Discriminating evidence – Use and misuse of the drug discrimination test in abuse potential assessments of novel CNS drugs

Abstract

Drug-discrimination is a mandatory test for evaluating abuse potential of CNS drug‑candidates. The major challenge is designing translationally valid protocols for drug-candidates with novel pharmacological mechanisms and no affinity for targets mediating the psychoactive effects of known substances of abuse. We assessed predictive validity of 3 experimental protocols using our and other published data. Experimental protocols: (1) Drug‑candidate tested in animals trained to discriminate a selected abused substance from vehicle, (2) Drug-candidate investigated in multiple studies with animals trained to recognize representatives from major abused drug classes; (3) Drug-candidate used as training cue with back-testing of various abused substances. Female rats were trained to discriminate d-amphetamine (0.25–0.5 mg/kg,ip) from saline. Full, Partial, and No Generalization: ≥75 %, ≥60 % and < 25 % drug‑associated lever-presses. Microdialysis measurements of extracellular dopamine in accumbens (ACB) or striatum (STR) were made in freely‑moving rats. Generalization to d-amphetamine (mg/kg[route]): lisdexamfetamine (3.4[po]); phentermine (3.0[po]); methylphenidate (5.0[po]); bupropion (30[po]); methamphetamine (0.5[ip]); fencamfamine (3.0[ip]); cocaine (10[ip]). Partial generalization <60 %: sibutramine (3.0[ip]); modafinil (100[ip]). No generalization: atomoxetine (10[po]); desipramine (5.0[ip]); venlafaxine (10[ip]). Only drugs markedly increasing dopamine neurotransmission in ACC/STR generalized to d‑amphetamine. Drugs producing small/moderate dopamine increases in ACB/STR and/or increased dopamine in prefrontal cortex did not produce full or ≥ 60 % generalization to d-amphetamine. Dopamine D₂ antagonists blocked d‑amphetamine's discriminative cue and its reinforcing effect in humans [1,2] showing the pharmacological specificity of both effects. Drug-discrimination findings with other substances of abuse, e.g. cannabinoids, psychedelics, GABA-A positive allosteric modulators (PAMs), glutamatergic ligands, confirm the hypothesis. Results obtained using the drug‑candidate as training cue showed this experimental approach is not viable [3,4]. Evaluation of approaches 1–3: (1) Has weaknesses, but highest predictive validity. (2) Invalid, except when drug-candidate has pharmacology shared with selected abused substances. Behavioral similarity does not produce generalization or have predictive validity. (3) Invalid. No evidence that generalization to drug-candidate would generate any meaningful information or predict abuse potential.

[1] Lile et al., 2005; [2] Rush et al., 2003; [3] Swedberg et al., 2014; [4] Swedberg & Raboisson, 2014.