Clinical characteristics of epilepsy in resource‐limited communities in Punjab, Northwest India

Abstract Objectives To describe clinical characteristics of a community‐based epilepsy cohort from resource‐limited communities in Punjab, Northwest India. Methods The cohort was gathered following a two‐stage screening survey. We cross‐sectionally examined and followed up the cohort for one year. A panel of neurologists assigned seizure types, syndromes, and putative etiologies and categorized drug responsiveness. Results The cohort of 240 included 161 (67.1%) men, 109 (45.4%) illiterates and 149 (62.1%) unemployed. Current age was >18 years in 155 (64.6%) but age at epilepsy onset was <18 years in 173 (72.1%). Epilepsies due to structural and metabolic causes were diagnosed in 99 (41.3%), but syndromic assignments were not possible in 97 (40.4%). After one year, drug‐resistant epilepsy was established in 74 (30.8%). Perinatal events (n = 35; 14.6%) followed by CNS infections (n = 32; 13.3%) and traumatic brain injury (n = 12; 5.0%) were common risk factors. Most of those with CNS infections (n = 19; 63.3%), perinatal antecedents (n = 23; 76.7%), and other acquired risk factors (n = 27; 90.0%) presented with epilepsy due to structural and metabolic causes. Perinatal events were the putative etiology for nearly 40.7% of generalized epilepsies due to structural and metabolic causes and 28.2% of all epilepsies with onset <10 years. Significance Existing classifications schemes should be better suited to field conditions in resource‐limited communities in low‐ and middle‐income countries. The finding of drug‐resistant epilepsy in nearly at least a third in a community‐based sample underscores an unmet need for enhancing services for this segment within healthcare systems. Perinatal events, CNS infections, and head injury account for a third of all epilepsies and hence preventative interventions focusing on these epilepsy risk factors should be stepped up.


| METHODS
We cross-sectionally assessed a sample of people with epilepsy and then followed them up for one year to characterize their clinical, imaging, and electroencephalographic (EEG) characteristics as well as drug responsiveness and outcomes. People were recruited to a cluster-randomized trial of home-based care following a community-based, two-step survey (described elsewhere 20 ) during which 59 509 people were screened for epilepsy in urban and periurban rural areas of Ludhiana in the Northwest Indian state of Punjab. Briefly, purposetrained field workers carried out a door-to-door screening in 24 clusters of around 2000 people each ( Figure 1) using a previously validated questionnaire between May 01, 2017, and June 22, 2018. Screen-positive people were then invited for evaluation by neurologists (including a pediatric neurologist) specialized in epilepsy at a tertiary-care hospital facility between May 23, 2017, and July 01, 2018 (Table S1). During this evaluation, they had a standard 1-to 2-hour awake and sleep (if possible) EEG recording (24 channel, Xltek, Ontario, Canada) and MRI brain scan using an epilepsy-appropriate protocol on a 3T scanner (Skyra, Siemens, Munich, Germany; 20 channels for head coil). Sequences in the scanning protocol included a T1-weighted 3D-volumetric acquisition, in addition to T1 and T2 oblique coronals and susceptibility weighted images.

| Clinical characteristics
Three neurologists jointly reviewed case notes. They confirmed diagnoses, elucidated semiology, and established seizure types, syndromes, and etiologies by consensus using standard terminology and the current conceptual and operational (2014) definitions of epilepsy. 16,17 The basic and advanced versions of the 2017 ILAE epilepsy syndromic classifications were used. 18 Attributable causes were likewise formulated according to a simplified and a more detailed scheme using structured forms. 21 The simplified scheme encompassed adverse perinatal events, CNS infections (neurocysticercosis included), other acquired cerebral insults (stroke, TBI, and tumors), presumed genetic generalized, and self-limited focal epilepsies and miscellaneous conditions. Perinatal disorders were considered to be causal, when interrogations disclosed premature birth, unattended (at home) delivery, prolonged labor, instrumental, or operative delivery with postnatal encephalopathy and/or poor feeding and excessive crying. 22 Neurocysticercosis was diagnosed and further classified into solitary active, solitary inactive, and multiple (mixed stage) forms according to recent consensus criteria. 23

