Cognitive trajectories in schizophrenia
Do they support the neurodevelopmental or neurodegenerative hypothesis?
The disorder now known as schizophrenia, was once called ‘dementia praecox’, meaning premature dementia. It was viewed by Emil Kraepelin as a progressive, degenerating illness with little hope of recovery.
This conceptualisation was challenged in the late 1980s by researchers such as Robin Murray, through the neurodevelopmental hypothesis. It proposes that some risk factors for schizophrenia such as winter birth and obstetric complications could exert an effect on neurodevelopment, making a person more vulnerable. Additionally, people with schizophrenia are more likely to have subtle motor problems and minor physical abnormalities in childhood - suggesting a common neurodevelopmental origin.
My own tenuous link to this topic came when I visited Robin Murray’s department as a medical student more than a decade ago. He was writing a rebuttal of the neurodegenerative model of schizophrenia with a psychiatrist who was visiting from Toronto, Robert Zipursky. They allowed me to assist, and the paper we published, The myth of schizophrenia as a progressive brain disease refutes the neurodegenerative model in three broad domains:
Clinical outcomes from schizophrenia. We argued that many patients experience remission and recovery. The unduly pessimistic prognosis schizophrenia might be a sampling bias in clinical populations, explained by the clinician’s illusion and Berkson’s fallacy.
Progressive brain changes in neuroimaging. Here we acknowledged that there are group-level changes in schizophrenia brain scans compared with healthy controls over time, but these were more likely to relate to confounding factors like antipsychotic medication, drugs, and lifestyle rather than a degenerative process.
Cognitive changes. We made the case that, while cognitive deficits exist in schizophrenia, these are present before, or around, illness onset and do not progress over time.
How do the arguments stand-up ten years later? In a previous post, I stated that the case for any neuroimaging abnormalities (never mind progressive changes) in schizophrenia are less certain than they were a decade ago.
The evidence for cognitive changes has recently been examined by Katherine Jonas and colleagues. The study is behind a paywall at JAMA Psychiatry but the lead author helpfully has PDFs of her author copies available on her lab site. Below, I’ll discuss their study in detail and try to put it in context - does it support the neurodevelopmental or neurodegenerative hypothesis?
The course of general cognitive ability in individuals with psychotic disorders
The premise of the study was explicitly to test whether the course of cognitive deficits in schizophrenia supports the neurodevelopmental or neurodegenerative hypothesis:
Indeed, schizophrenia has been alternatively conceptualized as a neurodevelopmental disorder or a neurodegenerative disorder. The neurodevelopmental model posits cognitive deficits emerge because of disruptions in brain development, marking the beginning of a disease process that ends in psychosis. The neurodegenerative model conceptualizes illness as the result of progressive deterioration. The former predicts cognitive deficits stabilize after illness onset, while the latter implies cognitive declines continue.
This provides a clear framework for examining cognitive deficits and determining whether they stabilise or deteriorate over time.
Summarising the research field, the authors describe clear differences in premorbid IQ between people with schizophrenia and healthy controls, which widens around the time of illness onset. However, there is mixed evidence of longitudinal progression following illness onset.
The study was ambitious - it tracked cognitive trajectory of schizophrenia and other psychotic disorders from childhood (pre-illness) through adulthood. Their study population was made up of schizophrenia spectrum disorders (n=216), and other psychotic disorders (n=212) like bipolar disorder, psychotic depression, and drug induced psychosis.
They managed to get pre-morbid cognitive measures for 218 individuals. These were taken from school records (either directly from IQ tests or imputed from academic achievement).
Follow-up cognitive assessments were obtained at the following time-points: 6 months, 24 months, 20 years, and 25 years. They then used a statistical analysis (multilevel modelling) to fit the data, determining the best model of cognitive trajectories using a measure called the Bayesian information criterion.
Cognitive changes by age are plotted below (reproduced with permission from the author’s copy). LOESS stands for Locally Estimated Scatterplot Smoothing.
Each point is a participant’s IQ score (vertical axis) plotted against their age (horizontal axis). We can see cognitive ability declining in both schizophrenia and other psychotic disorders. This starts in adolescence in the schizophrenia group, and much later in other psychotic disorders. After the age of 60, these two groups seem to be converging (with the caveat that we have substantially less data-points).
