Cognitive reserve (CR) is defined as a buffer against neuropathologies made possible through individual differences in neural networks that maintain efficient cognitive performance1 and thus actively compensate for the effects of a given disease.2 Previous research has shown that CR is a predictor of functionality and clinical and cognitive symptomatology.3–6 Thus, the higher the CR, the better the patient's performance in these areas, with CR functioning as a protective factor in patients with severe mental illness.7,8 CR is commonly estimated through the use of proxies and indices. One of the most common compositions of CR is composed of the proxies: years of education, premorbid IQ and employment status.6,9 Premorbid IQ proxy is related to patients’ functional outcomes10 and cognitive performance both at baseline and after disease onset,11 as well as to the levels of certain biomarkers12 Meanwhile, years of education and employment status are related to intellectual and environmental factors at different stages of life: years of education in early life, and employment status in middle and late life.13

Research focused on CR in Alzheimer's disease (AD) clarifies the importance of the proxy years of education, such that patients presenting with a higher level of education tended to have a lower incidence of developing the disease.14 Two explanations would justify this phenomenon. It is possible that highly educated patients have higher cognitive performance that would in turn require advanced pathology to be observed at clinical levels of symptomatology. It is possible as well that education stimulates neuroplasticity, providing resistance to adverse effects on brain function.15 Furthermore, previous heritability studies found that the genetic architecture of dementia was related to education,16,17 suggesting a causal effect on AD.18 However, neurological and psychiatric disorders are complex, polygenic and multifactorial. For this reason, endogenous and exogenous factors should be taken into account for the composition of CR. A recent line of research in the area of AD has recently developed to address this complexity. For instance, studies have integrated extra proxies into the measure of CR, such as the tau protein,13 the Aβ burden13 and the P300 latency.19

A previous study has described the influence of peripheral pathways in schizophrenia.20 This influence causes a link between schizophrenia and the metabolism. One reason for this link is that the probability of neuronal recruitment in the brain is related to the bioenergetic efficiency of neurons and astrocytes. There is thus a link between somatic metabolism and neurological deterioration, a relationship that is affected by external factors such as exercise and diet.21 Initial research reported that biomarkers of high-density lipoprotein (HDL), triglycerides, and fasting plasma glucose (FPG) were predictive of global cognitive impairment in schizophrenia.22 Subsequently, it was clarified that cognition was predicted by total cholesterol and glucose together with the proxy years of education.23 Recently, a study has been published reporting that the prediction model for social functioning in schizophrenia was formed by HDL, triglycerides and FPG, while the biomarker systolic blood pressure (SBP) needed to be added to the model to predict cognitive impairment.24 To date, many studies have highlighted the importance of HDL levels in schizophrenia, indicating a relationship between low HDL levels and poorer global functioning, global cognition,25–27 poor social relationships,28 as well as increased suicidal behaviour.29 On the other hand, although it has been reported that HDL levels tends to remain stable during the first three years of the disease,30 any changes in these lipids could produce changes in brain structure and inhibition of synaptogenesis, and consequently neurodegenerative changes.31 This is because cholesterol plays an important role in the brain due to its involvement in myelination, synaptic growth and regeneration,31 and it has been shown that high levels of HDL may act as a protective factor against hippocampal atrophy.32

To the best of our knowledge, no previous study has assessed the relationship between biological parameters and CR proxies in first-episode psychosis (FEP). This lack of knowledge justifies the present study, which focused on the search for neurobiological proxies of CR.

Based on the scientific evidence, we hypothesised that HDL levels are related to CR proxies and specifically, that HDL levels could be an extra proxy for CR in patients with FEP. The aim of this study was to construct an index of CR that is objective, quantifiable and multifactorial.

The main objective of the present study was to explore an objective and quantifiable CR, taking into account endogenous and exogenous factors. Following this line of work, it is interesting to note that, for both groups of participants, the results found that the proposed biological CR has a greater explanatory value of the phenomenon than the classic CR commonly found in the literature. Therefore, we can affirm that introducing HDL levels, as a biological proxy, together with the years of education proxy, is a way of obtaining a CR adjusted to the multifactorial nature of psychosis. Furthermore, it was found that in the classic CR the proxy unemployment did not provide the necessary information load to be taken into account in the total index.