Key points
• Little is known about the clinical characteristics of epilepsy in resource-limited settings in the South Asia.
• Perinatal adverse events, CNS infections, and traumatic brain injury are risk factors in nearly one-third of epilepsies in community samples.
• Perinatal adverse events, CNS infections, and traumatic brain injury are preventable risk factors for epilepsy.
• At least one-third of all epilepsies in communitybased samples might be drug-resistant.

| Follow-up
Participants were initiated (or continued, if already on appropriate treatment) on antiseizure medications (ASMs) when indicated using a pragmatic approach. They were followed up monthly either at a stand-alone clinic at the Government District Hospital or during home visits by a study nurse who could schedule interim consultations with the study neurologists at hospital for the participants in the home-based care group. Participants attending clinic were allowed unscheduled visits, and all could make emergency visits as required. A different researcher recorded seizure control, medication adherence, and adverse effects in the previous month during separate home visits. All participants received ASMs free of charge on a monthly basis. Previously used ASMs, currently used, and during one year of follow-up, maximal doses, formulations, seizure control, and adverse effects during their use and reasons for discontinuation (if applicable) were recorded. Previous use of ASMs was recorded at initial evaluation and subsequent use recorded prospectively. We stratified seizure outcome following each ASM used into three groups in keeping with current ILAE criteria for DRE. 19 These were as follows: (1) Complete seizure freedom with or without adverse effects over three times the pretrial interseizure interval or 12 months, whichever was longer, (2) Treatment failure (occurrence of seizures during the specified interval above) with or without adverse effects, (3) Undetermined outcome when the observation period was less than three times the longest pretrial interseizure interval.
Subsequently, the ILAE definition of drug-resistant epilepsy was applied to categorize epilepsy as drug-resistant, drug-responsive, or indeterminate.
Once a participant completed a year of follow-up, neurologists by consensus reassigned seizure, syndrome, and attributable causes, and entered into structured forms.

| Comorbidities
We adapted complains' inventories from an epilepsy-specific somatic comorbidity index and from a Canadian tool for psychiatric comorbidities to record somatic and psychiatric comorbidities during follow-up. 24,25 Spontaneous medical and psychiatric complaints during follow-up were also noted. Formal medical or psychiatric and psychometric assessments were requested to deal with problems arising. Medical and psychiatric comorbidities were inferred from review of the case notes and matched to the inventories.
In women with epilepsy in the reproductive age group (18-45 years), folic acid and contraceptive method use during the follow-up period were noted. The occurrence of pregnancies, terminations, and outcome of each pregnancy over one year were recorded.

| Sociodemographic categorization
During the initial evaluations, we ascertained household income, education, and employment to assign socioeconomic status according to the Kuppuswamy scale (revised, 2015 version). 26 This validated scale, widely used in India, assigns deprivation level to one of five ordinal classes.

| Statistical analysis
We used descriptive statistics to analyze sociodemographic variables and the proportions of various seizure types, epilepsy syndromes, etiological risk factors, drug resistance, and comorbidities. To establish associations between various syndromic diagnoses and attributable causes and various sociodemographic variables, we used univariate analyses (with the chi-square test) and multinomial regression. For both analyses, we consigned presumed genetic generalized epilepsies as the reference category. We obtained relative risk ratios by exponentiating the linear equations generated and regression coefficients for unitary changes in the predictor variables estimated. Lastly, Venn diagrams created to explore the relationship between different seizure types, syndromes, and etiologies. Stata ver. 15 (StataCorp, TX, USA) was used for the analyses.
The study was scrutinized and approved by the Institutional Ethics Committee of Dayanand Medical College & Hospital, Ludhiana, India (IEC no. 2017-281) and registered with the Clinical Trial Registry of India (Re.: 2017/09/015380). Participants provided written informed consent and parents or guardian provided assent for minors or those not having capacity.

