Common outcome measures between the 9 studies containing completely sufficient data studies were extracted (Table 1), with those being scores related to Aberrant Behavior Checklist (ABC) total and subscales (specifically the Lethargy, Hyperactivity, Irritability/Agitation elements, Inappropriate Speech, and Stereotypical Behavior), Clinical Global Improvement severity scale (CGI-S), Conners 3 Parent Rating Scale (specifically the subscales of Hyperactivity and Inattention), and Vineland (specifically the socialization, communication, and maladaptive behavior subscales). Forest plots with each overlapping outcome measure and corresponding effect sizes are included for each of the 9 studies. The direction of the effect size was adjusted to reflect an improvement or decline in each outcome measure.
Table 1
FXS Clinical Trial Studies with Overlapping Outcome Measures Used for Effect Size Calculations.
Study | Drug/Treatment | Population | Participants | Sample Size | Outcome Measures |
De La Torre (Study 6) (20) | Epigallocatechin-3-gallate | Fragile X | Adults | 12 | ABC subscales |
Bruno (Study 8) (21) | Donepezil | Fragile X | Anyone with FXS | 12 | ABC total, ABC subscales |
Hessel (Study 11) (22) | Cogmed | Fragile X | Children and adolescents | 49 | Conners 3 Parent Rating Scale |
Youssef (Study 15) (23) | Basimglurant | Fragile X | Adolescents and Adults | 25 | CGI-S, ABC Total |
Sahu (Study 35) (24) | Donepezil | Fragile X | Boys | 10 | Conners 3 Parent Rating Scale |
Berry-Kravis (Study 86) (25) | CX516 | Fragile X | Adult Males and Females aged 18–50 | 24 | ABC subscales |
Veenstra-VanderWeele(Study 98) (26) | Arbaclofen | Fragile X | Age 5–21 | 69 | ABC subscales, Vineland socialization, CGI-S |
Berry-Kravis (Study 112) (27) | Arbaclofen | Fragile X | Adolescents and adults | 62 | ABC subscales, Vineland socialization, CGI-S |
Torrioli(Study 147) (28) | L-acetylcarnitine | Fragile X | Boys | 24 | Vineland socialization, Conners 3 Parent Rating Scale (Hyperactivity) |
The most commonly used outcome measure was the ABC, which was found to be used in 5/9 studies (Fig. 2A-E), which also reported the numbers required for an effect size calculation. The two Arbaclofen studies interestingly had the largest positive treatment effect sizes on all the ABC subscales. Although the majority of effect sizes reported were not considered large according to Cohen, a large effect size of 0.822 for the ABC lethargy outcome measure was reported for patients undergoing placebo treatment, with a smaller effect size for the actual arbaclofen treatment (0.740). These two arbaclofen studies had a mildly positive effect on all 5 subscales of the ABC (ES Lethargy: 0.740 (study 112), 0.525 (study 98), hyperactivity: 0.545 (112), 0.371 (98), Irritability: 0.339 (112), 0.322 (98), Stereotypical Behaviour: 0.605 (112), 0.233 (98), Inappropriate Speech: 0.456 (112), 0.211 (98)). But there was also a similar mildly positive effect on all 5 subscales for placebo treatment (ES Lethargy: 0.822 (112), 0.687 (98), hyperactivity: 0.454 (112), 0.300 (98), Irritability: 0.411 (112), 0.315 (98), Stereotypical Behaviour: 0.433 (112), 0.362 (98), Inappropriate Speech: 0.529 (112), 0.234 (98)). Despite the mild effect sizes, no significant findings were found between the placebo and treatment groups. The phase 3 arbaclofen study was promising as it suggested improvements for younger patients, despite not seeing improvements in primary endpoints27.
The CX516 study appeared to have small effect sizes for all the ABC subscales25. In the case of epigallocatechin-3-gallate (EP3G), we noted that the drug (treatment group) had the opposite effect on inappropriate speech compared to the rest of the ABC subscales. A mildly positive effect size was reported for EP3G on all 5 subscales of ABC except for inappropriate speech, where a weakly negative effect size was reported (-0.132). The negative effect size reported for the inappropriate speech subscale suggests that these symptoms worsened for the EP3G treatment group (-0.132). However, the rest of the ABC subscales suggest an improvement in symptoms as reflected by positive effect sizes for EP3G.
The placebo group for basimglurant was interesting as it reported the largest positive effect size for the ABC Total outcome measure (0.844). Unexpectedly, the effect size was larger for the placebo group than the treatment groups for both 0.5mg (0.465) and 1.5 mg of basimglurant (0.573). Donepezil reported the smallest effect sizes for the placebo and treatment groups on the ABC Total measure. Because the confidence intervals overlap, no significant findings were found.
Next, we found that the Vineland socialization subscale was used by 3/9 (Fig. 3) studies reporting measures required for ES calculation.
