Estimation of prevalence and incidence of infantile spasms in Taiwan using capture–recapture method

Download (0)



Estimation of prevalence and incidence of infantile spasms in

Taiwan using capture–recapture method

Chih-Chuan Chen


, Ta-Fu Chen


, Haung-Chi Lin


, Pei-Ching Oon



Hui-Min Wu


, Pen-Jung Wang


, Tony Hsiu-Hsi Chen


, Horng-Huei Liou


aDepartment of Neurology and Pharmacology, National Taiwan University Hospital and National Taiwan University College of Medicine,

No. 1, Sec 1, Jan-Ai Road, Taipei 100, Taiwan, ROC

bDepartment of Pediatrics, National Taiwan University Hospital and National Taiwan University College of Medicine, Taipei, Taiwan, ROC cDepartment of Pediatrics, En Chu Kong Hospital, No. 399 Fuhfing Road, Sanhsia, Taipei 237, Taiwan, ROC

dDepartment of Internal Medicine, En Chu Kong Hospital, Taipei, Taiwan, ROC eGraduate Institute of Preventative Medicine, College of Public Health, National Taiwan University,

No. 19, Hsu-Chow Road, Taipei 100, Taiwan, ROC

fDepartment of Pediatrics, College of Medicine, Tzu Chi University, No. 701, Section 3, Chung-Yan Road, Hualien 970, Taiwan, ROC Received 5 August 2003; received in revised form 2 December 2003; accepted 15 December 2003


Purpose: To estimate the prevalence, incidence, and case-fatality of infantile spasms (IS) in Taiwan.

Methods: A retrospective cohort of patients with IS was obtained from one medical center to identify 69 IS cases from 1985 to

1997. This cohort, in conjunction with the claimed data from the National Health Insurance, was used to estimate the prevalence and incidence of IS by capture–recapture design, taking the case-fatality of IS into account.

Results: The prevalence rate of IS for aged 0–9 years was 0.046 per thousand. The incidence rate was estimated as 6 over

100,000 per year in Taiwan. Of the 69 IS cases, 8 deaths were ascertained. The case-fatality rate was 11.6%. The leading cause of death was status epilepticus.

Conclusions: We have demonstrated an efficient method to estimate the incidence and prevalence rates of IS in Taiwan. Our

results help to make a clear understanding of the disease burden of IS in this society. © 2004 Elsevier B.V. All rights reserved.

Keywords: Infantile spasms; Prevalence; Incidence; Capture–recapture; Epidemiology

Corresponding author. Tel.:+886-2-23123456x8325; fax:+886-2-23915297.

E-mail addresses: (T.H.-H. Chen), (H.-H. Liou).

1Co-corresponding author. Tel.:+886-2-23587620; fax:+886-2-23587707.

1. Introduction

Although infantile spasms (IS) accounts for only about 2% of all cases of childhood epilepsy (Hauser, 1994), it is one important type of epilepsy with unique clinical presentations. Previous studies showed factors affecting the risk of IS were highly related to the un-derlying neurological abnormalities or pre- and peri-natal insults (Rantala et al., 1996; Liou et al., 2001). 0920-1211/$ – see front matter © 2004 Elsevier B.V. All rights reserved.


The prevalence rate of IS has been reported to range from 0.14 to 0.2 per thousand children aged from birth through 9 years (Cowan and Hudson, 1991; Trevathan et al., 1999). Males had 1.5-fold risk for IS as com-pared with females (Cowan et al., 1989). Annual in-cidence rate of IS was reported to range from 16 to 42 over 100,000 live births per year (Riikonen and Donner, 1979; Cowan and Hudson, 1991; Ludvigsson et al., 1994; Trevathan et al., 1999).

Most studies on IS have been conducted in western countries but few in oriental countries. In addition, few descriptive studies were designed to simultane-ously estimate the incidence and the prevalence of IS, taking mortality into account. However, conducting a population-based survey to estimate descriptive epi-demiological profiles for a rare disease is costly and time-consuming. To tackle this problem, we proposed an efficient capture–recapture method (Bobo et al., 1994; Murphy et al., 1995; O’Callaghan et al., 1998; Debrock et al., 2000) in conjunction with epidemio-logical study design to estimate the prevalence and in-cidence rate of IS, taking cases of death into account. Two data resources were available for such a purpose. One was based on primary data using a retrospective cohort from a medical center, National Taiwan University Hospital (NTUH). To take cases of death into count, this cohort can be linked to the mortality registry to estimate the survival of patients with IS after its onset. The other relies on the clinical data from the National Health Insurance (NHI) sys-tem that has been implemented in Taiwan since 1995. Although NHI system covers 95% population and can be used to estimate the prevalence of IS, whether case-ascertainment, particular in early stage, was complete is worthy of being investigated. The aim of our study was therefore to use a retrospective cohort of IS from one medical center in Taiwan to esti-mate the incidence, prevalence and case-fatality of IS through capture–recapture design whereby informa-tion on this cohort was compared with the data from NHI that may have incomplete ascertainment of IS.

