Strong beliefs may make us comfortable, but they are not facts. Passion does not automatically turn opinion into truth. In science, we can test hypotheses and we can try models, but a model is not a hypothesis. There are plenty of examples of situations where a model is more explanatory than a hypothesis, especially when addressing complex questions that do not have a simple, or a single answer. While hypotheses may be right or wrong, models are better or worse. When trying to understand the nuances of cognitive impairment in bipolar disorder (Samamé, Reference Samamé2024), debating who is wrong, (those who believe in a neurodevelopmental cause, or those who think that it is driven by neuroprogression), leads to a black and white approach that will not solve the question, simply because this is not the question. Neuroprogression and neurodevelopment are not hypotheses, they are models, with their pros and cons. Neither model is proven right or wrong, and depending on the context, one may fit better than the other. There is also a crucial concept in science which is bias. Bias is particularly relevant in epidemiology and especially in observational studies. It is not the same to study the trajectory of certain variables in a sample of patients from a private outpatient clinic than from a representative sample of patients from a given catchment area, and one has always to be mindful of attrition in long-term studies.

Cognitive impairment in bipolar disorder is hugely heterogeneous (Burdick & Millett, Reference Burdick and Millett2021). Cluster analysis from cross-sectional studies shows that a subset of patients may show persistent cognitive deficits, while others are apparently cognitively intact; there is also a third group with selective impairments (Lima et al., Reference Lima, Rabelo-da-Ponte, Bücker, Czepielewski, Hasse-Sousa, Telesca and Rosa2019). Two decades ago, two independent studies reported the impact of manic episodes on cognitive impairment (Cavanagh, Van Beck, Muir, & Blackwood, Reference Cavanagh, Van Beck, Muir and Blackwood2002; Martínez-Arán et al., Reference Martínez-Arán, Vieta, Reinares, Colom, Torrent, Sánchez-Moreno and Salamero2004). Although those were cross-sectional studies, similar cognitive dysfunctions across bipolar mood states, including euthymia, were reported, as well as a correlation between the severity of cognitive impairment and poor functioning. This finding has been replicated several times, the last of which was a large, multicentre study with over 5800 patients, that showed an association with the number of episodes (Burdick et al., Reference Burdick, Millett, Yocum, Altimus, Andreassen, Aubin and McInnis2022), of which the manic ones seem to be the most impairing (Sanchez-Moreno et al., Reference Sanchez-Moreno, Bonnin, González-Pinto, Amann, Solé, Balanzá-Martinez and Vieta2018).

The main argument against a neuroprogression model, as argued by Samamé (Reference Samamé2024), is the fact that most longitudinal studies do not show cognitive impairment to progress over time. The problem, though, is that most longitudinal studies are very short, and they suffer from selection bias. The main bias is caused by the fact that the most severe and difficult-to-treat patients tend to be lost to follow-up or die. Hence, those patients may be less treatment adherent (Samalin & Belzeaux, Reference Samalin and Belzeaux2023) and relapse, and their cognition and functioning may decrease over time, in line with the concept of neuroprograssion. It is not a matter of the type of setting, but the ‘survivor’ effect, as described by Montejo et al. (Reference Montejo, Torrent, Jiménez, Martínez-Arán, Blumberg, Burdick and Vieta2022) in elderly population. In fact, the best longitudinal studies are those that start from premorbid states (supporting indeed a staging model), showing that subjects developing bipolar disorder were, on average, actually even smarter than those who remained healthy (Cannon et al., Reference Cannon, Jones, Gilvarry, Rifkin, McKenzie, Foerster and Murray1997; Zammit et al., Reference Zammit, Allebeck, David, Dalman, Hemmingsson, Lundberg and Lewis2004). If the majority of people with bipolar disorder do not show any premorbid cognitive deficits, and the majority of bipolar patients are cognitively impaired in clinical samples, it seems obvious that something happened in between. But of course, due to heterogeneity, it is likely that a relatively small subset of patients with bipolar disorder may actually have a neurodevelopmental condition (Kjærstad, Søhol, Vinberg, Kessing, & Miskowiak, Reference Kjærstad, Søhol, Vinberg, Kessing and Miskowiak2023); those are more likely to have an early onset of disease, and comorbidity with ADHD and other neurodevelopmental problems. One could think of these patients as neurodevelopmental phenocopies of bipolar patients with a neuroprogressive trajectory. Moreover, there is now evidence that not only cognitive dysfunction tends to worsen over time in the most severe and recurring patients, but structural brain changes such as frontocortical thinning are also associated with worse course outcomes and the number of manic episodes (Abé et al., Reference Abé, Ching, Liberg, Lebedev, Agartz, Akudjedu and Landén2022). Importantly, this is a longitudinal study, and the authors conclude that their results yield insights into disease progression in bipolar disorder and highlight the importance of mania prevention in bipolar disorder treatment. While it is true that the cluster analysis studies identified a subset of patients who were apparently cognitively intact, this term may be misleading, because it refers to those patients as having a cognitive performance similar to healthy controls, but in the context of cross-sectional studies, we do not know if those patients worsened with respect to their baseline performance before illness onset. Here, the concept of cognitive reserve comes into play (Sánchez-Torres et al., Reference Sánchez-Torres, Amoretti, Enguita-Germán, Mezquida, Moreno-Izco, Panadero-Gómez and Cuesta2023).

It is unfortunate that currently there is no pharmacological treatment for cognitive dysfunction in the context of bipolar disorder, schizophrenia, or depression (Miskowiak et al., Reference Miskowiak, Seeberg, Jensen, Balanzá-Martínez, Del Mar Bonnin, Bowie and Vieta2022). In fact, medication is a confounder as many of the drugs used to treat bipolar illness can actually harm cognition (Ilzarbe & Vieta, Reference Ilzarbe and Vieta2023; Vidal et al., Reference Vidal, Brunet-Gouet, Frileux, Aouizerate, Aubin, Belzeaux and Roux2023). The same applies to comorbidity (Miguel, Marquez-Arrico, Jodar, Navarro, & Adan, Reference Miguel, Marquez-Arrico, Jodar, Navarro and Adan2023). There are, however, effective treatments based on psychological interventions. Some meta-analyses mistakenly pool cognitive remediation trials with functional remediation ones, causing again heterogeneity and mixing apples and oranges (Samamé, Durante, Cattaneo, Aprahamian, & Strejilevich, Reference Samamé, Durante, Cattaneo, Aprahamian and Strejilevich2023). Cognitive remediation improves cognition (Lewandowski et al., Reference Lewandowski, Sperry, Cohen, Norris, Fitzmaurice, Ongur and Keshavan2017), and functional remediation improves functioning (Torrent et al., Reference Torrent, Bonnin, Martínez-Arán, Valle, Amann, González-Pinto and Vieta2013). Those trials were positive on their primary outcomes, even if they failed on some secondary outcomes.

In summary, in my opinion, there is enough information to say that cognitive impairment in bipolar disorder mostly fits into a neuroprogression model, although there is some heterogeneity and a subset of patients may better fit into a neurodevelopmental model, with premorbid deficits, early age of onset and complex comorbidities, and a course closer to schizoaffective disorder or schizophrenia. It may make sense to try to adjust staging models to this reality and try to tailor effective prevention and treatment strategies accordingly, aiming at personalized, precision psychiatry (Vieta, Reference Vieta2015). If bipolar disorder is mostly neuroprogressive, albeit heterogeneous, treatment and prevention programs should focus on effective prevention of relapse, early diagnosis and treatment, and enhancement of cognitive reserve and resilience.