Commonalities in biomarkers and phenotypes between mild cognitive impairment and cerebral palsy: a pilot exploratory study

Cohorts baseline characteristics

Table 1A summarizes the baseline characteristics of the study participants. We recruited a total of 127 participants, aged mean=24.97, SD=5.29 (CP cohort) and mean=71.28, SD=6.03 (MCI cohort). Most of the participants were female in the MCI cohort (70.8%), while the CP cohort had a balanced number of genders, with 48.1% female (Table 1A, 1B). Years of formal education also differed significantly between CP and MCI cohorts (mean, SD, CP=13.49±2.25 years, MCI=4.31±4.66 years). No significant differences in BMI were observed, but all the other eight outcome variables were significantly different based on bivariate testing. Table 1B presented the specific clinical characteristics of the study participants for both diagnostic entities, specifically the CP diagnoses or subtypes, i.e. quadriplegic, hemiplegic, diplegic, or triplegic. Furthermore, we also presented the distributions of the GMFCS levels of CP and the two MCI subtypes for patients with MCI.

Commonalities between the CP and MCI cohorts through overlapping biomarkers and phenotypes

Based on the results from Table 2, there were six biomarkers and phenotypes that were statistically non-significantly different between CP and MCI. The results remained statistically insignificant upon controlling for all the available covariates in their respective models 3. They were hs-CRP levels (model 3: β= 0.221, 95% CI=-0.074 to 0.515, p=0.141), semantic fluency test (model 3: β= -1.933, 95% CI=-4.713 to 0.847, p=0.171), WAIS-V Block Design (model 3: β= 2.834, 95% CI=-3.989 to 9.657, p=0.412), BMI (model 3: β= -0.538, 95% CI=-3.334 to 2.258, p=0.704), resting heart rate (model 3: β= -4.583, 95% CI=-10.914 to 1.748, p=0.154), and diastolic BP (model 3: β= -0.226, 95% CI=-5.376 to 4.924, p=0.931).

Distinct biomarkers, neurocognitive, and anthropometric measures between CP and MCI cohorts

The natural log-transformed delta FHS score was significantly different between the two cohorts (model 3: β= 2.017, 95% CI=1.606 to 2.428, p<0.001), with CP participants having significantly lower natural log-transformed delta FHS score compared to MCI cohort, after controlling for covariates (Table 2). Similarly, a significant difference in plasma log-transformed BDNF levels between CP and MCI was observed (model 3: β= 3.976, 95% CI=3.26 to 4.693, p<0.001), with CP participants having significantly lower log-transformed BDNF levels compared to MCI cohort, after controlling for covariates (Table 2). After controlling for covariates, systolic blood pressure was also higher in MCI (model 3: β= 21.939, 95% CI=12.38 to 31.499, p<0.001). Without controlling for any covariates, the other measures, except BMI, had significant differences. Nonetheless, after controlling for covariates, the other six measures had no significant differences between CP and MCI cohorts.

Associations of hs-CRP with biomarkers, neurocognitive, and anthropometric measures

Log-transformed hs-CRP was significantly associated with log-transformed BDNF (model 1: β=1.255, 95% CI=0.458 to 2.052, p=0.002). The relationship became borderline significant after adjusting for additional covariates in the final model (model 4: β=0.425, 95% CI= -0.004 to 0.854, p=0.052) (Table 3). Furthermore, log-transformed hs-CRP was also significantly associated with BMI (model 1: β= 3.032, 95% CI= 1.469 to 4.595), p<0.001 and model 4: β= 3.122, 95% CI=1.516 to 4.727, p<0.001). Lastly, log-transformed hs-CRP was also significantly associated with natural log-transformed delta FHS score (model 4: β= 0.266, 95% CI=0.021 to 0.511, p=0.034).

