It is apocryphal that the psychiatric uses of lithium were discovered due to high levels of the element in the drinking water of cities with low crime and mental illness rates, though healing waters from wells and natural baths in many countries may have contained high levels of lithium.[1],[2] The real credit for discovering the therapeutic value of lithium for mania goes to John HJ Cade, MD, another observant Australian (see Shoulders of Giants: Standing Up to the Crowd 20.12.20).[3]
An early but poorly tolerated treatment for gout, lithium was sold as a patent medicine for a wide variety of ailments in the late nineteenth century. This unregulated use resulted in many toxicities, cardiac among them. While searching for toxins that patients with manic psychosis might excrete, Dr. Cade detected a protective effect of uric acid in his animal model, with lithium providing the most solvent salt form. Hypothesizing that the lithium itself provided the value, he also tested lithium carbonate (the most common form used today to treat bipolar disorder). This was successful and Dr. Cade also observed a two-hour period of lethargy in his guinea pigs before they fully recovered.
In his outstanding article, Dr. Cade admits it may seem a stretch to jump from lethargic guinea pigs to treating patients with mania, but again, he already had that patient population in mind. It is valuable to review his entire 1949 article as it is another lesson in cogent thought and cautious research design in which he clearly delineates the steps necessary to disprove his hypothesis and the challenges of studying a self-limited and uncommon disorder. He provides what today we call an open label study describing ten prospective case reports: both the treating doctors and patients knew what treatment was given and the cases were considered separately. This leaves the results open to confirmation bias (seeing what you want to believe) but that is routinely accepted in these first research steps, as long as double-blind studies follow.
This initial report led to decades of effort to determine how lithium could be used effectively and safely while avoiding toxicity. As hospital and office prescription of lithium became feasible and more common by the 1980s, concerns developed about teratogenicity (birth defects in the offspring of treated patients). The first assessments of the effect of lithium on fetuses relied upon the Register of Lithium Babies, which started in Scandinavia and was later supplemented by data from other countries.
The previous post in this series contrasted prospective and retrospective studies, with emphasis on epidemiology (see Shoulders of Giants: Looking Forward or Backwards? 21.01.01). Most of the examples referenced utilized either case-control or cohort methods. Patient registries, though, are another form of data collection often used to gather information about risk and response to illness or treatment. Often, these are prospective, meaning that volunteers join the list and then future data is collected about them.
The Register of Lithium Babies, however, was a retrospective study: doctors were asked to submit cases of lithium exposure during pregnancy only after the babies were delivered. This is a design flaw because the data collection is not truly random, resulting in sampling error. This type of bias usually results in the estimate of risk being exaggerated, as people are more likely to report noticeable problems than no problems. In a prospective study, the data is available to be counted whether there is a problem or not.
Nevertheless, in this early retrospective registry the frequencies of any abnormalities reported in the infants were compared to expected incidences in the general population. By the late 1970s, this resulted in the assessment that a fetus exposed to at least one prescription of lithium during the first trimester carried a 3% chance of developing Ebstein’s anomaly, a specific heart defect. This represented an increased relative risk by a factor of 400 over the general population.[4],[5]
It was not until almost two decades later that data from controlled studies and prospective registries demonstrated a much lower relative risk (risk ratio) of this abnormality, ranging from 1.2 to 7.7. This is significantly lower than 400 and illustrates the bias that retrospective reports may carry.[6] This estimate has been replicated at a 2.6 adjusted risk ratio[7] in a recent study actually using retrospective case review, but this time it was a review of all pregnancies resulting in a live birth that were paid for by Medicaid during a 10-year period.[8] Unlike the Register of Lithium Babies, it did not rely upon random report gathering – every case was counted.
This change in risk ratio from 400 to an adjusted risk ratio less than 3 based on the methodology of data collection offers another important illustration of how research design can alter important conclusions. Methods of statistical analysis to be employed on the data gathered should also be determined at the time of research design and study registration, which is becoming mandatory. Due to Dr. Cade’s adroit thinking and Dr. Cohen and colleagues’ alertness to flaws in early methodology, people may now consider an effective treatment with a more accurate assessment of risk.
[1] Parker WF, Gorges RJ, Gao YN, et al: Association between groundwater lithium and the diagnosis of bipolar disorder and dementia in the United States. JAMA Psychiatry 75(7):751–754, 2018 [2] Henderson DK, Gillespie RD. Textbook of Psychiatry for Students and Practitioners, Sixth Edition, pg. 3. Oxford University Press (Oxford, England), 1944 [3] Cade JFJ. Lithium Salts in the Treatment of Psychotic Excitement. The Medical Journal of Australia. No 10, Vol 11, Sept 3, 1949 [4] Schou M, Goldfield MD, Weinstein MR, et al: Lithium and pregnancy. I. Report from the Register of Lithium Babies. BMJ 2(5859):135–136, 1973 [5] Weinstein MR, Goldfield M: Cardiovascular malformations with lithium use during pregnancy. Am J Psychiatry 132(5):529–531, 1975 [6] Cohen LS, Friedman JM, Jefferson JW, et al: A reevaluation of risk of in utero exposure to lithium. JAMA 271(2):146–150, 1994 [7] Adjusted risk ratio attempts to control for additional confounding variables (outside causes) [8] Patorno E, Huybrechts KF, Bateman BT, et al: Lithium use in pregnancy and the risk of cardiac malformations. N Engl J Med 376(23):2245–2254, 2017