Our research has found that years of education is the variable that best predicts HDL levels. These results are supported by previous studies suggesting that years of education in combination with HDL are necessary to form predictive models of cognitive impairment in schizophrenia.22–24 The fact that years of education is an indispensable proxy may be due to the fact that it often has an early onset, which makes it more important to take into account lived experiences and cognitive enrichment at an early age.46 Another reason why the need for this proxy might be justified is that education stimulates neuroplasticity before the development of the disease, which would promote cognitive protection against the adverse effects of psychosis.15

Furthermore, the need to include HDL levels is supported by previous arguments reporting its relationship with cognitive performance and patient functionality,25,28 independently of the effect of practice, lifestyle, improvements in clinical symptomatology or levels of other biological parameters.32 The introduction of HDL levels in CR implies taking into account biochemical processes that have not been considered until now, since, at the brain level, HDL is located in the hippocampus and is related to the performance of cognitive domains.32,47–49 If we know that an imbalance in HDL levels causes hippocampal atrophy32 with consequent cognitive impairment and increased neurodegenerative speed,31 it is conceivable that it also affects the alternative neural networks referred to in the definition of CR.1 Ultimately, these problems could explain the patients’ inabilities to cope with the onset of their first episode of psychosis and subsequent cognitive declines. This fact supports the claim that CR is a protective factor for the development of severe mental illness, as suggested by previous research.7,8

The combination of years of education and HDL levels seems to have an important implication for CR. On the one hand, if a patient has low cognitive enrichment due to a low level of education, this influences the patient's understanding, learning and reasoning about their health and achieving problem-solving skills to prevent alterations in their biological parameters and future chronic diseases.50 Moreover, if a patient has low HDL levels, it will start an inhibition of synaptogenesis and lead to neurodegenerative changes31 that in turn lead to worse cognitive functioning, making it difficult for the patient to increase their years of education. Therefore, as we can see, the relationship between these two proxies appears to be reciprocal.

Another notable finding of our research is that the proxies premorbid IQ and employment status (unemployment) were discarded. One possible explanation is that years of education and HDL levels can be considered secondary. The results indicated a strong relationship between premorbid IQ and years of education, which leads us to believe that, if a patient has a low educational level, his or her cognitive enrichment will be poor and therefore will not reach a high premorbid IQ. Thus, we come to the possible conclusion that years of education is a predictor of the patient's crystallised intelligence, making the premorbid IQ proxy redundant in the CR composite score. On the other hand, having low HDL levels is related to poorer performance in the domains of verbal learning and memory.47–49 These domains represent the receiving, processing and storing of long-term knowledge, suggesting that low HDL affects the development of crystallised intelligence and, as a result, premorbid IQ. Secondly, if functionality is impaired, as predicted by low HDL levels, it will be more difficult for the patient to find a job, which helps to explain the relationship between HDL and employment.

This paper cannot end without mentioning some limitations and strengths. One of the main limitations lies in not being able to explore the relationship with CR in other inflammatory or neuroimmune biomarkers such as interleukin receptor SHC signalling, hematopoietic cell lineage, interleukin-1/2 family signalling51 and activation of HERV-W.20 We also did not take into account the hormone levels of the participants and especially of the patients. Previous research has shown how antipsychotic medication influences hormone production and how hormones mediate HDL production.52 Another possible limitation lies in the collection of blood tests, as they were not performed on all participants. We also consider it a limitation that due to the lack of uniform information for occupation, it was used as a dichotomous variable (unemployed vs. active). As a final limitation, we can mention that we have not taken into account other factors that could potentially interfere with the results, such as habitual physical exercise, diet, substance use and type of antipsychotic. However, our study has several strengths consisting in the size and homogeneity of the samples, which allows us to consider our results for future CR research. To our knowledge, this is the first study to explore the relationships and predictions between CR proxies with biomarkers, as well as to explore their use as an additional proxy. The gap in the literature justified the present study's focus on the search for neurobiological CR proxies, as patients with FEP tend to have unhealthy lifestyle habits that alter their metabolisms53,54 and thus their CR. Future research is needed to expand the use of biological CR based on years of education and HDL levels.

In conclusion, this study proposes a new composition of CR with exogenous and endogenous factors, consisting of the proxies years of education and HDL levels. We showed that, in patients and HC, the strongest correlations were between years of education and HDL levels. In fact, years of education outperformed the other variables in terms of predicting HDL levels. A notable observation was that our proposed biological CR had a higher percentage of explanation of the phenomenon with only two proxies compared to the classic CR consisting of three proxies. In addition, its two proxies contributed information in a balanced way to the CR composite score. These findings allowed us to develop a quantifiable, objective and multifactorial CR. The possible impact of the biological CR is that is provides a more precise composition taking into account biomarkers through a question and a blood test, or, desirably, through non-invasive measures yet to be developed. These insights will allow for the selection and establishment of the most appropriate treatments for each FEP patient.

All the authors have participated and have made substantial contributions to this paper.

RMH: conception, design, statistical analysis, interpretations of data and drafting the article.

NMG, AYR, BCF: interpretations of data and revising the article.

KN: editing.

RAA: conception, design, interpretations of data and revising the article.

The PAFIP project carried out on Marqués de Valdecilla Research Institute was supported by the Carlos III Health Institute (PI14/00639 and PI14/00918). In addition, this work was financed by a Miguel Servet from the Carlos III Health Institute contract (Dra. Rosa Ayesa-Arriola) (CP18/00003), and a predoctoral contract (Rebeca Magdaleno Herrero) (PREVAL22/03) from the Valdecilla Biomedical Research Institute and the University of Cantabria. No pharmaceutical company has financially supported the study.

The authors have no conflict of interest to declare.