| RESULTS
We recruited 240 individuals between 1 and 80 years of age and of whom, 161 (67.1%) were males. EEGs were recorded in 237, and an MRI was performed in 200. At enrollment, 230 (95.8%) had a net family income of less than US $ 250/month, 149 (62.1%) were unemployed with 43 (17.9%) on daily wages, and 152 (63.3%) were either illiterate or had barely completed primary school (Table 1). Thus, 194 (80.8%) were in the lower segment on the Kuppuswamy scale. 26 Epilepsy duration varied from 1 to 62 years (median: 11 years; 95% confidence intervals, 5-21 years). Over half of the participants (53.3%) had had seizures for more than 10 years. Age at onset was <5 years in 75 (31.3%), <10 years in 108 (45.0%), and less than 18 years in 173 (72.1%) people. Age at neurological evaluation was over 18 years in 155 (64.6%) people. Prior to the neurological evaluation, 116 (48.3%) reported at least one seizure in a month.
Putative etiologies for focal epilepsies due to structural and metabolic causes included CNS infections (n = 29; 29.9%), perinatal risk factors (n = 23; 23.7%), and other acquired epilepsy risk factors (n = 27; 27.8%). In those with focal seizures (n = 47), epilepsy could be attributed to acquired risk factors in 13 (27.7%), to perinatal risk factors in 11 (23.4%), and to CNS infections in 9 (19.2%) (Figure 2A-C). The probabilities in someone with both focal and generalized seizures and a syndromic diagnosis of epilepsy due to structural and metabolic causes (n = 16) could be attributed to acquired risk factors were 18.8%, to perinatal risk factors were 31.3%, and to CNS infections 12.5% ( Figure 2D-F). Most participants (n = 34; 97.1%) with genetic generalized epilepsy had generalized seizures ( Figure 2G), and in comparison only one (6.3%) of those with prior CNS infections ( Figure 2I) and six (37.5%) with perinatal insults ( Figure 2J) had generalized seizures. Of those with generalized seizures (n = 128; 53.3%), 37 (29%) were considered to be having presumed genetic generalized epilepsies and 16 (12.5%) generalized epilepsy due to structural/metabolic causes. Among those with a syndromic diagnosis of generalized epilepsy due to structural/metabolic causes (n = 16), perinatal risk factors were identified in seven (44%) but in none and only one, respectively, were other acquired epilepsy risk factors and CNS infections identified.

| Drug responsiveness
After one-year of follow-up, 74 (30.8%) participants fulfilled ILAE criteria for DRE. 19 Drug responsiveness could not be established in 72 (30.0%) due to an inadequate observation period. People with DRE were younger (mean age: 10 ± 10 years; P < .0001), more often developmentally delayed on a global scale (n = 15 out of 74; 20.3%; P = .05), more likely to have concomitant psychogenic nonepileptic attacks (n = 8 out of 74; 10.8%; P = .001), less likely to use alcohol and other recreational substances (n = 2 out of 74; 2.7%; P = .01), and less likely to be employed (n = 20; [27.0%] out of 74 compared to 43 [45.7%] of 94 with DRE; P = .01) in comparison with their drug-responsive counterparts (Table 3). Besides, a trend for greater frequency of generalized epilepsy due to structural and metabolic causes (

