Efficacy of Early Intervention for Infants With Cerebral Palsy in an LMIC: An RCT Free

This single-blind randomized control trial (RCT) of LEAP-CP was conducted in West Bengal, India from March 2017 to November 2020.²⁵  Ethics approval was gained through the Children’s Health Queensland Hospital and Health Service Human Research Ethics Committee (HREC/16/QRCH/214), The University of Queensland Medical Research Ethics Committee (2016001073), Apollo Gleneagles Hospital Kolkata Institutional Ethics Committee (IEC/2016/12/35), and the Dr BC Roy Postgraduate Institute Institutional Ethics Committee (BCH/ME/PR/54). All families gave written informed consent to participate.

Infants aged 12 to 40 weeks corrected age (CA) with birth or infant identifiable risk factors who were screened as “high risk of CP” and resided in a study geographical area (2 rural districts, with community and tertiary hospital referrals; 1 urban informal settlement or slum with community referral) were referred to the trial.²⁵  CP risk was determined using the General Movements Assessment or Hammersmith Infant Neurologic Examination (HINE). Infants were defined at risk if General Movements Assessment at 12 to 17 weeks CA was classified as absent or abnormal fidgety, or HINE was less than 56, 59, and 62 at 3, 6, and 9 months CA (respectively).²⁶  Infants with known or suspected congenital or chromosomal abnormalities (based on medical records) that were likely to affect their neurodevelopmental outcome, diagnosed with neurodegenerative conditions, or considered medically fragile (determined by primary physician) were excluded.

Infants with birth or infant detectable risk factors were identified in the community through field visits by community disability workers (CDWs, n = 11) who were responsible for both participant identification and delivery of the intervention with support from the child-health site coordinator. CDWs identified potential participants using multiple identification strategies, including expansion of existing services (routine immunization clinics, primary health centers, community antenatal or primary care); conducting community activities or education alongside recruitment; and household screening. Infants in the hospital were identified by discharge record review and screened at Special Care Nursery follow-up clinics.

After baseline assessment, an independent researcher randomized infants to LEAP-CP or HA using simple randomization based on computer generated sequences, stored centrally (REDCap v8.0.2 database, Vanderbilt University, USA 2009). Twins were randomized as a single family unit. The intervention and comparator were labeled with neutral terms (“Play Program” and “Health Program”) to ensure caregivers receiving the intervention and CDWs administering the intervention did not know which intervention was active versus comparison.

Infants received their assigned intervention from enrollment for 15 fortnightly visits or until they reached 18 months CA (whichever occurred first). Infants from both study arms were visited in the home by a CDW and were encouraged to access medical or allied health services available in their community.²⁵ 

LEAP-CP is a peer-delivered multidomain intervention provided in the home, contextualized for LMICs.²⁵  It is based on principles of active goal-directed training (parent-identified goals); responsive parenting, building caregiver capacity, and caregiver mental health grounded in Acceptance Commitment Therapy (parent education); and environmental enrichment, including cognition (CP Learning Games).^27,28  CDWs, who were women from the local community, provided support to significant family members, coached goal-directed tasks, and engaged in collaborative education and decision making. CDWs were supported through fortnightly training by a treatment coordinator (child health professional).

HA was based on the World Health Organization’s Integrated Management of Childhood Illness Key Family Practices (breastfeeding, complementary nutrition, hygiene practices, vaccinations, management of illness, maternal mental health).²⁵  This included clinical signs indicating referral to existing health services.

Outcomes at enrolment and study completion (18 months CA) were conducted by a masked assessor (S.C.) trained by lead authors (K.B. and R.B.).²⁵ 

Infant function was assessed using the PEDI-CAT, a parent-reported measure including self-care, mobility, and social function.²⁹  The mobility domain was selected as the primary infant outcome.