Associations of natural log-transformed delta FHS score with biomarker, neurocognitive, and anthropometric measures

In Table 4A, we showed that in the CP cohort, the natural log-transformed delta FHS score was significantly associated with log-transformed hs-CRP. The relationship was significant after adjusting for additional covariates in the final model (model 4: β=0.509, 95% CI= 0.044 to 0.974, p=0.032). Natural log-transformed delta FHS score was also significantly associated with semantic fluency through adjusted models (model 3: β=5.705, 95% CI=0.997 to 10.413, p=0.018). Chronological age was a not significant covariate for all the variables of which natural log-transformed delta FHS score regressed on, except with long-transformed hs-CRP, with the addition of chronological age, the model was significantly improved (R² change=0.071, p-value of R² =0.025). Whereas in the MCI cohort, Table 4B, natural log-transformed delta FHS score had two significant associations upon adjusting for covariates, namely with BMI and systolic blood pressure.

Associations of BDNF with biomarkers, neurocognitive, and anthropometric measures

In the CP cohort, log-transformed BDNF was significantly associated with log-transformed hs-CRP (model 1: β=0.355, 95% CI= 0.085 to 0.624, p=0.011) (Table 5A). The relationship remained significant after adjusting for additional covariates through the final model (model 4: β=0.36, 95% CI=0.082 to 0.639, p=0.012). Log-transformed BDNF was also significantly associated with BMI through adjusted models (model 1: β= 2.998, 95% CI= 0.393 to 5.602, p=0.025 and model 4: β= 2.814, 95% CI=0.079 to 5.550, p=0.044). Chronological age was not significant covariate for all the variables of which log-transformed BDNF regressed on Table 5A, except for the natural log-transformed delta FHS score.

In the MCI cohort, log-transformed BDNF has no significant correlation with log-transformed hs-CRP, neurocognitive measures, anthropometric, and physiological measures (Table 5B).

Sub-group analyses based on clinical characteristics of CP

Furthermore, we performed two sub-group analyses (Supplementary Tables 1–4), with one set of analyses stratified by CP type (hemiplegic versus non-hemiplegic) and another set stratified by STMS results. Regardless of the sub-group stratifications, the presence and the lack thereof of the associations in the sub-groups were very similar to those of the total sample.

Limitations

We acknowledged several limitations which present in this study, mainly conferred by the study’s pilot and exploratory nature. First, we could not completely exclude the possibility of residual confounding effects, since our study cohorts were recruited from two different countries, including participants of different ethnicities. Hence, these findings are preliminary and require validation in larger studies. Several pertinent covariates to be taken account in future comparisons, including the batch effects across the two cohorts in examining biomarkers, the BDNF and APOE genotypes, exercise, diets, intakes of supplements, and changes in medication consumption. However, such extensive controls for potential confounders would only be feasible in large cohort studies with both clinical conditions present, which is unlikely as of present. Second, there is also contention in the literature on how well blood markers reflect brain-based biomarkers in general, particularly BDNF. Conversely, there have been increasingly overwhelming evidence supporting the utility of blood-based biomarkers to examine neurological disorders. Third, also due to the pilot and exploratory nature of this study, we did not control for multiple testing. Similar practice has been adopted by other studies of pilot and exploratory nature [52, 53]. With our encouraging pilot findings, we provided strong preliminary data for future validation studies. Fourth, we did not examine the key biomarkers for Alzheimer’s dementia, such as Tau and amyloid beta. Lastly, we examined the associations between the measures cross-sectionally, hence the proposed sequence of events presented in Figure 2 required future empirical validation utilizing longitudinal cohort study. However, to our best knowledge, this is the first hypothetical comparative aging model that is backed by preliminary data postulating the connections between CP and MCI, providing encouraging impetus and supporting future pursuance in this direction. In fact, we are following up with these participants longitudinally to validate the proposed model.

Strengths

Despite these limitations, this study made significant contributions on several aspects. To our knowledge, this study was the first to compare multiple characteristics of patients with CP and MCI directly, investigating the commonalities and differences in various biological and phenotypical measures. Furthermore, the participants were also clinically well-characterized by clinical experts in their respective fields of expertise, coupled with a number of biomarkers and clinical measures compared and contrasted across these two cohorts. The moderate sample sizes of the two cohorts also enabled us to unravel several significant associations, revealing the common and distinct biomarkers and phenotypes in these conditions, and further proposing hs-CRP as a prominent biomarker for adults with CP.

