This was a prospective study. The data was collected from June 2019 to May 2020. We recruited patients with acute ischemic stroke in Dr. Sardjito General Hospital Yogyakarta, Indonesia.

We included patients with: (1) first-ever acute ischemic stroke with <5 days after stroke onset, which is diagnosed based on clinical examination and head CT scan immediately after patients’ initial assessment in the emergency room; and (2) aged >18-years old. We excluded those with: (1) recurrent stroke patients; (2) brain tumor, brain trauma, encephalitis, Parkinson’s disease, epilepsy, pregnant females, and dementia; and (3) depression or history of psychiatric diseases.

Demographic, clinical, and laboratory data included serum BDNF were collected at day-5 of the stroke onset during hospitalization. Serum BDNF was collected at day 5 of the stroke onset during hospitalization and at day 30 during the home visit. Cognitive functioning was measured at day 5, 30, and 90. The follow-up was carried out by the study team through home visits.

We estimated the minimum sample size and found that it might be acceptable to represent the entire population. The sample size was calculated using the following for two population proportions (one-sided test) ¹⁷ : n = Z 1 − ∝ 2 P 1 − P + Z 1 − β P 1 1 − P + P 2 1 − P 2 2 P 1 − P 2 2

n = sample size

P1 = proportion of PSCI with lower serum BDNF

P2 = proportion of PSCI with higher serum BDNF

Based on previous study, the proportion of PSCI with lower serum BDNF is 88.7% ⁶ and the proportion of PSCI with higher serum BDNF is 65.5%. ¹⁸ With Z 1 − ∝ = 1.96 and Z 1 − ∝ = 0.84, the minimum sample size is as follows: n = 1.96 2 0.887 1 − 0.877 + 0.84 0.887 1 − 0.887 + 0.655 1 − 0.655 2 0.887 − 0.655 2

n = 40 participants for each representing group

With the 10% of dropout rate, the minimum sample size is 88 participants. Initially, we collected 99 participants. However, 10 participants were loss to follow up due to some reasons. In the end, there were only 89 participants who completed the study.

Outcome variable

The outcome variable was PSCI developed 3 months after stroke onset. We assessed the cognitive function at day 5, 30, and 90 after stroke onset by using Montreal Cognitive Assessment-Indonesian version (MoCA-INA) which has been validated for Indonesian population. ¹⁹

The longer follow-up is needed to identify PSCI identification because cognitive functioning after stroke may have been fluctuated due to the change of cerebral hemodynamics. However, prior studies have demonstrated that the definition of PSCI is any severity of cognitive impairment, regardless of cause, noted after an overt stroke, and it occurs from 3-6 months after stroke. ^{20 – 23} Therefore, in the present study we identified PSCI at 90 days (3 months) after stroke.

PSCI group was defined if the total score of MoCA-INA at 90 days (3 months) after stroke was < 26. ²⁴ Meanwhile, a non-PSCI group was defined for those with MoCA-INA score ≥ 26. The follow-up measurements of cognitive function at day 30 and 90 were conducted by home visit.

Demographic and clinical characteristics

Demographic factors consisted of age (> 60 and ≤ 60 years old) and sex (male or female). Clinical characteristics consisted of body mass index (BMI) (obese or non-obese), hypertension (yes or no), systolic blood pressure (SBP), diastolic blood pressure (DBP), diabetes (yes or no), HbA1C, dyslipidemia (yes or no), lipid profile (the levels of triglycerides, HDL, LDL, and total cholesterol), as well as serum BDNF level at day 5 and 30 after stroke onset.

We classified the BMI based on WHO criteria for Asian population ^{25 , 26} and further categorized the subjects into obese and non-obese (consisted of those with overweight and normal weight; there were no subjects with underweight). Systolic and diastolic blood pressure were measured by using sphygmomanometer performed at admission. We collected serum BDNF level at day 5 based on prior study which revealed that BDNF was reduced at acute stroke. ²⁷ We also collected serum BDNF at day 30 based on another study reported that BDNF level in acute stroke would be increased steadily and reached the highest level after 30 days. ²⁸

Measurement of serum BDNF level

We took the samples from patients’ serum on days 5 and 30 following the onset of the stroke. Between 7:00 and 9:00 a.m., a serum separator tube was filled with 5 mL venous whole blood from the patients. Patients were also told not to eat or drink anything for 10 hours before getting their blood drawn. Within 60 minutes of blood sampling, the venous whole blood was allowed to coagulate for 30 minutes at room temperature before being centrifuged at 1,000 × g for 15 minutes at 40°C to separate the serum and blood clot. The serum was extracted and stored at −20°C in a separate tube. The serum sample was transported to Prodia Laboratory Jakarta in a frozen state, sealed in a heat-insulated container with dry ice. The sample was labelled by the participant’s initial name and their order number. The laboratory staffs were kept unaware of the participants’ group assignment.

We measure the level of serum BDNF by an enzyme-linked immunosorbent assay (ELISA) using Human BDNF Immunoassay (Quantikine ^® R&D systems, Minneapolis, USA, Cat No: DBD00). The assay's minimal detection limit was 20 pg/ml, and its intra- and inter-assay coefficients of variation were both less than 10%. A monoclonal antibody specific for human BDNF was pre-coated in a 96 well polystyrene microplate (BDNF, R&D System, RD1S) at a concentration of 100L in a buffered protein base with preservative and then lyophilized. The test samples were applied in duplicate (50 L/well). In duplicate wells, a standard curve was created using BDNF (R&D systems, RD1S) at concentrations of 0, 62.5, 125, 250, 500, 1000, 2000, and 4000 pg/mL. In each well, a standard, control, or sample was added and incubated for 2 hours before adding the BDNF conjugate.

Another hour was spent incubating the mixture. Each well was aspired and washed once, then twice more for a total of three washes. Washing buffer (400 L) was used to wash each well with a squirt bottle, multi-channel pipette, manifold dispenser, or autowasher. Each well was filled with 200 mL of substrate solution and incubated at room temperature for 30 minutes, protected from light. Each well received 50 mL of Stop Solution. We lightly tapped the plate to ensure thorough mixing if the color change did not appear uniform. Using a microplate reader tuned to 450 nm, the optical density of each well was calculated in under 30 minutes.

We used Kolmogorov-Smirnov test to determine the normality of numerical data; the independent t-test, Mann-Whitney, and Chi-square test for analyzing the statistical differences between variables. We performed a receiver operating characteristics (ROC) curve analysis by using Youden Index to determine the cut-off point of BDNF level. Bivariate analysis was conducted to analyze the relative risk of associated factors (including BDNF) between PSCI and non-PSCI group. We further performed multivariate analysis by using logistic regression to independently measure the effect (Odds ratio) of serum BDNF level on PSCI with some prediction models after controlling the covariates. In Model I: we controlled the covariates such as age and sex. Model II: Model I plus controlling BMI, hypertension, diabetes mellitus, and dyslipidemia. Model III: Model II plus controlling some variables such as systolic and diastolic blood pressure, level of triglyceride, HDL, LDL, total cholesterol, and HbA1C. All of the analyses were performed using the SPSS software version 25.0 (IBM Co. Ltd, NY, USA). A p value of < 0.05 indicated statistical significance.

This study has received ethical approval from the Medical and Health Research Ethics Committee of the Faculty of Medicine, Universitas Gadjah Mada, Indonesia (EC No. KE/FK/0682/EC/2020). All participants were provided the information regarding the study and signed the informed consent form.