Skip to main content
. 2026 Apr 1;17:1782968. doi: 10.3389/fphar.2026.1782968

TABLE 5.

Top drugs associated with high-risk PARS and low-risk PARS. Coefficients are from the logistic regression model; odds ratios are reported with 95% confidence intervals (patient-level bootstrap). Visits denote the number of cohort visits containing the drug. Few drugs met the inclusion criterion (Visits ≥20 ) with negative coefficients; therefore all eligible low-risk drugs are shown.

High-risk PARS drugs (coef > 0)
Drug name Coefficient Odds ratio (95% CI) Visits
Quetiapine 3.299 27.075 (16.856–43.183) 459
Haloperidol 2.850 17.290 (4.913–37.712) 67
Entacapone 2.661 14.304 (1.699–26.758) 37
Amiodarone 2.362 10.611 (1.028–20.445) 24
Trazodone 1.998 7.372 (3.937–11.176) 423
Venlafaxine 1.977 7.221 (3.229–11.170) 216
Risperidone 1.934 6.918 (3.008–11.406) 276
Tramadol 1.930 6.887 (3.133–10.829) 23
Clozapine 1.830 6.235 (1.493–16.009) 108
Trihexyphenidyl 1.731 5.645 (2.240–17.157) 130
Formoterol 1.711 5.536 (2.540–11.290) 73
Sertraline 1.705 5.502 (1.634–9.845) 92
Olanzapine 1.701 5.480 (2.435–9.581) 155
Bisoprolol 1.619 5.047 (2.655–10.827) 297
Rivaroxaban 1.616 5.033 (0.833–11.379) 21
Nebivolol 1.602 4.963 (2.810–8.783) 205
Escitalopram 1.594 4.925 (2.230–7.370) 272
Tiapride 1.594 4.924 (1.836–8.862) 74
Flupentixol 1.544 4.682 (0.714–10.908) 58
Acetaminophen 1.526 4.602 (2.206–7.056) 21
Low-risk PARS drugs (coef < 0) a
Drug name Coefficient Odds ratio (95% CI) Visits
Ivabradine − 0.830 0.436 (0.250–1.324) 39
Piracetam − 0.353 0.702 (0.247–2.002) 52
Amisulpride − 0.347 0.707 (0.270–1.983) 52
Sitagliptin − 0.263 0.769 (0.476–1.487) 34
Ramipril − 0.255 0.775 (0.280–2.753) 46
Nitroglycerin − 0.254 0.776 (0.369–1.602) 44
Rilmenidine − 0.245 0.783 (0.326–2.553) 67
Doxepin − 0.184 0.832 (0.573–1.000) 32
Alfacalcidol − 0.109 0.896 (0.416–2.532) 53
Trimetazidine − 0.057 0.944 (0.436–2.118) 127
Perindopril − 0.019 0.981 (0.518–2.017) 316
a

Negative coefficients are reported for interpretability and transparency and do not imply protective effects or suggest medication changes; the actionable output is the model’s predicted High-risk probability.