Conclusion and future directions

In all, these preliminary findings on the common and distinct biological underpinnings and phenotypes between CP and MCI are novel and encouraging. Coupled with the fact that psychosocial interventions have been demonstrated to improve both outcomes and biomarkers in a wide range of neurological conditions, we believe this approach deserves further study. Nevertheless, due to the pilot and preliminary nature of the study, there is still limited evidence at this stage to draw definite conclusions on the common and distinct biological underpinnings and phenotypes in these two populations, or to recommend interventions in targeting them. Further validation of these findings is warranted, particularly in large-scale longitudinal follow-up cohorts recruiting participants with both CP and MCI. A number of critical future directions include taking into account of different subtypes of CP, as there could be different biomarkers associated with different etiologies characterizing the varied symptoms defining the different CP sub-types. Importantly, a prominent symptom in CP is motor dysfunction, often in the form of increased muscle tone and poor motor control. Searching for biological signatures of these symptoms could help elucidate the biological underpinnings and illuminate biological targets for individualized intervention. A life-course perspective should also be considered, with annual or bi-annual follow-ups, to validate our hypothesized effect of accelerated aging trajectory of adults with CP to developing cognitive impairment and ultimately dementia. Lastly, an examination of more comprehensive biomarkers, including nutritional status, amyloid beta and tau, and neurocognitive domains, including global cognition, are imperative to understand the complex interplay between common and distinct measures further. By comparing MCI and CP with these multi-unit examinations, future studies could shed light on how adults with CP could have increased geriatric-associated pathology and accelerated aging that may further impair function, ultimately contributing to the heightened risk of developing geriatric syndromes, earlier than their peers who were not beset with a pediatric-onset condition.

Settings, study design, and participants

Colorado site (USA): Adults with cerebral palsy (CP)

The CP cross-sectional study was approved by the Colorado Multiple Institutional Review Board (COMIRB Reference No: 14-0367) and registered with the clinical trial database (https://clinicaltrials.gov/ct2/show/NCT02137005). The study was conducted at a clinical motion analysis laboratory at the Children’s Hospital Colorado. The laboratory has a specialized team of clinicians and researchers (MDs, nurses, physical therapists, biomechanists, nurses, biogerontologists and psychologists) and is internationally accredited by the Commission for Motion Laboratory Accreditation (CMLA) (http://www.cmlainc.org/). Clinical and research staff were trained in the systematic conduct of the study procedures, such as physical examination, medical history, psychological assessments, and blood collection and composition analysis, under a standard human ethics approved protocol.

Colorado cohort inclusion and exclusion criteria: Participants with a confirmed diagnosis of CP were identified from an internal patient registry comprised of approximately 526 participants, aged 18 and above. Potential research participants underwent a short telephone screening survey to confirm eligibility. Participants were included in the study if they were (1) interested and able to participate in the study, (2) previous patient from the study clinic, (3) had a medical record on file at the clinical site, and (4) had mild CP by being able to walk across a 35-foot (10.6m) walkway, with or without assistive devices, at least three times. A total of 72 ambulatory participants, who passed the study screening criteria, were enrolled [35].

CP diagnosis: Cerebral palsy (CP) is a group of disorders that affect a person’s ability to move and maintain balance and posture. CP is a lifetime disability and the most common motor disability in childhood. CP is caused by abnormal brain development, or damage, to the developing brain that affects a person’s ability to control muscles. Many individuals with CP might also have other neurological conditions such as intellectual disability; seizures; problems with vision, hearing, or speech [36]. Although abnormal gait patterns in CP are related to posture and movement impairment, the symptoms of CP vary from person to person [37]. A person with mild CP might be able to walk independently while someone with severe CP might need lifelong care. Presently, there is limited knowledge explaining how a person with CP experience aging such as how the symptoms change over time and the types of secondary health conditions they might develop as they age [14].