Looking at the Vineland 2 subscales, the 3 studies reported relatively small effect sizes. However, arbaclofen showed some promising results as mildly positive effect sizes were reported on all 3 of the subscales, with the exception of the phase 3 clinical trial on the maladaptive behaviour subscale (where the largest negative effect size of -0.452 was found). However, the negative effect size suggests that arbaclofen tended to worsen maladaptive behaviour. A mildly positive effect size was found for L-acetylcarnitine (0.379) on the socialization subscale, with smaller positive effect sizes being found for the 2 arbaclofen studies. However, because the confidence intervals overlap, no significant findings were found.
Another commonly used scale is the Clinical Global Impression Severity Scale (CGI-S) to report clinical improvements from multiple domains related to the condition. This was used by 3/9 studies (Fig. 4).
Again, the 2 arbaclofen studies are promising as they both reported mildly positive treatment effect sizes (0.634 and 0.469), which were greater than both basimulgrant doses (0.367 and 0.337). However, all of the placebo effect sizes for the 3 studies were very similar (0.316, 0.333 and 0.332). It is important to note that there was a significant difference between the treatment and placebo groups for Arbaclofen phase 2 as determined by an unpaired t-test. However, a phase 3 clinical trial failed to reproduce a significant effect between the placebo and treatment groups27.
Next, we found that the Conners’ Parent Rating Scale was used by 3/9 (Fig. 5) studies reporting measures required for ES calculation.
L-acetylcarnitine showed promising results as the treatment group reported the largest effect size on the Conners 3 Parent Rating subscale for hyperactivity (1.16), representing a reduction in hyperactivity symptoms. However, it is unclear if this large effect size is valid, as the reporting of the quantitative data may have been poor in the study. Unfortunately, the study did not measure inattention. Donepezil also showed promising results, as the treatment group reported a mildly positive effect size for hyperactivity (0.433) but a smaller effect size for inattention (0.0955). It is also interesting that the placebo group produced a larger effect size (0.772) than the treatment group for inattention (0.0955), speaking to the large placebo effect. In addition, a Cogmed intervention reported a small positive effect size for both the treatment and placebo groups for both hyperactivity and inattention.
Studies that did not include a placebo group but contained only a treatment group with overlapping outcome measures were graphed (Table 2). In particular, the ABC subscales (Fig. 6) and CGI-S (Fig. 7) outcome measures were graphed. In these studies without a placebo group, a larger effect size was observed compared to those with placebo groups.
Table 2
FXS clinical trial studies with overlapping outcome measures but without placebo groups used for effect size calculations.
Study | Drug/Treatment | Population | Participants | Sample Size | Outcome Measures |
Heussler (Study 7) (29) | Transdermal cannabidiol | Fragile X | Children | 18 | ABC subscales |
Caku (Study 24) (30) | Lovastatin | Fragile X | Anyone with FXS | 15 | ABC subscales |
Schneider (Study 31) (31) | Minocycline | Fragile X | Children | 5 | CGI-S |
Erickson (Study 34) (32) | Acamprosate | Fragile X | Youth | 12 | ABC subscales |
Erickson (Study 41) (33) | Aripiprazole | Fragile X | Boys | 10 | ABC subscales |
Erickson (Study 42) (34) | Riluzole | Fragile X | Adults | 6 | ABC subscales |
Paribello (Study 44) (35) | Minocycline | Fragile X | Ages 13–32 | 20 | ABC subscales, CGI-S |
Torrioli (Study 45) (36) | Valproic Acid | Fragile X | Boys | 10 | Conners Parent Rating, CGI-S |
Kesler (Study 48) (37) | Donepezil 5mg and 10 mg | Fragile X | Males | 8 | ABC subscales |
Berry-Kravis (Study 49) (38) | Lithium | Fragile X | Ages 6–23 | 15 | ABC subscales, CGI-S |
Erickson (Study 97) (39) | Arbaclofen | Fragile X | Children and adolescents | 32 | ABC subscales, CGI-S |
Biag (Study 106) (40) | Metformin | Fragile X | Children | 9 | ABC subscales |
Erickson (Study 121) (41) | Memantine | Fragile X | Ages 13–22 | 6 | CGI-S |
One interesting finding worth mentioning is that aripiprazole reported large effect sizes for multiple ABC subscales. However, as this study did not contain a placebo group that can be compared to, it is difficult to determine its true efficacy. Future clinical trials should include a placebo group to investigate aripiprazole’s true efficacy.
Another interesting finding is that lovastatin reported a weakly negative effect size for ABC hyperactivity (-0.06), indicating a worsening of hyperactivity symptoms in some subjects.
Another promising finding is that two different minocycline studies reported very large effect sizes for the CGI-S outcome measure (2.83 and 3.60). However, both of these studies did not include a placebo group that could be used for comparison. Given these limitations, further clinical trials that include a placebo group with sufficient data are required.