In the schizophrenia group, there was a decline of 16 IQ points over the study period. That’s roughly the equivalent of going from average cognitive ability to borderline learning disability. A big deterioration. Things really get interesting when cognitive ability is plotted against time of illness onset and the trajectories calculated by the statistical model are added:
According to the Bayesian information criterion, this was the best model of the underlying data. It has two notches or transition points. You can see again that the vertical axis shows IQ score while the horizontal axis is time relative to psychosis onset (time-point 0), ranging from more than 20 years pre-illness onset to more than 30 years post-onset.
The model is interesting, because it shows cognitive decline in psychosis, more than 10 years before the illness presents. At this point, the average participant would have been aged 13. Furthermore, the decline was steeper in schizophrenia than other psychotic disorders. The model has cognitive ability declining and then deteriorating more steeply from approximately 20 years post-onset.
The authors name three phases of cognitive change in schizophrenia as: normative (normal cognitive development), declining (decreasing cognitive ability) and deteriorating (more severe worsening). Cognition seems to be on a downward trajectory many years before the illness presents and the rate of decline then gets worse approximately 22 years after onset.
They interpret the second inflection point as being consistent with a degenerative process and cite the increased incidence of dementia in individuals with schizophrenia.
They conclude that their findings support both the neurodevelopmental (cognitive decline beginning in adolescence well before illness onset) and neurodegenerative models (decline continuing after illness onset and accelerating as participants age).
Limitations
This is the largest study of cognitive deficits over time in schizophrenia. It is also the first to chart cognitive trajectories over the lifespan (from childhood to old age). It is well conducted and transparent (with information about additional analyses in the supplementary material). Nevertheless like all studies, it has limitations, some of which are already described by the authors.
Firstly, despite a relatively large sample size, the number of participants with schizophrenia was only 216 and the number with complete data (pre-admission cognitive tests) would have been even less. As the authors acknowledge, childhood cognitive tests had to be gathered retrospectively from school reports. In the era of big data and registry studies, I feel there is an even bigger study of prospectively measured IQ waiting to be done.
I couldn’t see a pre-registration of the analysis plan. This is important because in statistical modelling there is wide researcher degrees of freedom (i.e. different choices that can be made in the analysis that affect the results). In general, we should have more confidence in results with a pre-registered analysis and those that survive replication in an independent sample.
Along these lines, in the supplementary material, while the best model for the whole sample was the one with two inflection points, it looks like the best fitting model of cognitive trajectories for the schizophrenia group was actually one with only one notch at 10 years before illness onset.
Though not perfect, overall this is the best study (to my knowledge) of cognitive deficits over time in schizophrenia.
The wider picture
The study presents an interesting pattern of cognitive deficits in schizophrenia which, as the authors say, could be consistent with both a neurodevelopmental and neurodegenerative disease process.
As others have argued though, neurodevelopmental versus neurodegenerative may be a false dichotomy for schizophrenia. While evidence from cognition supports both hypotheses, in many ways schizophrenia does not fit neatly into either category.
Classic neurodevelopmental disorders, like autism and ADHD, almost by definition present in childhood. While there is a rare subset with childhood onset schizophrenia, the disorder does not typically present until late adolescence / early adulthood and it is not uncommon for it to present in midlife (particularly in women).
Meanwhile, schizophrenia lacks many features of classic neurodegenerative disorders, like Alzheimer’s and Parkinson’s disease. We do not see characteristic degenerative pathology when the brains of people with schizophrenia are examined post-mortem.
Furthermore, there are lots of confounding factors that accompany severe mental illnesses like schizophrenia which accumulate over time: poor physical health, drug use, sedentary lifestyle, poverty, social isolation, antipsychotic medication. Therefore a decline in cognition might not necessarily be a primary result of a disease process.
So, while Jonas and colleagues provide evidence of cognitive trajectories consistent with both the neurodevelopmental and neurodegenerative hypotheses, perhaps neither tells the whole story. Alternative views categorise it as an inflammatory or autoimmune disorder. Like the parable of the blind men and the elephant, there may be some truth in each of these theories but none yet unifies what we know about schizophrenia.
It seems to me that schizophrenia (and especially MDD) are imperfect, albeit useful, categories that umbrella together multiple independent etiologies.
If so, the question becomes 'why not both neurodevelopmental and neuro degenerative?'. From what I can tell, there are a subset of people given a schizophrenia diagnosis who follow the predictable neuro developmental course you describe. The map in this case matches the territory very well.
Others' prognosis are more variable (but still meet criteria for schizophrenia). Like you mention, maybe immune or neuro degenerative processes are at play.
I really don't know what to make of the studies you go through because of concern of this problem. It's promising to find any correlations!