Gross motor function classification system (GMFCS): The Gross Motor Function Classification System (GMFCS) is a multi-level categorization tool that helps to describe varying levels of severity in people with CP [37, 38]. The GMFCS is categorized in five different levels (I, II, III, IV, V); the lower levels (I-III) correspond with milder forms of CP, while the higher levels (IV, V) indicate increased severity. The GMFCS can be used to describe all types and severity levels of CP. This classification provides both the patient and the clinician with a description of the patient’s current motor function [38].

Cerebral palsy topographical classification: The topographical classification of CP is used to diagnose and describe the body part(s) and side(s) that are affected by the condition [39]. Usually, these are described as 1) paresis for a weakened part and plegia/plegic for paralyzed; 2) monoplegia/monoparesis when only one limb is affected and hemiplegia/hemiparesis when the limb is significantly impaired; 3) diplegia/diparesis usually indicates when the legs are the part of the body that are severely affected; 4) hemiplegia/hemiparesis is used when the arm and leg on one side of the body are affected; 5) paraplegia/paraparesis means the lower half of the body is affected, including both legs; 6) triplegia/triparesis indicates that three limbs are affected (i.e. both arms and a leg or both legs and arm) as well as one upper and one lower extremity and the face; 7) double hemiplegia/double hemiparesis indicates all four limbs are involved, but one side of the body is more affected than the other; 8) tetraplegia/tetraparesis indicates that all four limbs are involved, but three limbs are more affected than the fourth; 9) quadriplegia/quadriparesis is used when all four limbs are involved; and 10) pentaplegia/pentaparesis means all four limbs are involved, with neck and head paralysis often accompanied by eating and breathing complications [39].

Mild cognitive impairment (MCI) screening in cerebral palsy (CP): The Short Test of Mental Status (STMS) [40, 41] was used to identify mild cognitive impairment in the CP cohort. STMS has been widely used in neurological clinical settings as a brief and reliable standardize cognitive screening tool for mild cognitive impairment detection. Cognitive screening tools are commonly used in the older adult population, but many of them lack sensitivity and specificity in mild cognitive impairment (MCI) detection. In addition, MCI is under-diagnosed in adults with CP. The Short Test of Mental Status (STMS) has a higher sensitivity and specificity in detecting early cognitive deficits, such as MCI, as compared to the Mini Mental State Examination (MMSE) [42, 43]. Since the CP cohort was relatively young, the study clinical team recommended the STMS over the MMSE as a more suitable cognitive screening tool for the CP cohort.

Singapore site (Republic of Singapore): Older adults with MCI

This study was approved by the National University of Singapore ethics committee, Institutional Review Board (NUS-IRB Reference No: B-14-110), and registered with the clinical trial database (https://clinicaltrials.gov/ct2/show/NCT02286791). The participants were older adults aged 60 and above, who have participated in the Mindfulness Awareness Practice randomized controlled trial, at a community-based research center established by the NUS Psychological Medicine department. The research nurses, research assistants, and a Ph.D. student obtained informed consent before screening for potentially eligible participants.

Singapore cohort inclusion and exclusion criteria: The inclusion criterion was fulfilling the operational criteria of MCI based on The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) [44]. We excluded older adults with either dementia or normal aging, had a neurological or major psychiatric condition, had a terminal illness, had visual or hearing impairments, had upper and lower limb motor difficulties, and those who were participating in another intervention at the time of the screening. To derive the cognitive status of the participants, there was a two-tier procedure. First, the assessors, comprised of a team of trained research assistants and a Ph.D. candidate, administered the clinical dementia rating (CDR) and neurocognitive assessments (NCA) to all screened participants at the research center and derived at preliminary research diagnosis. Final research diagnoses of MCI were made during the study’s consensus meetings by a panel consisting of at least two consultant-ranked psychiatrists, clinical scientists, and the trained assessors who administered the tests.

Mild cognitive impairment (MCI) diagnosis in community-dwelling older adults: A diagnosis of amnestic MCI and non-amnestic MCI was made during the MCI cohort’s monthly consensus meeting, by a panel consisting of at least two consultant-ranked psychiatrists, clinical scientists, and the trained assessors who administered the tests. A diagnosis of amnestic MCI is mostly indicative of memory loss as the predominant symptom and is a prodromal stage of Alzheimer’s disease. On the other hand, a diagnosis of non-amnestic MCI is often associated with non-memory related cognitive impairment and is a prodromal stage of other dementia types [45].

