Patients diagnosed with schizophrenia (F20.x, SCH group) according to the tenth revision of the International Classification of Disease [ICD-10, (20)] criteria were recruited from the Academic Psychiatry Clinics associated with the Medical University of Lublin. All individuals had to fall within the age range of 21–65 years and had to have at least 12 years of education. Subjects from the clinical group were excluded if they had a serious neurological or medical condition or a history of psychoactive substance (excluding nicotine) addiction in the past 6 months. All patients were treated with antipsychotic medications. Daily dosages were converted into risperidone equivalents (21). The assessment took place during psychiatric hospitalization, after a minimum of 4 weeks of treatment with antipsychotic drugs, after which functional improvement was achieved, enabling patients to participate in the study. Patients' clinical characteristics were reconstructed on the basis of the available medical records. Healthy controls matched for age, gender, and education were recruited from among those reporting for checkups to non-public healthcare facilities in the Lubelskie Voivodeship. All participants gave their written informed consent after receiving detailed information. The study was conducted following the Declaration of Helsinki and received written approval from the Bioethics Committee of the Medical University of Lublin (KE-0254/248/2020).

Regarding all participants, the following exclusion criteria were implemented: any previously diagnosed ophthalmological disorders (e.g., glaucoma, macular degeneration, and diabetic retinopathy), diabetes mellitus, non-treated arterial hypertension (if treated, BP should be > 140/90 mm Hg during an evaluation), obesity (BMI above 30), history of ocular trauma, ocular surgery, eye refraction above ± 5 dpt, glaucoma suspect (DDLS ≥ 6), dementia, and any relevant, concomitant psychiatric disorder (patients with F25.x were excluded from the study). Ophthalmological abnormalities were detected by a certified ophthalmologist. Interviews and ophthalmologic and cognitive evaluations were conducted on the same day. The psychiatric assessment took place within the 1-week time of the OCT test.

All participants underwent retinal imaging using the OPTOPOL COPERNICUS REVO^® Spectral Domain – OCT (OPTOPOL Technology, Zawiercie, Poland) with an upgraded scanning speed of up to 80, 000 measurements/s. The Spectral Domain – OCT is characterized by a SLED light source operating on a wavelength of 830 nm, an axial resolution of 2.6 μm, a transverse resolution of 12 μm, and a scan depth of 2.4 mm. All the results were processed with built-in OPTOPOL SOCT 11.0.7 software utilizing automatic segmentation of individual retinal layers with the manual correction of an experienced ophthalmologist. Only high-quality images were accepted (QI ≥7) for further analyses (Figure 1).

Measures of MV and MT were subdivided according to the ETDRS grid for both eyes. ETDRS grid divides the macula into three circles: central circle (1 mm), inner circle (3 mm), and outer circle (6 mm). The inner and outer circles could be additionally divided into four quadrants, namely, inferior, superior, nasal, and temporal. Measures of GCC (mRNFL+GCL+IPL) were subdivided into six segments, namely, superior, inferior, nasal superior, nasal inferior, temporal superior, and temporal inferior.

Images were acquired using a protocol consisting of 640 A-Scans and 85 B-scans in a 7 × 7 mm area centered in the fovea. MT and MV were measured from ILM to OS/RPE border. Peripapillary RNFL measures were obtained with a protocol of 512 A-scans and 112 B-scans in a 6 × 6 mm area centered in the optic disk. The measurements of RNFL were taken from the ring area of 2.4 mm diameter and 0.4 mm thickness from the optic disk. This area was divided into four quadrants, namely, inferior, superior, nasal, and temporal.

Schizophrenia psychopathology was evaluated using the Positive and Negative Syndrome Scale (PANSS). The PANSS employs a semi-structured interview assessing three domains, namely, Positive Symptoms Scale (7 items), Negative Symptoms Scale (7 items), and the General Psychopathology Scale (16 items) (22).

