Back

Benzodiazepine Withdrawal Symptom Clusters: Distinct Phenotypes with Treatment Implications

Madhava, V. S.

2025-10-09 psychiatry and clinical psychology
10.1101/2025.10.07.25336923 medRxiv
Show abstract

BackgroundBenzodiazepine withdrawal is highly heterogeneous, ranging from mild, transient symptoms to a severe, protracted syndrome. Existing descriptive approaches catalog symptoms without explaining the underlying biology, which often makes the tapering schedule the sole focus of management--an approach that provides only partial relief. MethodsBaseline data from 39 patients in a specialty taper program were analyzed. To assign phenotypes, symptoms reported in a comprehensive 233-item questionnaire were mapped onto five pre-specified mechanistic axes reflecting neurobiological and immune pathways. ResultsMost patients (79.5%) clustered into one of three phenotypes: corticotropin-releasing hormone (CRH)-heavy (35.9%), excitatory-neuroinflammatory (ENI) (33.3%), or Autonomic (10.3%). The remaining patients were classified as Low-symptom (17.9%) or Mixed (2.6%); no patients were classified as Basal Ganglia-Cerebellar (BG/Cer). Features consistent with mast cell activation (MCAS-overlap) were observed in just over half of the cohort (53.8%). PROMIS-29 scores demonstrated severe, multi-domain functional impairment, with notably elevated T-scores in Anxiety and Fatigue. ConclusionBenzodiazepine withdrawal is not a unitary syndrome but clusters into reproducible, biologically grounded phenotypes. Notably, nearly 80% of patients in this cohort fell into just three of these phenotypes, suggesting that withdrawal, though heterogeneous, may follow structured and predictable patterns. Recognizing these patterns enables clinicians to anticipate vulnerabilities and personalize management, shifting the therapeutic goal from merely executing a taper to stabilizing the underlying neurobiology.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.