Overlapping measurements across both cohorts

After examining the datasets for the two cohorts, we identified nine overlapping measures that have been associated to aging and early cognitive impairment progression, which we divided into two categories, namely biomarkers (N=3) and phenotypic measurements, and sub-divided into two sub-classes: neurocognitive (N=2) and anthropometric measures (N=4).

Biomarker measurements across both cohorts

Bio-specimen collections: For both cohorts, blood collections were scheduled between 9:00 and 11:00 in the morning to minimize diurnal variations. The participants stopped the consumption of foods after 10 pm the night before venipuncture. The consumption of only water was advised. The participants were advised not to exercise or perform rigorous physical activities before the collections and not to rush to the centers in the case that they were late. Blood draw via venipuncture was performed by the research nurses on the day that the participants visited the research center. The blood was kept at 4° C for a maximum of three hours before being processed in the respective laboratories.

Biomarker pre-processing, storage, and measurements: The blood samples were sent to the laboratory located at the University of Colorado and Singapore Immunology Network (SIgN), for the CP and MCI cohorts, respectively. Subsequently, the whole blood samples were centrifuged at 1650×g for 25 minutes at room temperature to obtain the plasma. The plasma samples were then stored at -80° C until further analyses. After sample collections from all the participants were completed, all samples were assayed on the same day and on the same plates in the respective laboratories to avoid batch effect.

Biomarker levels were examined using commercially available enzyme-linked immunosorbent assay (ELISA) kits. A total of two overlapping biomarkers were measured, namely high-sensitivity (hs)-CRP (Tecan, Männedorf, Switzerland) and BDNF (Promega Corporation, Madison, USA). All the experiments were performed as per the instructions of respective manufacturers of the kits.

10-year Framingham heart study (FHS) measure: We utilized the equations with recommended measures from the FHS to determine the risk percentage for the development of CVD for each subject [35]. The FHS cardiovascular (CVD) 10-year risk factor estimation for the BMI-based results were used to determine the percentage of CVD risk in both cohorts [46]. Sex, age, systolic blood pressure, BMI, information on whether the participant was a smoker, had diabetes, or was on medication for hypertension were utilized to calculate the risk percentage for CVD. Based on the same set of measures, we also derived the estimates for the general population, which were obtained from the FHS database [46, 47]. Since we were comparing two cohorts with a large age gap, we derived the delta FHS score, by subtracting each subject’s FHS individual risk score from the corresponding risk estimates for the general population. The derived “delta FHS score” was used in all the subsequent analyses and regression models. Due to its skewed nature, we performed natural log-transformed on the delta FHS score, successfully transforming it to conform to statistical normality.

Phenotypic measurements across both cohorts

Neurocognitive assessments (NCA): Cognitive functions were examined using two neurocognitive tests that have been validated in both Singapore and the United States to have good psychometric properties [48, 49]. First, the Wechsler Adult Intelligence Scale (WAIS)-V Block Design is a sub-test that is administered as part of the WAIS-V test battery, primarily measures visual-spatial and organizational processing abilities, as well as non-verbal problem-solving skills [49, 50]. As it is a timed task, it is also influenced by fine motor skills. Second, semantic verbal fluency (60-seconds Animal Naming) taps lexical knowledge and semantic memory organization [50]. Hence, for both tests, the higher the scores, the better the participant’s cognitive function [51]. First, for the Wechsler Adult Intelligence Scale (WAIS)-V Block Design, the participants are presented with nine identical cubed blocks with two surfaces of solid red, two surfaces of solid white and two surfaces that are half red and half white. The participants were required to assemble pieces of blocks to match the figures presented to them within a specific time limit. As the participants completed more figures, the more complex the subsequent figures and more time were given. Total scores were calculated after the participants failed three consecutive trials on assembling the tasks. Hence, higher scores indicate higher cognitive functioning [51]. Second, semantic verbal fluency (60-seconds Animal Naming) taps lexical knowledge and semantic memory organization [50]. Participants were required to generate certain words corresponding to a specific semantic category (in this study, animals) within a 1-minute time limit. Animal is the most frequently used category given that: (1) it is a clear semantic category across languages and cultures; (2) it is a relatively easy semantic category with only minor differences among people living in different countries, different educational systems, or belonging to different generations; and (3) it is an easy-to-administer, short, and common test included in different cognitive tests. The total numbers of words were summed up. Optimal fluency performance involves generating words within a sub-category and, when a sub-category is exhausted, switching to a new sub-category. Hence, the higher the number of words, the better the participant’s cognitive function. The tests were administered by either the trained research nurses, a Ph.D. candidate or research assistants.

