*3.1. Participants*

Between 11 January 2017 and 12 April 2018, 150 children and adolescents (mean age 11.8 ± 4.1 years, median 11.25, range 5.1–20.8; 80% boys) entered the trial. The ASD symptoms were 'severe' in 78.7% per ADOS-2 (comparison score = 8–10) [31] and the adaptive levels were 'low' (composite score ≤ 70) in 88%, as per the Vineland Behavior Scales [32].

The participant's characteristics are provided in Table 1. Fifty participants were randomly assigned to each of the three treatments in Period 1 and 44 participants per group completed the study (Figure 2).

Among the 150 participants who underwent randomization, 131 (87%) submitted valid questionnaires at the onset and end of the first treatment period (Figure 2), enabling a between-subject analysis in this period (i.e., to compare the change in sleep parameters between the participants who received cannabinoids and the participants who received the placebo). In total, 107 participants (71%) submitted valid questionnaires at the onset and the end of both the first and second treatment period, enabling a within-subject analysis (i.e., to compare the change in sleep parameters while receiving cannabinoids, while receiving the placebo, and in participants who received both treatments).

The participants' baseline characteristics, including sleep disturbances, as indicated by the CSHQ total and sub scores, were similar in the three treatment arms (Table 1).

Overall, 18 participants (12%) withdrew from the trial for the following reasons: 13 for reasons unrelated to treatment, three due to adverse events, and two due to ineffectiveness. In total, 131 participants (87%) had valid CSHQ scores before and after the treatment in the first treatment period. In total, 107 participants (71%) had valid pre-and post-treatment scores in both treatment periods, allowing a within-subject comparison.

#### *3.2. Baseline Sleep Disturbances*

Among the 146 participants who had valid CSHQ scores at the baseline, 125 (86%) had a CSHQ total score ≥41, indicating a sleep disorder. Higher CSHQ scores (indicating more prominent sleep disorder symptoms) at the baseline were correlated with a younger age (Pearson correlation r = −0.288, *p* < 0.001) and with higher SRS total scores, indicating more severe core autistic traits (r = 0.175, *p* = 0.036). The CSHQ scores were not associated with sex or adaptive behavior, as indicated by the VABS composite scores.

Notably, the baseline characteristics were not different between the participants included in the per-protocol analysis and the participants who were excluded due to withdrawal or missing data, including age (*p* = 0.83); sex (*p* = 0.86); the severity of sleep disorders, as reflected by the CSHQ total score (*p* = 0.63); adaptive behavior, as evaluated by the VABS Composite scores (*p* = 0.57); and the severity of the core autistic symptoms, as assessed by the ADOS-2 (*p* = 0.58), CARS (*p* = 0.75), and the SRS (*p* = 0.25).

#### *3.3. Impact of Cannabinoid Treatment on Sleep*

The impact of the cannabinoid treatment on sleep disturbances was assessed using the CSHQ. In total, 131 participants had valid CSHQ scores, both pre-treatment and post-treatment, in the first 12-week treatment period. Among these 131 participants, 44 received a whole-plant extract (BOL-DP-O-01-W, CBD:THC ratio = 20:1), 42 received pure cannabinoids (BOL-DP-O-01, CBD, and THC at a 20:1 ratio), and 45 received a placebo. The CSHQ total scores and the subscale scores did not differ significantly between the participants who received cannabinoids and the participants who received the placebo

(Table 2). None of these measures differed significantly between the participants who received the whole-plant extract versus the pure cannabinoids (Table 2).


**Table 2.** Impact of cannabinoid treatment on sleep. Comparison of treatment effects in the 1st 12-week period.

Between-subject analyses of the change in the CSHQ scores following treatment in the first treatment period. CSHQ—Children's Sleep Habits Questionnaire. Positive change (increment of CSHQ scores) indicates worsening of the sleep disorder. Change in the CSHQ scores from baseline following treatment is compared between the 3 treatment arms. ˆ One-way ANOVA for influence of treatments between study groups. Notably, the difference between cannabinoid treatment and placebo was not statistically significant, even when combining the two cannabinoid treatments into one group, compared to placebo (data not shown).

Similar negative results were found in the second treatment period (Table S2) and when comparing the two treatments that each participant received, using a within-participant analysis (Table S3).

#### *3.4. Longitudinal Associations between Sleep, Behavior, and Autistic Core Symptoms*

Regardless of the treatment, improvements in the sleep disturbances, as indicated by a decline in the CSHQ total score, were associated with improvements in the autistic core symptoms, as well as the associated disruptive behaviors in both treatment periods.

The autistic core symptoms were assessed using the SRS total score (higher scores indicate higher severity of symptoms). Changes in the SRS total score correlated with changes in the CSHQ total score in Period 1 (Pearson correlation: r = 0.266, *p* = 0.008) and Period 2 (r = 0.309, *p* = 0.004).

Improvements in the ASD-associated disruptive behaviors were evaluated by the Clinical Global Impression–Improvement rate (CGI-I: lower rates indicate improvement). The CGI-I rate was associated with a change in the CSHQ total score in Period 1 (one-way ANOVA: f = 4.5, *p* = 0.013) and Period 2 (f = 3.36, *p* = 0.038).