2. Methods

2.1. Study design

NTUH is a 2000-bed teaching hospital and one of the major tertiary referral centers in Taiwan. There

Fig. 1. Study design for estimating prevalence, incidence and case-fatality rates.

are nine medical centers on the Taiwan Island. As all could diagnose and treat of IS, not all IS were remit-ted to NTUH. We therefore used a capture–recapture method in conjunction with a retrospective cohort. Fig. 1shows the study design for estimating the preva-lence rate, incidence rate and case-fatality of IS. We identified IS from NTUH during the period from 1985 to 1997. Primary data from NTUH were compared with NHI claimed data, making allowance for cases of death by linking NTUH cohort data with Taiwan Mortality Registry. Details of the study design are de-picted in the following:

(1) Prevalence rate: There were two datasets used for estimating the underlying prevalence rate of IS. The major dataset was derived from a retro-spective cohort consisting of a total of 69 IS pa-tients that had been diagnosed as IS by pediatric neurologists from 1985 to 1997 in NTUH, the teaching hospital that captured a substantial pro-portion of IS cases in Taiwan. In order to estimate nation-wide prevalence rate of IS in Taiwan, a nation-wide data from the NHI was also accrued,


including 35 IS and 3,064,804 children aged 0–9 years derived from claimed data on NHI in 1997. However, as data were available from only 1 year and some IS cases may be under-ascertained in early period of NHI, a capture–recapture method (see the description later) was used to estimate the nation-wide prevalence rate by combining the two datasets.

(2) Case-fatality rate: Linking the above retrospec-tive cohort with the national Taiwan mortality registry gives the details of information on death among the 69 IS cases. This provided the base for calculating the case-fatality rate of IS. (3) Incidence rate per 100,000 live birth: A

retrospec-tive cohort was also used to estimate the annual incidence rate for the underlying population who will seek for maternity care in NTUH by calcu-lating numbers of IS cases delivered at NTUH divided by a total of live births at NTUH during the same period.

2.2. Diagnosis of IS

The diagnosis of IS was made by the presence of a clinical history of spasms, and a characteristic elec-troencephalogram (EEG) with or without psychomo-tor retardation after the onset of spasms (Lacy and Penry, 1976; Hrachovy and Frost, 1989). The spasms were classified as flexor, extensor or mixed pattern according to Kellaway et al. (1979). The EEG was classified as either hypsarrhythmic or modified hyp-sarrhythmic by the criteria ofGibbs and Gibbs (1952) andHrachovy et al. (1984), respectively. In addition, subjects having spasms and psychomotor retardation with multifocal EEG activity were included. Accord-ing to the 1989 Classification of Epilepsy and Epilep-tic Syndromes of the International League Against Epilepsy (Commission, 1989), patients were decided into symptomatic and cryptogenic groups. For etio-logical investigation, every patient received thorough clinical history evaluation, developmental and neuro-logical assessment, neuroimaging such as computed tomography or magnetic resonance imaging of the head, chromosomal studies and metabolic studies, such as serum amino acids and urine organic acids. Cryptogenic group was characterized by (1) normal pregnancy and birth, (2) normal development before the onset of spasms and absence of neurological

ab-normalities at the onset of spasms, (3) absence of any other type of seizure before the onset of spasm, and (4) normal laboratory and neuroimaging findings at onset (Watanabe, 1998). When an etiology or signs of previous brain damage could be found or the above criteria were not fulfilled, the patient was classified as symptomatic group.

2.3. Statistical methods

The principle of capture–recapture is that the size of a population can be estimated by trapping and marking a sample and then releasing the marked sam-ple and allowing it to mix with the other members of the population before trapping a second sample. The ratio of marked samples to all individuals in the second catch is the same as the ratio of the number in the first catch to the number in the whole popula-tion. To estimate the nation-wide prevalence, a simple capture–recapture design was used to combine the retrospective cohort with the claimed data on NHI. The procedure is described as follows: Suppose 69 subjects who were diagnosed as IS during the period from 1985 to 1997 were captured by NTUH, labeled with a mark “NTUH-IS” and released into the pop-ulation pool. The NHI nationwide data recaptured these cases. Since not all 69 subjects were recap-tured by NHI, the ratio of recaprecap-tured cases with the mark “NTUH-IS” to all cases identified from NHI in conjunction with 69 subjects was used to indirectly estimate nationwide IS cases. The detailed calcula-tions are shown in Table 2. Let N11, N10, N01, N00 be denoted as IS cases present in both data resources (N11), NTUH but not NHI (N10), NHI but not NTUH (N01) and absent from both data sets (N00). The pur-pose of capture–recapture is to first estimate N which can be estimated by N.1× N1./N11. The number of under-ascertainment, N00, was therefore estimated by subtracting N11, N10, and N01from N.