Anthropometric measures across both cohorts: Anthropometric data were obtained from physical examinations, administered by either the research nurses or trained research assistants. Body-mass index (BMI) was calculated by body mass divided by the square of the standing height, resulting in a unit of kg/m², with a BMI higher than 24.9 considered overweight. We measured systolic and diastolic blood pressure (BP) and resting heart rate using a blood pressure monitor and cuff which we secured around the participants’ left arm. We took three blood pressure measures using a medical-grade electronic vital sign monitor (Welch Allyn Spot Vital Signs Monitor, Welch Allyn, Skaneateles Falls, NY, USA).

Statistical analyses: comparing CP and MCI participants’ outcomes

Since this is a post-hoc exploratory study, no effect size was assumed and thus no sample size calculation was performed. All measures were expressed as mean ± standard error (SE), except gender with percentage. The differences in baseline variables were examined using Student’s t-test, chi-square or Fisher’s exact tests according to the nature of the data. The raw values of the biomarker measurements did not fulfill the normality assumption; therefore, the raw values of the biomarkers were natural log- or log-transformed for subsequent analyses and were successfully normalized, based on dot plots, skewness, and kurtosis. We performed linear regression analyses using the dummy variable participant cohort as the independent variable, associating with the biomarkers, anthropometric, and neurocognitive measures independently in investigating aim 1. To investigate aim 2, we used the respective biomarkers as the independent variables and associated it with the other biomarkers, anthropometric and neurocognitive measures, which acted as dependent variables. All the regression models controlled for a number of relevant covariates; We performed stepwise regression analyses, with the covariates sequentially entered into the regression models. Model 1 did not control for covariates, model 2 controlled for sex and model 3 further controlled for years of formal education. In further investigating aim 3, we further controlled for age of the participants in model 4. However, for regression models shown in Tables 2, 3, multicollinearity with the cohort effect occurred, and hence age was not included. In Tables 4A, 4B, 5A, 5B, and supplementary tables, since the samples were stratified based on cohorts, no issues of multicollinearity happened, and age was thus included in the model 4. For all the models, the participants did not have to complete all the assessments to be included in the analysis and missing values were replaced by mean substitutions. All the analyses were performed using the Statistical Package for the Social Sciences (SPSS) Statistics for Windows, version 24.0 (IBM Corp., Armonk, N.Y., USA). A two-tailed p-value of <0·05 was considered statistically significant. Due to the pilot and exploratory nature of this study, we did not control for multiple testing, similar to other studies of pilot and exploratory nature [52, 53]. For all the regression models 3 presented in Table 2, statistically significant p-values represent distinct whilst statistically insignificant p-values represent commonalities in measures between CP and MCI. A recent study indicated that adults with intellectual and developmental disabilities are more likely to develop early dementia [54]. Hence, the association between CP and dementia might be driven by neurologic or intellectual comorbidities, rather than as a direct effect of CP. Since we did not collect measures for ID, we used CP subtypes as a proxy, as certain CP subtypes are associated with a higher incidence of ID. We performed two sub-group analyses, with one set of analyses stratified by CP subtypes (hemiplegic versus non-hemiplegic) and another set of analyses stratified by the Short Test of Mental Status (STMS) results (screened with MCI versus without MCI).