| DISCUSSION
The spectrum of seizures, syndromes, and attributable causes with imaging and EEG findings in community-based cohorts from South/South-East Asian Region have been sparingly described previously. 15,27,28 Here, we reflect on some distinctive findings in our sample. The age of epilepsy onset in our cohort was consistent with other reports from other resource-limited settings. For instance, over half of Kenyan and Honduran population-based cohorts had onsets before age, 10 years. 7,29 In comparison, less than one-third of a prevalent sample from Rochester, Minnesota, USA, dating back to 1980s had onset < 9 years of age. 30 Epilepsy duration in our cohort was similar to Rochester, of whom over half had epilepsy for more than 10 years. 30 Predictably, nearly two-thirds in our cohort were over 18 years of age. The gender bias in our sample with women representing only a third merits discussion. In Honduras, the survey found higher prevalence in women, and in Kenya, women constituted a little over half in a representative sample of people with epilepsy. 7,8,29 When we examined the segment of screen-positive people not attending neurological evaluation, the proportion of women was higher but this only partly explained the underrepresentation of women in our final sample. In many communities with limited resources in South-East Asia, women are not able to travel and this could preclude some of them to attend the hospital for evaluation. 31 Stigma associated with disclosure widespread in the region might also have precluded attendance or even disclosure during the screening campaign. 32 The higher proportion of generalized seizures in comparison with focal seizures in our sample is at variance from other studies. 8,33 Some of the generalized seizures might have in reality been focal seizures evolving to bilateral tonic-clonic seizures. An under-reporting of focal seizures might also have been due to cultural issues. Even so, the finding underscores the difficulties in classifying seizures in population-based settings. Problems associated with assigning syndromes in population-based samples are well known. 34,35 The complex relationship, however, between seizure types, epilepsy syndromes, and attributable causes is illustrated by the Venn diagrams (Figure 2A-H).
Adverse perinatal events were the commonest risk factor in our cohort and higher than in some recent studies from sub-Saharan Africa. Several settings including premature births, unattended home deliveries, cesarean sections followed by postnatal events such as hypoglycemia in institutional care, all more frequent in resource-limited communities have been associated with epilepsies. 7,36,37,38 The estimated median fraction of epilepsies attributed to adverse perinatal events is roughly 11%. 39 Epilepsies following perinatal insults are frequently associated with comorbidities including motor deficits, mental challenge, and poor scholastic performance. 22,36,40 In our sample, individuals with an adverse perinatal event as the putative etiology of epilepsy had significantly lower educational grades and a trend toward drug resistance (Table 3). Adverse perinatal events were also more common in younger people. This is consistent with findings from previous studies of prevalence samples of epilepsies worldwide. 7,30 A lesser frequency in adults could mean that either many died prematurely or less likely, moved into remission. Nonetheless, this suggests the need to contextualize antenatal and perinatal care in resource-limited communities in LMICs. The lack of skilled birth attendants, poor capacity of health systems, and access issues such as cultural beliefs, distance, and financial limitations are barriers to quality maternal and childcare, much required for improving perinatal outcomes.
The proportion of pharmacoresistance in our sample appears high; however, this might still be an underestimate, as the categorization into DRE or not was not possible in nearly a third of the cases. 41 Not many reports of proportion of DRE in population-based samples of active epilepsy are available. An Italian study found DRE in less than a fifth of people with active epilepsy using older criteria. 42 An Egyptian study reported uncontrolled epilepsy, defined as at least a monthly seizure frequency during the previous six months, in about a third of those with active epilepsy. 43 This was after selecting out those with pseudointractability due to inappropriate drug choice or doses or due to incorrect diagnosis. Rather our findings align with a Kenyan study reporting seizures on at least a monthly basis in 58% of people. 8 We surmise that the high proportion of pharmacoresistance might be due to engaging a prevalent sample, paucity of resources, and access to appropriate care, that is, surgical treatment in impoverished communities. Regardless of the reason, DRE is associated with substantially higher disease burden in comparison with controlled epilepsy and justifies the need to scale up provisions for tackling it in healthcare systems. 34 Caution should be exercised in the interpreting our findings due to some limitations. Firstly, we included less than half of those who screened positive at the survey stage. Many did not attend the evaluation following screening. The characteristics and outcome of this group of people are described elsewhere. 44 Secondly, some potential participants may have been missed during screening due to fear of the consequences of disclosure. Thirdly, most came from a low SES category. This, however, does not mean that the communities from which our sample was drawn were comparably impoverished as we did not collect or have access to general population data. Enlistment of somatic and psychiatric comorbidities was based on spontaneous reporting rather than a structured inquiry, and this may have led to an underenumeration of the conditions. Not least, the classification of epilepsies and attributable causes often remained inadequate, emphasizing the challenges we faced, and hence an urgent need for pragmatic classifications suited to resource-limited community settings.
Adverse perinatal events, CNS infections (including neurocysticercosis), and acquired risk factors (including traumatic brain injury (TBI) and stroke) accounted for over a third of our sample. All these factors are amenable to primary prevention, and it could be argued that third of epilepsies in our sample could have been prevented. 39 Population-based interventions and initiatives to improve obstetric outcomes, eliminate, or eradicate CNS infections (eg, neurocysticercosis) and prevent stroke and TBI are warranted to reduce the burden of epilepsy in LMICs. 45,46 Currently, a handful of trials to improve obstetrical outcome in LMICs are underway, of which only one, a pragmatic trial of improved intrapartum care is assessing its impact on epilepsy incidence in India. 47,48 Primary prevention trials and initiatives should likewise target a variety of CNS infections, TBI, and stroke. 49,50 Lastly, there is a compelling need for translational research assessing secondary prevention, referring to the interruption of epileptogenesis after the occurrence of adverse perinatal events, CNS infection, stroke, or TBI to reduce the burden of epilepsy.