || || || |
|Other|| || || || |
|Cerebrovascular disease||Cerebral hemorrhage|| || || || |
|Cerebral infarct or stroke|| || || || |
|External injury||Trauma or nonaccidental|| || || || |
|Other group (timing of disease not known)|
N = 18; 9 responders (50%)
| ||Cortical atrophyb||1b (1)||1 (0)||1 (1)|| |
| ||Cortical scarring|| ||1 (1)|| || |
| ||Hemimegalencephaly|| || ||1 (0)|| |
| ||Unexplained calcification|| ||1 (0)|| || |
| ||Unexplained basal ganglia abnormality|| ||1 (1)|| ||1 (0)|| |
| ||Dysmorphic but no disorder identified|| || ||1 (0)|| |
| ||Microcephaly|| ||1 (1)|| || |
| ||Cerebral palsy (no further abnormality identified)||1 (1)|| ||2 (0)|| |
| ||Stroke or infarct (time period not identified)|| || ||4 (2)|| |
| ||Possible arachnoid cyst|| || ||1 (1)|| |
Of the 207 infants studied, 127 (61%) had proven etiology, 68 (33%) had no identified etiology, and 12 (6%) were not fully investigated: of these 12, the missing investigation was a cranial scan in 4, whereas for the remaining 8, no information on etiology was obtained. Of those with a proven etiology, the classification using the adaptation of ICD 10 is shown in Table 1, where the response to treatment is also given. The breakdown into subgroupings of proven etiology is also shown: there were 63 prenatal, 38 perinatal, 8 postnatal, and 18 other. Through examination of the data, we found that it was also possible to amalgamate some of the ICD 10 classifications into subgroups to give meaningful groups with significant numbers. This allows classification as hypoxic–ischemic encephalopathy (HIE) in 21 (10%), chromosomal in 16 (8%), TSC in 15 (7%), and periventricular leukomalacia or hemorrhage (PVL/PVH) in 11 (5%). If all the malformations (excluding TSC) are counted together, they also make a significant group of 16 (8%) and stroke, including porencephaly, accounts for another 16 (8%). The remaining 32 etiologies were all individually uncommon and did not seem to allow significant or sensible amalgamation.
At presentation with infantile spasms, the clinical history (including past investigations) identified the etiology without the need for further investigation in 45, the clinical examination explained 7 (including funduscopy in 2), a cranial scan explained 55, chromosomal analysis explained 5 (and confirmed Down syndrome in 11), and other investigations explained the etiology in 3 (muscle biopsy in 2 and pyridoxine challenge in 1), whereas in one case the method of diagnosis was not clear. A urine metabolic screen did not explain the etiology in any case.
Moving away from the old classification of symptomatic, cryptogenic, and idiopathic to a better defined and more systematic classification, including specific diagnoses and subgroupings, will allow better understanding and analysis of the results of trials and of cohort studies. It is also consistent with the recommendations of the ILAE commission on classification and terminology. Their suggested etiologic groupings of genetic, structural/metabolic, and unknown are less specific and untried, but are easily formulated from the structure we suggest. At present, infantile spasms rarely have a proven genetic etiology. Our more detailed structure will also facilitate meta-analyses of subgroups of infants and of specific diagnoses (using ICD 10) in the future. We believe that these forms of classification will also make it easier to detect the influence of new investigative tools, for example, the use of new gene mutations that are associated with infantile spasms—such as CDKL 5, ARX, and MAG12. If the results of these new investigations are reported, then it will be possible to see the outcomes for these new groups and to see what effect removing those with these new proven conditions has on those remaining in the group still classified as no identified etiology. For example, suppose a new genetic mutation was identified in 10% of cases of infantile spasms that previously had been classified as no identified etiology. First, we could look at the outcome for this group alone, but we could also see what it told us about the outcome for the 90% remaining in the no identified etiology group. Suppose none of these newly identified infants responded to a particular therapy: That would also tell us that those remaining in the no identified etiology group would now have a better response to that therapy than previously expected for that group as a whole. In UKISS, we collected DNA from 63 infants in our cohort and have reported finding one SCN1A mutation (Wallace et al., 2003), one CDKL 5 mutation (Archer et al., 2006), and no ARX mutations: We are still examining other candidate genes. Had these results been known and had they been found in infants enrolled into the randomized trial, the results would have moved infants from no identified etiology to proven etiology; however, we could still follow the effect this would have on group results. One infant classified as enzyme deficiency due to pyridoxine dependency has now been shown through planned withdrawal of pyridoxine not to be dependent, and would now be classified as no identified etiology.
In addition, subgroups of the proven etiology group can be identified by classification into prenatal, perinatal, postnatal, and other (where other includes those that cannot be placed in the first three categories). It may also prove useful to consider reporting on the number of infants with HIE, TSC, chromosomal disorder, PVL/PVH, malformations, and stroke. This makes it possible to identify studies with a higher than average number of participants in one subgroup such as might occur from countries where perinatal complications are more likely. One subgroup might carry a worse, or better, prognosis, and large numbers in that subgroup would then affect the overall results of the study. These subgroupings can also be used to see if any one subgroup has a consistently better prognosis or response to a particular treatment—as TSC is reported to have to vigabatrin. All this can be done while still showing the individual results using ICD 10, so that meta-analyses can subsequently be done on individual etiologies.