Caregiver mental health symptoms were assessed using the DASS, a self-reported norm-referenced measure.³⁰ 

Secondary infant developmental outcomes included goal attainment (Canadian Occupational Performance Measure³¹ ), motor (Peabody Developmental Motor Scales – second edition (PDMS-2)³² ), cognition or language (Bayley Scales of Infant Development — third edition (BSID-III)³³ ), vision (Near Detection Scale³⁴ ), nutritional status (length, weight, and BMI Z-scores³⁵ ), neurologic status (HINE²⁶ ); home environment (Home Observation for Measurement of the Environment (HOME) Inventory³⁶ ); and caregiver outcomes (Parenting Sense of Competence Scale [PSOC]³⁷ ).

Information collected at enrollment included caregiver-reported birth and developmental history,^6,38  family economic status (Poverty Measurement Tool [PMT])³⁹  and Social Risk Index [SRI]).⁴⁰  At 18-months CA, a differential diagnosis of “CP and alike conditions” (motor conditions of presumed central origin) was determined^41,42 ; and classification of the Gross Motor Function Classification System (GMFCS) (<2 year age band)⁴³ ; motor type or distribution.⁴⁴  These were classified from videoed clinical assessment by an Australian pediatrician (>25 years experience) masked to clinical history.

The PEDI-CAT mobility domain can detect a 1 SD change after ≥3 months of physiotherapy in children <3 years.⁴⁵  With 64 children per group, a medium effect size ≥0.5 SD between-groups is detectable (α = .05, β = .8); total n = 142 (10% attrition).²⁵ 

Generalized linear models with Gaussian families and identity links were constructed to compare between-group differences on primary (PEDI-CAT mobility domain scaled score, DASS total score) and secondary outcomes. Models were adjusted for baseline scores. Missing data were not imputed, and a complete-case analysis was performed. Analyses were conducted following the Intention-To-Treat (ITT) principle. Analyses were repeated using a modified-ITT (mITT) sample that excluded participants with typical development at 18 months CA. Secondary analyses tested between-group and within-group pre to post differences stratified by GMFCS (I–II, III–IV, and V). The stratification by GMFCS differs from the a priori analysis plan to stratify according to the covariable of neurologic severity (binary HINE score < 40). As we had incomplete data for the HINE at 18 months, and HINE was being used as a proxy for gross motor function (HINE < 40 indicates nonambulatory CP), we opted to use the GMFCS classification. Finally, the association between participant characteristics and outcome measures was explored. Formal adjustment for multiplicity was not performed, and these findings should be interpreted cautiously. All analyses were undertaken using Stata V.13.1 (Stata statistical software, College Station, TX, USA). Significance was set at P < .05.

There were 165 infants screened as “high risk of CP” and invited to participate (Fig 1). 153 infants (93%) were randomized (LEAP-CP n = 77; HA n = 76; 86 males; mean age 7.1±2.7 months CA; Table 1). Three sets of twins participated (n = 2 LEAP-CP, n = 1 HA). There were 32 infants lost to follow-up and 3 discontinued (LEAP-CP n = 18, retention 77%; HA n = 17, retention 78%; Fig 1). No significant differences were found between participants completing and dropouts for sex, age, or HINE score at baseline.

Demographic and baseline characteristics are presented in Tables 1 and 2. Groups were generally similar, although there was a greater nonsignificant proportion of families in LEAP-CP compared with HA who were (1) “very poor” families (14% vs 8%), (2) primary caregivers with no formal schooling (10% vs 4%), (3) infants classified as severe-profound undernutrition (36% vs 22%), and (4) infants with poor vision (42% vs 32%).

Families completed a median interquartile range of 15 (0) visits over a mean duration of 11.6 months (LEAP-CP = 15 [0]; HA = 15 [0]). In the LEAP-CP and HA groups, 75% and 82% (respectively) completed ≥12 visits (93% and 92% with final outcome data, respectively). Three participants were >50 weeks CA when screened because of misreported age when eligibility was confirmed.

At study completion, infant motor function (PEDI-CAT mean difference [MD] = 0.8 [95% confidence interval (CI) −1.9 to 3.6], P = .54) and caregiver mental health (DASS MD = 0.4 [95% CI −4.9 to 5.6], P = .90) did not differ significantly between-groups (Table 3).