3. Results

Table 1 shows the distribution of age, sex and eti-ologies in IS cases. Male and female were equally af-fected. The peak age of IS onset was between 4 and 9 months after birth. Among the 69 cases, 55 cases (79.7%) were symptomatic and 14 cases (20.3%) were


Table 1

Distribution of age, gender and etiology of 69 infantile spasms (IS) cases captured by NTUH

Case number (%) Male Female Age (years) 0–3 5 4 4–6 10 15 7–9 11 10 ≥10 8 6 Total 34 (49%) 35 (51%) Symptomatic 55(79.7%) Tuberous sclerosis 9 Asphyxia 8 CNS malformation 7 Prematurity 6 Neurodegenerative 5 CNS infection 4

Traumatic brain injury 4 Chromosome abnormality 3 Early infantile epileptic


3 Pyridoxine dependent seizure 1 Hypoxic-ischemic

encephalopathy, postnatal


Hypomelanosis 1

Unidentified with cerebral palsy 3

Cryptogenic 14 (20.3%)

cryptogenic. The associated diseases of the 55 symp-tomatic cases were also listed in theTable 1.

Linking the NTUH cohort with the NHI data yields the result of the estimated prevalence rate of IS by the capture–recapture method (Table 2). Seventeen cases (N11) were both captured by NHI data and NTUH cohort. Eighteen cases (N01) were found in NHI but not in NTUH. Fifty-two cases (N10) were identified from NTUH but not captured by NHI data. Cases that were

Table 2

Estimates of prevalence of infantile spasms (IS) in Taiwan with capture–recapture method Cases present in

NHI database

Cases absent in NHI database


Cases present in NTUH-IS cohort 17 (N11) 52 (N10) 69 (N1.)

Cases absent in NTUH-IS cohort 18 (N01) 55 (estimated N00) 73 (estimated N0)

Total 35 (N.1) 107 (estimated N.0) 142 (estimated N)

(1) Assumption: consultation habits of all IS patients do not change from year to year. Therefore,N = 35 × 69/17 = 142, and N00= 55. (2) A total of 3,064,804 children aged between 0 and 9 years old were presented in NHI data in 1997. Hence, prevalence of IS = 142/3,064,804= 0.046/1000.

not captured in both datasets (N00) were estimated as 55.

The application of capture–recapture method yields 142 (35× 69/17) estimated prevalent cases of IS (N) from 1985 to 1997 in Taiwan. The number of the de-nominator was 3,064,804. Hence, the prevalence rate of IS aged 0–9 years could be estimated as 0.046 per thousand. It should be noted that of 142 IS cases es-timated from capture–recapture method, both found in two data source accounted for 11.97% (17). Cases found in NHI but not in NTUH explained 12.68% (18). Cases found in NTUH but not in NHI were 36.61% (52). Those cases not found in both data sources were 38.73% (55).

There were only 2 out of 69 IS cases were delivered in NTUH, and the rest of the cases were referred from other hospitals. There were a total of 33,263 live births in NTUH in the study period. The incidence rate of IS was estimated as 6 over 100,000 live births per year.

Of the 69 IS cases, there were eight deaths (five males and three females) ascertained after the linkage with the Taiwan Mortality Registry. The case-fatality rate was 11.6%. The estimated death rate per year using the exponential distribution was 0.01 per case. Cause of death in these patients was status epilepticus (4), pneumonia (1), accident (1) and unknown (2).

4. Discussion

4.1. Credibility of results

Our study used capture–recapture method in combi-nation with a retrospective cohort design and NHI data to estimate disease prevalence and incidence in Tai-wan, taking case-fatality rate of IS into account. The results showed that the estimated prevalence rate for


children aged less than 10 was 0.046 per thousand and annual incidence rate was 6 over 100,000 live birth per year. The case-fatality rate was estimated as 11.6%.

