Characterization of affective lability across subgroups of psychosis spectrum disorders

The current study is part of the larger Thematically Organized Psychosis (TOP) study at the Norwegian Center for Mental Disorders Research (NORMENT) in Oslo, Norway. Recruitment to the TOP study has been ongoing since 2003 and potential participants are referred via psychiatric inpatient and outpatient units, including specialized psychosis units as well as community teams, in a catchment area that is comprised of all the major hospitals in Oslo. As such, the overall representability of the sample is considered to be very good. The participants are given thorough information about the voluntary nature of the study and the possibility to withdraw at any time. The participants included in the study have to meet diagnostic criteria for a Diagnostic and Statistical Manual of Mental Disorders 4th Edition (DSM-IV) diagnosis of schizophrenia- or bipolar spectrum disorder and provide informed consent. Both individuals with established diagnoses and individuals who are diagnosed for the first time are included in the study. Further inclusion criteria are intelligence quotient (IQ) above 70, no prior history of severe head trauma and sufficient understanding of a Scandinavian language (Ringen et al. 2008).

The sample for the current study was comprised of two hundred and ninety-seven participants with psychosis spectrum disorders from the TOP study, and only participants who had completed the ALS-SF were included. The ALS-SF was originally introduced in a TOP study sub-protocol for participants with first episode mania, and a few years later included in the main protocol (i.e. to patients with other diagnoses than BD-I). It was mainly presented for participants with low levels of current affective symptoms. The diagnostic grouping in the current study was as follows: SZ (including schizophreniform [n = 14]) n = 76, BD-I n = 105, BD-II = 68 and MP (including psychosis NOS [n = 32] and schizoaffective disorder [n = 16]) n = 48. The rationale for combining psychosis NOS and schizoaffective disorder into one “mixed” group was that these categories typically include patients with both psychotic- and affective symptoms that are diagnostically more heterogeneous than the other groups (Santelmann et al. 2015, 2016; Widing et al. 2020). Of the present sample, n = 222 were used in our previous study investigating affective lability across the psychosis spectrum (Høegh et al. 2020), and are now re-analyzed along with the newly added participants to shed light on putative differences in affective lability between the specific psychotic disorders.

The diagnoses in the study were established by the Structured Clinical Interview for DSM-IV axis I disorders (SCID-1), modules A-E (First et al. 1995) which was carried out by trained medical doctors, psychiatrists or clinical psychologists. In the TOP study, diagnostic reliability is assessed with regular intervals and Cohen’s kappa for diagnosis in the range between 0.92 and 0.99 has been found across different assessment teams. To assess current symptom state, the positive subscale of the Positive and Negative Syndrome Scale [PANSS (Kay et al. 1987)] was used for positive psychotic symptoms, and the depression item (G6) and the anxiety item (G2) in the general scale of the PANSS were used for depressive- and anxiety symptoms, respectively. PANSS G6 and PANSS G2 were chosen because they were the only measures of depression and anxiety that were collected at the same time point as the ALS-SF for all participants. The rating for G6 is based on the answer to one initial question (“how has your mood been in the past week, mostly good or mostly bad?”) followed by 1–11 follow-up questions concerning the extent of the depressive state and its behavioral consequences. For G2, the rating is based on the same algorithm; one initial question (“have you been feeling worried or nervous in the past week?”) and then 1 to 6 follow-up questions depending on the response to the first question. The Young Mania Rating Scale [YMRS (Young et al. 1978)] was used to assess manic symptoms. The Alcohol Use Disorders Identification Test [AUDIT (Saunders et al. 1993)] was used to evaluate the degree of harmful alcohol consumption, and drug related problems were measured with the Drug Use Disorders Identification Test [DUDIT (Berman et al. 2005)]. Duration of illness was estimated based on the age of onset of the first SCID-verified episode of psychosis for SZ and MP, and the first SCID-verified affective episode for BD-I and BD-II.