The overall level of cognitive functioning was assessed with the application of the ACE-III Scale (23). The cutoff point excluding dementia was set at 70 points (24). A vocabulary test from the WAIS-R(PL) Scale (25) was used to measure word knowledge and linguistic concept formation ability. Two indicators of cognitive functions dependent on performance speed were used: a Trail Making Test (TMT (26)) measuring mainly attention and psychomotor speed (part A) and mental flexibility (part B). The main outputs of TMT are time of performance and a number of errors informing about execution accuracy. In addition, to use a task with a time limit, but not dependent on the visual modality, the verbal fluency test in the phoneme version was applied. The subjects were asked to give as many unique words as possible for 60 s, beginning with a given letter (K, A, and T were used). The given words cannot be proper names and could not be repeated within one trial. The test result was the average number of unique words given in three attempts and the number of errors. Errors included perseveration and words that did not follow the instructions (27).

Due to the results analysis plan containing many multivariate methods and the assessment of complex within- and between-group effects, all statistical analyses included the averaged parameters of the OCT results initially measured in separated quadrants. The raw values of MT, MV, mRNFL, pRNFL, and GCC were z-standardized to obtain outcomes adhering to the normal distribution. To verify whether retinal age-related trajectories were significantly different in groups of patients and controls, a two-way ANOVA was implemented, with diagnosis (HC vs. SCH) as the first factor and age range (younger, middle-aged, and older) as the second factor. The significance level was set at a p-value of <0.05, but it was adjusted to the impact of multiple comparisons by applying Bonferroni correction. If the age subgroups among the SCH sample differed significantly regarding OCT outcomes, an ANCOVA was used to test the potential impact of antipsychotic doses on these effects. For all ANOVAs, partial eta squared (ηp2) was an effect size indicator. Pearson's correlation coefficient (r) was applied to assess relationships between studied variables, with the statistical significance level corrected for the number of performed analyses. If the correlations were found to be significant also after adjusting for multiple testing, the relationship was validated using a linear regression analysis including the risperidone equivalent as a controlled variable. In the Results section, tables contain raw metrics to show the values typical for the studied groups.

Considering all previously described inclusion and exclusion criteria and the quality of OCT recordings, data from 60 SCH patients and 61 healthy controls (HC) have been ultimately analyzed. Taking into account participants' age distributions, each sample was divided into three subgroups. The younger (Y) participants were aged between 20 and 31 years, the middle-aged (M) participants were aged between 32 and 45 years, and the older (O) participants were aged between 46 and 65 years. Such age ranges enabled distinguishing subgroups with relatively similar participant numbers.

As shown in Table 1, the age subgroups in SCH and HC samples did not differ significantly regarding sex and years of education; in the HC sample, the age subgroups were also similar regarding the overall score on the ACE-III Scale; and in the SCH sample, the older ones scored significantly worse than the younger ones (p < 0.001). As for the SCH sample, the duration of untreated psychosis, number of hospitalizations, and the score in the PANSS subscale of Positive Symptoms did not differ in the subgroups; however, scores of Negative Symptoms (N) and General Pathophysiology Scale (G) subscales were significantly higher in the O or M subgroup compared with Y. Middle-aged patients were also treated with significantly higher doses of antipsychotics compared to the older ones. The SCH subgroups differed significantly regarding performance time of the Trail Making Test parts A and B and with reference to the verbal fluency test outcomes (Supplementary Table S1).

Table 2 consists of all evaluated retinal measures (MT, MV, mRNFL, pRNFL, GCC) in the age subgroups with effects of age (Y vs. M vs. O) and effects of diagnosis (HC vs. SCH). According to the presented results, none of the interaction effects turned out to be significant.

For the right MT, the age effect was non-significant concerning the adjusted level of statistical significance (p > 0.016). The group effect was significant (HC > SCH, p < 0.001), and as shown via the post-hoc analysis, it originated from the difference between the middle-aged and older subgroups of controls and patients. Within the SCH sample, right MT was significantly decreased in older patients compared with younger ones, and this difference remained significant also after controlling for risperidone equivalents (p = 0.004). The diagnosis effect was significant also regarding left MT (p < 0.001) and resulted from significant differences between middle-aged patients and controls. As for the right MV, a significant effect of diagnosis (p < 0.001) was driven by the difference between middle-aged and older patients and controls. Younger patients had significantly greater right MV than the older ones, and again, this difference remained significant also after controlling for risperidone equivalents (p = 0.002). Regarding left MV, the effect of diagnosis was significant (p < 0.001) and again resulted from the differences between middle-aged patients and controls. No significant effects were noticed regarding right and left mRNFL. Right peripapillary RNFL was significantly smaller in patients compared with controls (p < 0.001); however, no post-hoc tests reached the adjusted level of significance. Regarding left pRNFL, the results were similar, with a significant diagnosis effect (p < 0.001). Significant diagnosis effects concerned also right and left GCC (both p < 0.002), but without significant interaction post-hoc tests. Figure 2 represents the results of the analyzed variables in the HC and SCH age subgroups for MT, MV, pRNFL, and GCC in the right and left eyes.