To examine the validity of our results, we applied annual incidence rate per live birth and case fatal-ity rate to estimate nation-wide prevalence rate. We compared this estimate with that calculated from capture–recapture method. As we had around 280,000 live births per year for the birth cohort from 1985 to 1997 among children aged 0–9 years, this yields 168 IS on the basis of 6 over 100,000 liver births per year. As the annual death rate was 0.01 per case, this gives 17 deaths among the 168 IS cases. After subtracting these cases, the estimated prevalent cases for 0–9 years old were 151 that are close to our estimate from capture–recapture method. This suggests that our es-timated prevalence, incidence and case fatality may be representative of current epidemiological profile of IS in Taiwan.

It could be argued that the characteristics of age groups from NTUH may be different from those used in this validation. Subjects from NTUH may be more likely to yield more IS cases than the latter. The slightly higher estimated cases in validated sample in comparison with the estimate from capture–recapture method can support this argument. However, as the difference is trivial, this problem is not serious. 4.2. Methodological consideration

From methodological viewpoint, our study design was efficient in estimating the magnitude of morbid-ity and mortalmorbid-ity of IS. A retrospective cohort plus the NHI data using the capture–recapture method pro-vides an efficient way to estimate the prevalence and incidence of IS. This method dispenses with using a large nationwide survey for estimating these basic epi-demiological figures.

Another strength of using capture–recapture method in our study is to detect the degree of underascertain-ment of NHI system, particularly in early period. In our study, there were 36.6% (52) cases ascertained in NTUH but not found in NHI system. This suggests that the direct use of NHI system for estimation of prevalence of IS may lead to an underestimate.

It could be also argued that the estimation using the capture–recapture method requires some assump-tions. For instance, it was assumed that the

probabil-ity of being listed in NHI system was independent of that identified in NTUH. In addition, we also assumed that the characteristics of IS cases in other hospitals in the same period and not listed in NHI system would be similar to that of the 52 IS cases in NTUH and not identified in NHI system. These two assumptions were not unreasonable since the NHI system covered almost 95% of the population in Taiwan and sub-jects with IS seeking medical services might be rather similar.

Although capture–recapture method provides an efficient method for the estimation of prevalence as shown in previous studies on epilepsy to cope with in-complete case-ascertainment data (Bobo et al., 1994; Hook and Regal, 1995; Murphy et al., 1995; Debrock et al., 2000). It is still possible subject to the under-estimation of IS due to cases of death, particularly in a retrospective analysis. To cope with this problem, we used a retrospective cohort study design to take this concern into account. Results of case-fatality rate give 12% attrition due to death during 13-year follow-up. The underestimation could be possible without using this design. However, the estimates of incidence from NTUH data may also be skewed if the obstetric service caters to high risk pregnancies. This may lead to an underestimates of IS. Nevertheless, we believe the impact of mortality on the underestimates of IS is minor because only 12% attrition due to death was ascertained.

4.3. Comparison with other studies

The prevalence and incidence rates of IS estimated in our study were far lower than those from Western countries, such as in Finland (Riikonen and Donner, 1979), Atlanta, GA, USA (Trevathan et al., 1999) and Tokyo, Japan (Tsuboi, 1988). We believe lower incidence or prevalence of IS compared with other countries is not due to methodological problem but possible due to two causes. First, there are certainly other epilepsy centers in Taiwan and the assumption that NTUH has the potential to see all IS cases may be an overestimate. This could explain the lower estimates in our results compares with those from the previous studies. Second, underdiagnosis is the other possibility. It should be noted capture–recapture method can cope with incomplete ascertainment of IS cases but can not solve the problem of


under-diagnosis. The problem of underdiagnosis should be explored in the future study.

The case-fatality rate reported from previous stud-ies (Lacy and Penry, 1976; Hrachovy and Frost, 1989; Matsumoto et al., 1981; Ohtahara, 1984) was between 5 and 22%, and that in our study was approximately 12%. This finding suggested that the quality of treat-ment on IS in Taiwan may not be worse than that in western countries. Among deaths of IS cases, status epilepticus is the leading cause of death, which under-scores the importance of advanced seizure control.

In conclusion, we have demonstrated an efficient method to estimate the prevalence and incidence rate of IS in Taiwan. These estimates enable one to make a clear understanding of the impact of the disease burden of IS in this society. Using capture–recapture method is helpful for the association of ascertainment of IS and medical care system, such as NHI.


This work was supported by grants NSC 89-2314-B-002-145 from the National Science Council and NTUH 91-S-072, 91-A-21-3 from National Taiwan University Hospital.


Bobo, J.K., Thapa, P.B., Anderson, J.R., Gale, J.L., 1994. Acute encephalopathy and seizure rates in children under age two years in Oregon and Washington State. Am. J. Epidemiol. 149, 27–38.

