Lithium Studies Revisited

I found studies on diagnostic accuracy for bipolar disorder: Carvalho et al. (2015, Journal of Affective Disorders): Sensitivity 69–72 %, Specificity 74–86 % (screening tools). Mitchell et al. (2008, The Lancet): MDQ, Sensitivity 66 %, Specificity 79 % (clinical settings). Zimmerman et al. (2011, Journal of Clinical Psychiatry): Sensitivity 40–60 % (clinical practice). Hirschfeld et al. (2003, American Journal of Psychiatry): MDQ, Sensitivity 73 %, Specificity 90 % (outpatient psychiatry). Phelps & Ghaemi (2006, Bipolar Disorders): Broad range, 50–80 % depending on setting. These studies oddly omit Positive Predictive Value (PPV) and prevalence – a glaring gap, given how critical these are to assess diagnostic reality. So, I took it upon myself to calculate them, using Bayes’ Theorem with reasonable prevalence assumptions: Today’s general population prevalence: 1 % (e.g., Merikangas et al., 2011). PPV range: 1.53 % (40 %/74 %) to 6.87 % (73 %/90 %). Meaning: A bipolar diagnosis is correct in only 1.53–6.87 % of cases – 93–98 % are false positives, i.e., “healthy” people misdiagnosed. For 1970 (lithium approval, FDA), prevalence likely lower: 0.5 % (estimated from retrospective data, e.g., Lewinsohn et al., 1995). Realistic 1970 values (Sensitivity 60 %, Specificity 80 %): PPV = 1.49 %. Lowest (40 %/74 %): PPV = 0.77 %. Highest (73 %/90 %): PPV = 3.54 %. Meaning: Even worse – 96–99 % false positives in the 1970s. Conclusion:The approval studies (e.g., Cade 1949: 10 patients, Schou 1954: 38, Baastrup & Schou 1967: 88, Baastrup et al. 1970: 84) totaled ~220–250 participants. At PPV 1.49 % (0.5 % prevalence), only 3–4 were likely truly bipolar – 246–247 were “healthy” (false positives). In Baastrup 1970 (84 patients, pivotal for FDA): ~1 real case – 0 relapses (lithium) = sedation of “healthy,” 54 % relapses (placebo) = withdrawal misinterpreted as “mania.” This suggests lithium was almost exclusively tested on “healthy” individuals (e.g., with drug use, sleep deprivation, or personality traits), not bipolar patients. The “relapses” were likely withdrawal symptoms, not disease recurrence – a conclusion bolstered by my PPV estimates. Moncrieff’s Viewpoint:In this talk by Joanna Moncrieff (Link: https://www.youtube.com/watch?v=-lBoyHCIuGE), she describes how withdrawal symptoms (e.g., agitation, insomnia) are often misread as relapse, especially when stopping sedative drugs like antipsychotics. "It's only recently been recognized that when people stop taking psychiatric drugs, they have a withdrawal reaction. The withdrawal reaction occurs for the same reasons that people get a withdrawal reaction from heroin or alcohol. If you take a drug for a long period of time, the body gets used to having it on board, and if you stop it, the body reacts to it not being present anymore. Over the last few decades, it's become apparent that if you stop taking any sort of psychiatric drug, at least some people will get a withdrawal reaction. Different sorts of psychiatric drugs produce different sorts of symptoms when they're stopped, and it depends on the chemical nature of the drug and what effects it has when you're taking it as to what withdrawal symptoms you'll get. When people stop taking antipsychotics, for example, because these drugs are very sedative, people will become agitated, anxious, they might find it difficult to sleep, and there'll be physical symptoms related to the physical effects of the antipsychotics too. If someone who's had a previous psychosis starts to become very agitated, stops sleeping, and maybe becomes angry because the emotional suppression that the drugs cause has been lifted, it may be thought that they're having a relapse. Frequently, withdrawal symptoms are mistaken for a relapse of the underlying disorder because no one thinks about withdrawal symptoms or, at least until recently, people have given them very little consideration." This perfectly aligns with my thesis: the lithium studies likely observed withdrawal effects in misdiagnosed individuals, not true prophylaxis in bipolar patients. Questions: Can you validate my PPV calculations (1.53–6.87 % at 1 %, 0.77–3.54 % at 0.5 %)? What’s your take on the thesis that lithium was tested almost exclusively on “healthy” individuals, with withdrawal symptoms driving the “relapse” narrative? Are there other studies or PPV data I’ve missed that could refine this?