Most trials reviewed in this study relied on spontaneous reporting of DICI by participants, relatives, or caregivers instead of actively assessing cognitive safety. However, DICI is not synonymous with manifest dementia. It may remain mild yet measurable with appropriate neuropsychological tests. Drug-induced cognitive impairment is not always recognized by patients or their relatives and even if cognitive impairment is noticed, it may be interpreted as symptoms of disorders such as depression or Alzheimer’s disease [23]. This implies a considerable risk of non-detection and, therefore, under-reporting in clinical trials. The risk of under-reporting is further increased as patients wishing to receive the experimental treatment may be afraid of being excluded from the study if they report adverse drug reactions. Hence, it is not sufficient to rely on self-reporting of DICI.

There are obvious reasons why DICI should be recognized and evaluated in clinical trials. For one, DICI accelerates cognitive aging [15, 24,25,26]. It may lead to a new diagnosis in addition to the patient’s original condition, thereby increasing the number of drugs prescribed to this patient, which in turn may increase the risk of DICI. Second, DICI is not always so severe that it is noticed by mere observation.

However, even subtle changes in cognition and behavior may have a significant impact on academic achievement, work, or social functioning. Such subtle changes are rarely detected by using crude screening tools such as the Mini-Mental State Examination (MMSE) or the Montreal Cognitive Assessment [27]. Screening tools or questionnaires such as those developed by the European Organisation for the Research and Treatment of Cancer have low sensitivity for detecting non-serious cognitive adverse effects [27,28,29,30]. For example, only three out of the maximum achievable 30 points in the MMSE account for learning and long-term retrieval. This implies that if a person with apparent memory problems has little difficulty with the remaining items in the MMSE, they may achieve a total score of 27 points. This would be interpreted as normal cognitive function. A deterioration in memory function in an otherwise highly functioning, intelligent and/or younger person may thus remain undetected.

This is further illustrated by the fact that several trials that actively assessed cognitive safety did not detect DICI even when the trial drug’s potential to induce them was declared in its prescribing information (see Table 1 of the ESM). This may have been due to coincidence or the study design but in many cases, it appears more likely to be a consequence of using questionnaires, tests with low sensitivity (e.g., the MMSE), or tests that are not suitable to detect DICI. For example, tests focusing on behavior (e.g., Behavior Rating Inventory of Executive Function Preschool Version) or intelligence (e.g., Wechsler Adult Intelligence Scale, Third Edition) are poorly suited to detect impaired cognitive function. By contrast, some trials did report cognitive adverse events with frequencies >5% on www.clinicaltrials.gov, but these are neither reported in the formal publication nor in the respective prescribing information. Examples of the latter are NCT01982955 (tefotinib) and NCT02308020 (abemaciclib).

Seven of the 15 trials that used neuropsychological tests reported cognitive adverse events. Notably, two of the four trials with esketamine and one trial with an anti-seizure medication (lacosamide) did not. Their prescribing information declares cognitive adverse effects. However, the lacosamide trial used the Behavior Rating Inventory of Executive Function Preschool Version. The esketamine trials used a non-specified test battery that possibly was inappropriately composed.

It may be argued that it is not imperative to evaluate the cognitive safety of a drug already approved and clinically used for years, perhaps even decades. This is not a valid argument because, not least according to the findings in the present study, the cognitive safety of most drugs has very likely never been appropriately examined.

It might be reasonable to expect that new unapproved drugs or drugs marketed for no more than a year would be more thoroughly assessed regarding cognitive safety than older drugs. Our results do not support this notion, even if it is well known that previously unreported adverse drug reactions may keep emerging for years after a drug’s first approval [31].

Given the absence of appropriate assessment of cognitive safety in clinical trials and the focus on manifest dementia in post-marketing DICI research, it is not surprising that meta-analyses reviewing the evidence for DICI of specific drugs find inconclusive, weak, or no such evidence at all [32, 33]. As a side note, most trials in this study used a public reporting threshold of 5% for non-serious adverse events, meaning that even if cognitive adverse events should occur in as many as 49 out of 1000 patients, they would not be publicly reported.

Some publications stated that the severity grading of adverse events was performed by the investigator, and only reported “serious or severe common adverse events”. An example is the shared benefit-risk assessment article of the esketamine trials (see Table 1 of the ESM) [34]. This publication does not report any cognitive adverse events at all even if a confusional state and mental impairment were observed in 5.9% and 5.2% in two of the trials (according to clinicaltrials.gov). Instead, it states: “There was also no difference between treatment groups in any of the cognitive tests performed during TRANSFORM‐1 or TRANSFORM‐2”. This statement seems to be in contrast to the reported adverse events of NCT02417064 (TRANSFORM-1) on www.clinicaltrials.gov where mental impairment was observed in 5.22% of the 56-mg 2 × /week esketamine arm and in 0.88% of the placebo arm. Mental impairment is used here as a Medical Dictionary for Regulatory Activities term, meaning that it is a subclass of Cognitive and attention disorders [35].

Most of the observed cognitive adverse events are not trivial (see Table 1 of the ESM). For instance, aphasia, amnesia, and memory impairment can have a devastating impact on a person’s professional and social life. If not recognized as DICI, they may be misinterpreted as symptoms of an underlying disease and the patient may be given a new diagnosis as well as, presumably, additional drug treatment [20, 36].

Decades ago, it was found that DICI is the cause of 10% of dementia cases [1, 2]. In the absence of more recent epidemiological data, the current prevalence of drug-induced dementia and other forms of DICI is unknown. Cognitive impairment and dementia can have enormous consequences for those who are affected, and for their families. In addition, they strain public health systems and, through sick leave or early retirement, deprive enterprises and national economies of individuals with professional knowledge and experience [37, 38]. Regulatory agencies should demand active cognitive monitoring in all clinical drug trials. We recommend that each cognitive domain be assessed. Ideally, the tests should be selected by an experienced neuropsychologist according to the type of the expected possible cognitive effects. This should be mandatory for drugs known to, intended to, or suspected to affect the CNS, and desirable for all other drugs. The minimum should be a cognitive screening tool of the more domain-specific type, selected according to the study population, disease, and drug studied, such as EpiTrack®, the Rowland Universal Dementia Assessment Scale, or the Mini-Cog®.