Knowing that a diagnosis may soon be made does not help classification at the time of diagnosis of the spasms. As a result, a classification that requires the results of investigations not yet performed cannot be used to determine treatment. Yet if such a diagnosis would be important because it gave a strong lead to the best treatment, then such an investigation would need to be undertaken urgently, if that were possible. Perhaps the best current example might be the urgency to obtain a cranial scan to exclude TSC so that hormonal treatment was not offered to such infants who may perhaps be better treated with vigabatrin Although we accept that there is insufficient information from randomized trials to confirm this, other information suggests this to be a reasonable assumption (Hancock & Osborne, 1998). Given that the cranial scan was frequently a useful investigation in determining proven etiology [explaining 55 of the 132 infants (42%) in whom etiology remained not proven after history and examination], and that such a diagnosis often also gives the opportunity for prognostic information, we recommend that as soon as the diagnosis of infantile spasms is made, a cranial MRI scan becomes an urgent investigation for those infants with no underlying etiology as yet identified. Occasionally a cranial CT scan may be needed in addition to a cranial MRI scan to rule out TSC, since the diagnostic calcifications of subependymal nodules can be difficult to detect on MRI in young infants (personal observation).
How many further investigations to undertake in infants in whom history, examination, and cranial scan fail to explain the underlying etiology remains problematic. There are numerous single case reports of rare disorders identified as an underlying etiology, but where each investigation is unlikely to frequently find a cause. Many of these investigations are also expensive or invasive. However, treatable conditions such as phenylketonuria (detected through a metabolic screen that will also detect rare organic acidurias), GLUT 1 deficiency [cerebrospinal fluid (CSF) glucose], vitamin B12, and folate deficiencies and pyridoxine dependency should be considered at least in those who fail to respond promptly to treatment, precisely because they are treatable. We believe that pyridoxine dependency needs to be considered only where other seizure types also are present and where there is as yet no other identified etiology. New investigations such as analysis of neurotransmitters could also be considered in those failing to respond to treatment, and their place in the etiology of infantile spasms will then become clear.
One effect of our classification system is that any underlying neurologic disorder identified is reported as a proven etiology without regard for whether the investigator thinks that this disorder is, or is not, the cause. This allows others to make that decision in the light of subsequent information from many sources, and would allow retrospective analyses to be undertaken if new information was published. For example, we have found three infants each with an arachnoid cyst. Although their EEG studies might give some clue as to the likelihood that such a lesion had provoked the spasms, the absence of a focus overlying the cyst does not prove that the cyst is not the cause of infantile spasms. Focal lesions, such as those found in TSC, frequently cause symmetrical spasms and often no significant asymmetry in the EEG. Meta-analyses will show in time whether arachnoid cysts are more frequent in this population than expected by chance, although the site of the lesion might also turn out to be important.
There are some problems still to resolve—such as whether to classify an enzyme deficiency as prenatal (as seems appropriate for pyridoxine dependency where seizures can occur in utero) or as postnatal as the ICD 10 suggests. However, as long as the details are reported, such discrepancies are visible and can be taken into account by those reading the publications. However, we would suggest that, in order to avoid confusion, ICD 10 is followed, as we have done in this article, and is not amended. It is possible to add details, for example, of the type of enzyme deficiency: This can be done within the enzyme deficiency category as we have shown in Table 1.
Our study could be criticized because not all infants had the same detailed investigations (because the decision on such investigations was left to the local clinician and there is no consensus on which investigations to perform), so we have been clear about important missing data. However, it is clear that a good history and examination (including examination of the skin with UV light and funduscopy) will detect many underlying etiologies, and that a cranial scan is essential unless the etiology is already known from previous investigations (when it may still be advisable).
In conclusion, we recommend that future studies report on etiology using the ICD 10 pediatric classification, and that the results are also grouped into no identified etiology, proven etiology, and not fully investigated where an important piece of information is missing. In addition, we also recommend reporting the subgroupings of HIE, TSC, chromosomal disorder, PVL/PVH, malformations, and stroke.
We thank the parents who gave their time and their permission for their infants to take part in UKISS at a difficult and traumatic time. We also thank all the clinicians who enrolled patients into this study (their details can be found at http://image.thelancet.com/extras/03art11384webappendix.pdf). This study, UKISS, and AL were supported by a grant from the Bath Unit for Research in Paediatrics (BURP) that included support from Cow and Gate. JPO was also supported by a separate grant from BURP. FJKO’C was supported by the Wellcome Trust and the Castang Foundation. We thank Patricia Shepherd for administrative support.
We confirm that we have read the journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. The authors declare no conflicts of interest.