The secondary outcomes of infant function, motor, cognition, language, goal-attainment, vision, neurologic status, home environment, nutritional status, and parenting competence did not differ significantly between-groups (Table 3).

There were 96 infants diagnosed as having CP and alike conditions who were included in mITT analyses (Table 4). Overall, only length Z-scores showed statistically significant between-group differences, which favored HA (MD −0.7 [95% CI −1.4 to 0.0], P = .05). When stratified for GMFCS, children classified GMFCS I or II in the LEAP-CP group scored better than HA on the PEDI-CAT mobility domain (MD = 4.0 [95% CI 1.4 to 6.5], P = .003) and the HINE (MD = 5.5 95% CI = 0.0 to 11.1], P = .05).

There were significant improvements between enrollment and study completion for all outcomes overall for both study arms (Table 4). There were, however, no significant changes on (1) weight and BMI Z-scores (HA) and (2) length Z-scores (LEAP-CP, HA). There were significant declines on (1) PDMS total motor quotient (TMQ) (LEAP-CP, HA), and (2) BSID-III language composites (LEAP-CP, HA). Improvements were observed for all outcomes for children classified GMFCS I and II in both study arms; except for no significant change on (1) HOME (LEAP-CP), (2) length Z-score (LEAP-CP, HA), and (3) weight and BMI Z-scores (HA); and significant declines on (1) TMQ (LEAP-CP) and (2) BSID language composite (LEAP-CP, HA).

Child factors were associated with a greater number of study outcomes compared with family or community factors (Supplemental Table 5 and 6). The family and community factors most frequently associated with better scores on outcomes were site (hospital referrals had poorer scores), lower poverty (PMT and SRI), and older caregiver age. Of broader influence, better child function (GMFCS, vision, undernutrition, and seizures) had significant positive associations with almost all outcomes. GMFCS was also associated with exposure variables of site, child’s age, vision, undernutrition, and seizures.

In this large pragmatic RCT of a peer-delivered multidomain early intervention for infants at risk for CP in a LMIC there were no significant overall improvements for infants or caregivers receiving LEAP-CP compared with dose-matched peer-delivered HA at 18 months CA. LEAP-CP did, however, improve parent-reported motor skills (PEDI-CAT) and neurologic status (HINE) in ambulatory children with CP (on secondary subgroup analysis). This is consistent with the literature from high-income countries, showing ambulatory children with CP have greater response to targeted motor training, with little to no response reported in nonambulatory children.^46 –48  There were also clinically and statistically significant improvements observed on almost all child and caregiver outcomes for both intervention arms during the course of the study, which suggests that the parent-delivered model in LMIC may be beneficial. In the absence of a true “control arm,” it is not possible, however, to determine the extent of improvements related to the interventions, as opposed to time and child-development. The age-standardized BSID-cognition composite evidenced no-significant change during the study period, which is a positive finding given standard scores for children with CP frequently decline relative to their same aged-peers.⁴⁶  This pattern of age-referenced decline was noted on PDMS-TMQ and BSID language-composites, which suggests that although these outcomes improved in absolute terms, they did not track according to the child’s age.

The lack of between-group differences found in the primary analysis and overall mITT analysis may have been influenced by a number of factors, including the provision of an active dose-matched control program (including all infants accessing the community-based screening, which reduced the age of identification), greater severity of the sample (representative of CP in LMIC),⁵  and trends toward greater disadvantage in the LEAP-CP group (almost twice the rates of extreme poverty, higher rates of severe-profound undernutrition and poor vision). These factors are among those that have been reported to reduce responsiveness to intervention and between-group treatment effects in other early intervention studies for infants with CP.^48,49 

Both study arms had clinically and statistically important improvements for almost all outcomes, indicating the benefits of parent support delivered by peer trainers in this context. LEAP-CP and HA were delivered by CDWs, who were women from the local community with a wealth of relevant lived experience (motherhood, having a child with a disability or complex health condition, having a physical disability themselves) and/or related work experience (special education, developmental therapy, maternal-child health, or nutrition). Although no active training component was provided to HA and efforts were made to minimize contamination, there were likely benefits arising from existing skills of the CDWs and the therapeutic alliance itself. The literature shows that peer supporters instill hope and facilitate service involvement⁵⁰ ; and that building strong therapeutic alliances encourage engagement, satisfaction, and positive outcomes in allied health programs.^51,52  Our findings suggest that the peer supported parent-delivered model used in both study arms may be effective in this LMIC context and that gains in the active control group may have obscured some of the between-group differences.