The next step of the analysis involved verifying the associations between age and retinal parameters in the SCH and HC groups. To reduce the number of variables analyzed at once, the retinal measures for both eyes were averaged. No significant correlations were noted in the HC sample. In the complete SCH sample, the following retinal parameters correlated significantly with age: MT (r = −0.47, p < 0.0001), MV (r = −0.49, p < 0.0001), and GCC (r = −0.29, p = 0.023, this p-level, however, does not survive the correction for multiple analyses). The correlations between age and pRNFL and between mRNFL and age were non-significant (p > 0.1). For MT and MV as dependent variables, a linear regression analysis was performed to verify whether correlations with patients' age remained significant also when controlling for the risperidone equivalent (RE). Age was not correlated with risperidone equivalents. Both regression models were statistically significant (Supplementary Table S2). Despite the relatively weak correlation between GCC and age, a regression analysis was carried out to determine a possible relationship also after controlling for risperidone equivalent. The regression model turned out to be statistically significant (p = 0.022), although the range of explanation of the dependent variable was small (corrected R² = 0.09, Supplementary Table S2). Considering the significant results of the regression analysis for age, MT, MV, and GCC, additional correlational analyzes were performed in the age subgroups of patients. In the middle-aged subgroup, age correlated with MT (r = −0.43, p = 0.045) and with MV (r = −0.45, p = 0.033); however, the significance level of these correlations did not survive the multiple testing correction.

In the whole SCH group, there were no significant correlations between MT, MV, pRNFL, GCC, years of education, and vocabulary (all p > 0.09). After controlling for multiple testing, the performance time of the Trail Making Test part A correlated significantly with MV: r = −0.52, p < 0.001, GCC: r = −0.50, p < 0.001. A number of errors committed in this task correlated with MV: r = −0.31, p = 0.033; however, the level of statistical significance did not survive the correction for multiple testing. The performance time of TMT part B correlated significantly, also after controlling for multiple testing with GCC: r = −0.43, p = 0.003. The number of errors committed during TMT part B execution correlated significantly with GCC: r = −0.54, p < 0.001. Regression analyses including the above cognitive measures, risperidone equivalents, and identified OCT correlates turned out to be significant (p < 0.001, Supplementary Table S3), with one exception: the relationship between performance time of TMT part B and GCC ceased to be statistically significant after adding the risperidone equivalent as a control variable.

Verbal fluency output correlated significantly with GCC: r = 0.44, p = 0.002. This relationship remained significant also in regression analysis, with risperidone equivalent as a controlled variable (p < 0.001, Supplementary Table S3). The number of errors committed in the fluency test did not correlate significantly with any of the OCT results in the group of patients (p > 0.055). No significant correlations between analyzed cognitive and retinal measures were noted in the SCH age subgroups.

All performed analyses concerned only the averaged parameters of retinal layers' morphology; therefore, the obtained results may not reflect the potential changes occurring in individual retinal quadrants with the disease progress (39). As shown in Figure 2, macular thickness and volume decreased more prominently with age in the SCH groups compared with controls; however, no interaction effects reached the statistical significance level. These results might be conditioned by relatively small age subgroups, which probably increased the impact of morphological measures variability. The conclusions about the possible aging acceleration of macula in schizophrenia patients are not based on advanced computational methods applied in neuroimaging studies using MRI or DTI but rather a preliminary suggestion regarding the possibility of interpreting the obtained results. Considering the anatomical complexity of the retina (40, 41) it seems possible to apply machine learning to calculate retinal layers' age in the large research group, and such analysis would ultimately determine the validity of our hypothesis.