Commission on Classification and Terminology of the International League Against Epilepsy, 1989. Proposal for revised classification of epilepsies and epileptic syndromes. Epilepsia 30, 389–399.

Cowan, L.D., Hudson, L.S., 1991. The epidemiology and natural history of infantile spasms. J. Child Neurol. 6, 355–364. Cowan, L.D., Bodensteiner, J.B., Leviton, A., Doherty, L., 1989.

Prevalence of the epilepsies in children and adolescents. Epilepsia 30, 94–106.

Debrock, C., Preux, P.M., Houinato, D., Druet-Cabanac, M., Kassa, F., Adjien, C., Avode, G., Denis, F., Boutros-Toni, F., Dumas, M., 2000. Estimation of the prevalence of epilepsy in the Benin region of Zinvie using the capture–recapture method. Int. J. Epidemiol. 29, 330–335.

Gibbs, F.A., Gibbs, E.L., 1952. Atlas of Electroencephalography. II. Epilepsy. Addision-Wesley, Reading, MA.

Hauser, W.A., 1994. The prevalence and incidence of convulsion disorders in children. Epilepsia 35 (Suppl. 2), S1–S6. Hook, E.B., Regal, R.R., 1995. Capture–recapture methods in

epidemiology methods and limitations. Epidemiol. Rev. 17, 243–264.

Hrachovy, R.A., Frost Jr., J.D., 1989. Infantile spasms: a disorder of the developing nervous system. In: Kellaway, P., Noebels, J.L. (Eds.), Problems and Concepts in Developmental Neurophysiology. Johns Hopkins University Press, Baltimore, MD, pp. 131–147.

Hrachovy, R.A., Frost, J.D., Kellaway, P., 1984. Hypsarrhythmia: variations on the theme. Epilepsia 25, 317–325.

Kellaway, P., Hrachovy, R.A., Frost, J.D., Zion, T., 1979. Precise characterization and quantification of infantile spasms. Ann. Neurol. 6, 214–218.

Lacy, J.R., Penry, J.K., 1976. Infantile Spasms. Raven Press, New York.

Liou, H.H., Oon, P.C., Lin, H.C., Wang, P.J., Chen, T.H.H., 2001. Risk factors associated with infantile spasms: a hospital-based case-control study in Taiwan. Epilepsy Res. 47, 91–98. Ludvigsson, P., Olafsson, E., Sigurthardottir, S., Hauser, W.A.,

1994. Epidemiologic features of infantile spasms in Iceland. Epilepsia 35, 802–805.

Matsumoto, A., Watanabe, F., Negoro, T., Sugiura, M., Iwase, K., Hara, K., Miyazaki, S., 1981. Long-term prognosis after infantile spasms: a statistical study of prognosis factors in 200 cases. Dev. Med. Child. Neurol. 23, 51–65.

Murphy, C.C., Trevathan, E., Yeargin-Allsopp, M., 1995. Prevalence of epilepsy and epileptic seizures in 10-year-old children: results from the Metropolitan Atlanta Developmental Disabilities Study. Epilepsia 36, 866–872.

O’Callaghan, F.J.K., Shiell, A.W., Osborne, J.P., Martyn, C.N., 1998. Prevalence of tuberous sclerosis estimated by capture–recapture analysis. Lancet 351 (9114), 1490. Ohtahara, S., 1984. Seizure disorders in infancy and childhood.

Brain Dev. 6, 509–519.

Rantala, H., Shields, W.D., Christenson, P.D., Nielsen, C., Buch, D., Jacobsen, V., Zachau-Christiansen, B., Uhari, M., Cherry, J.D., 1996. Risk factors of infantile spasms compared with other seizures in children under 2 years of age. Epilepsia 37, 362–366.

Riikonen, R., Donner, M., 1979. Incidence and etiology of infantile spasms from 1960 to 1976: a population study in Finland. Dev. Med. Child. Neurol. 21, 333–343.

Trevathan, E., Murphy, C.C., Allsopp, M.Y., 1999. The descriptive epidemiology of infantile spasms among Atlanta children. Epilepsia 40, 748–751.

Tsuboi, T., 1988. Prevalence and incidence of epilepsy in Tokyo. Epilepsia 29, 103–110.

Watanabe, K., 1998. West Syndrome: etiological and prognostic aspects. Brain Dev. 20, 1–8.


Fig. 1. Study design for estimating prevalence, incidence and case-fatality rates.
Fig. 1. Study design for estimating prevalence, incidence and case-fatality rates. p.2