Differences in the CP etiology and brain lesions in LMICs⁶  result in a population skewed toward poorer motor function and greater rates of comorbidities.⁵  Only a third (38%) of children in the current study were independently ambulant (compared with >60% in high-income countries), consistent with the LMIC CP registers (15% to 45%).⁵  This skewed sample may result in (1) reduced ability of tools to detect change; (2) greater proportion of infants with lower response to the active training intervention (ie, GMFCS IV–V); and (3) reduced efficacy of the LEAP-CP content for the sample overall. Measuring change remains a significant challenge in heterogenous samples, as tools that are sensitive to measure change in GMFCS Vs, often have large variance in GMFCS I and IIs (as seen in the current study with the PDMS-2). This may have obscured between-group differences, particularly in the unstratified primary analysis. There was also limited assessment and medical management of comorbidities (seizures, cortical visual impairment and severe-profound undernutrition) for this study’s participants, which are known confounders to intervention responsiveness.⁵³  Considering these context specific factors, the focus of the LEAP-CP intervention content, which was informed by the GAME trial²⁷  and Learning Games curriculum (positive effects in children with diplegia)⁵⁴  may need further development (including the addition of low-cost technology) to better support infants with complex clinical profiles common in LMICs. Additionally, future studies could consider limiting the recruitment to GMFCS I and II to reduce heterogeneity.

Length Z-scores were the only other outcome with significant between-group differences (mITT) that favored HA, but this finding may be a false positive (P = .05). This difference is driven by a marked decrease in length Z-score for the LEAP-CP GMFCS-V group. Although this decrease may represent a true deceleration in linear growth, it is likely caused by measurement error, which can be challenging in infants.⁵⁵  Both LEAP-CP and HA received equivalent input regarding feeding and nutrition. A myriad of factors influence growth in this context, and nutrient intake may have been insufficiently modified to facilitate catch-up growth in a context where stunting is prevalent even in the general population (31%).^56,57  Despite this, weight and BMI Z-scores of the LEAP-CP GMFCS-I and II group improved significantly from baseline to final, which was not observed for HA. Longer term follow-up is required to determine the effects of LEAP-CP on growth.

A significant difference was found in parenting competence (PSOC) during the course of the study for LEAP-CP (for all GMFCS levels), whereas a significant improvement was only detected for GMFCS-I and II for HA (mITT). This pattern was further reflected in changes to caregiver mental health postintervention. Caregiver mental health improved significantly for both arms overall and for GMFCS-I and II. There were additional improvements post-LEAP-CP for caregivers of infants GMFCS-III and IV for stress, and GMFCS-V for anxiety. This suggests that LEAP-CP may have further supported perceived competence of caregivers of children with nonambulatory CP above and beyond HA, which may have had associated benefits for caregiver stress and anxiety. LEAP-CP may have supported this through (1) emphasizing parental values, experiential acceptance, mindfulness, and responsive caregiving (informed by Acceptance Commitment Therapy)^58,59 ; (2) peer-delivered support and capacity building of the caregiver as the primary change agent (education, coaching, collaborative problem-solving, picture-based home programs)^60,61 ; and (3) prioritizing caregiver-identified meaningful goals. Improvements to parenting competence and mental health after relationally-focused early intervention is supported in the literature.⁶² 

This large pragmatic RCT of a novel peer-delivered intervention tailored to the local context was conducted across 3 geographies and partner agencies from different sectors (government, nongovernment, private) in West Bengal India. It was delivered as a free home-based service to families from vulnerable circumstances with low health literacy and was embedded within the organizational and community contexts. These implementation choices brought richness to understanding the program in this real-world setting but also presented some limitations that may have reduced the between-group differences needed to demonstrate efficacy. Most notably, the active control arm (World Health Organization HA) poses challenges to disentangle the effects resulting from the LEAP-CP content versus the service delivery model or shared content for both study arms (fortnightly peer home visits, including feeding, nutrition, and health advice). In addition, given the context and service delivery model, it is plausible that families with very low program engagement participated, which may have reduced intervention effects compared with other studies in LMICs (particularly center-based programs). A measure of family’s engagement with the program and daily caregiver-dose were not measured, and this may have provided greater indication of the quality and quantity of enactment and its impact on outcomes.

Almost a fifth of the sample included in the primary analysis had typical development based on the eligibility screening in the community. The better outcomes of children with typical development may have obscured overall between-group differences for the CP sample. This limitation was addressed by conducting mITT analysis (excluding children with typical development); however, this also decreased the sample size. Power was further compounded by large variance in many outcomes and attrition >20%. Use of the GMFCS in infants at 18 months CA is also less accurate than use in later childhood and may have influenced the stratified analysis.⁶³ 

For pragmatic and ethical reasons, the study did not exclude children with significant comorbidities, including uncontrolled seizures, severe-profound undernutrition, significant cortical visual impairment; all of which are known to reduce intervention responsiveness. Finally, as this RCT was conducted in conjunction with the LEAP-CP program’s first implementation, the CDW’s fidelity may have been reduced while workers gained experience and confidence in delivering the program.

This large pragmatic RCT of the LEAP-CP multidomain early intervention compared with dose-matched HA in an LMIC found no significant between-group differences. Consistent with what is known about targeted training, children with ambulatory CP receiving LEAP-CP did, however, have better motor outcomes. LEAP-CP may have led to better outcomes in parenting competence and caregiver mental health in children GMFCS III through V. There were significant differences from baseline to postintervention on almost all outcomes (caregiver mental health, motor, cognition, language, goal performance, vision, home environment, nutritional status, parenting competence), which supports further investigation into the efficacy of the peer-delivered model for families with an infant with CP living in LMICs.

We thank the invaluable contributions of the implementing partner organizations for the LEAP-CP trial, specifically, the teams at the Child in Need Institute (Ms Monidipa Ghosh and Ms Shirin Parveen); Asha Bhavan Centre (Mr Johnmary Barui, Ms Mohua Manna); Dr BC Roy Postgraduate Institute of Pediatric Sciences (Mr Debasis Gantait). We would also like to acknowlede the contribution of Apollo Gleneagles Hospital as the host of the Endeavour Scholarship (Mr Jewel Chakraborty); and the Indian Institute of Cerebral Palsy as research partner (Dr Reena Sen). Finally, we acknowledge the support of the Queensland Cerebral Palsy and Rehabilitation Research Centre clinical research team (Ms Bernadette Shannon, Ms Kym Morris, Ms Christine Finn) for their clinical inputs and role as advanced General Movements Assessment raters.

BSID-III

Bayley Scales of Infant Development, 3rd edition

CA

corrected age

CDW

community disability worker

CP

cerebral palsy

DASS

Depression Anxiety and Stress Scale

GMFCS

Gross Motor Function Classification System

HA

Health Advice (program)

HINE

Hammersmith Infant Neurologic Examination

HOME

Home Observation for Measurement of the Environment

LEAP-CP

Learning Through Everyday Activities for Parents of Infants at Risk for Cerebral Palsy

LMIC

low-middle income country

MD

mean difference

PDMS-2

Peabody Developmental Motor Scales, 2nd edition

PEDI-CAT

Pediatric Evaluation of Disability Inventory-Computer Adaptive Test

PMT

Poverty Measurement Tool

PSOC

Parenting Sense of Competence Scale

RCT

randomized controlled trial