To measure affective lability, we used the ALS-SF (Oliver and Simons 2004). The ALS-SF captures the total level of affective lability reported by an individual, but also subdimensions of affective lability covering oscillations between three subdomains: anxiety-depression, depression-elation and anger and normal mood. Thus, the scale provides an indication of whether affective lability is predominantly driven by specific- or a combination of affects. The ALS-SF has been found to have good psychometric properties (Aas et al. 2015; Look et al. 2010) and is widely used across different clinical populations. The 18 items of the scale are rated on a four-point Likert scale ranging from 0 (“very uncharacteristic of me”) to 3 (“very characteristic of me”) and yields a total score of affective lability as well as scores for the three subdomains. There are no validated cut-off scores for evaluating the severity of affective lability, but in our previous study we found mean ALS-SF total and subscale scores in the range of 0.17–0.39 for healthy controls, 0.85–1.33 for individuals with BD, and 0.69–1.34 for individuals with SZ (Høegh et al. 2020). This corresponds well with what has been found in at least one similar study (Marwaha et al. 2018).

Descriptive statistics, including means with standard deviations or frequencies with percentages where relevant, were used to investigate demographical and clinical characteristics of the different diagnostic groups. The groups were then compared using one-way between-groups analyses of variance (ANOVA) and chi-square tests. The Tukey’s honestly significant difference (HSD) test was used for post-hoc comparisons where appropriate. To investigate group differences in the level of affective lability as measured by scores on the total- and subdimensions of the ALS-SF, a one-way between-groups multivariate analysis of variance (MANOVA) was conducted with Bonferroni post-hoc tests. A multiple analysis of covariance (MANCOVA) was then carried out to investigate if statistically significant group differences in affective lability remained significant when current symptom and substance use status were entered as covariates. The variables which were significantly associated with diagnostic group and/or with the ALS-SF domains in bivariate correlation analyses were entered as covariates. The selected covariates were as follows: current level of positive psychotic symptoms (PANSS P), manic symptoms (YMRS), depression (PANSS G6), anxiety (PANSS G2), alcohol use (AUDIT) and drug use (DUDIT) (see Table 1 for correlation coefficients). In addition to the symptom and substance use variables, sex was entered as a covariate as there was a significant difference in the number of females in the different groups and previous research has indicated that affective lability is higher in females in general (Marwaha et al. 2013b; Winkler et al. 2004). Age, which has also been found to be associated with affective lability previously (Broome et al. 2015b), was not associated with the diagnostic groups or with the ALS-SF scores in our sample and therefore not included in further analyses. Effect sizes were calculated by partial eta squared. To investigate which of the ALS-SF subdimensions contributed the most to the total affective lability, a one-way repeated measures ANOVA was performed for each group. The Statistical Package for the Social Sciences (SPSS Inc., Chicago, IL, version 26) was used for all statistical analyses and a significance level of p ≤ 0.05 (two-tailed tests) was employed. For aim 2, ALS-SF scores for all dimensions for each diagnostic group were plotted into the Graphpad Prism tool (GraphPad Software, La Jolla California USA, version 8.0 for Windows) and converted into violin plots to illustrate score distributions. To further investigate variability in the ALS-SF scores, double generalized linear models (DGLM, Y ~ Dx, ~ Dx) were carried out using R (R core team 2017) to test if the ALS-SF score dispersions were significantly different between groups (for more detailed information see Additional file 1).

Demographic and clinical characteristics for SZ, BD-I, BD-II and MP are presented in Table 2. A statistically significant difference in sex between the groups was found, with fewer women in the SZ group compared to remaining groups. There were also significantly more women in the BD-II compared to the BD-I group. In addition, the duration of illness between the groups was significantly different; the BD-I and BD-II groups had been ill longer than the SZ and MP groups (p < 0.001 and p < 0.05, respectively). Regarding clinical features, there were statistically significant differences between the groups for PANSS total, PANSS P, YMRS, G2 anxiety and AUDIT (see Table 2).

The mean scores for the ALS-SF dimensions for the different diagnostic groups are presented in Fig. 1. The MANOVA showed that there was a statistically significant difference in affective lability between the groups for all of the ALS dimensions: ALS-total F = 8.446, df = 3, p < 0.001; ALS anxiety-depression F = 9.298, df = 3, p < 0.001; ALS depression-elation F = 7.281, p < 0.001 and ALS anger F = 4.252, p = 0.006. Effect sizes were moderate; partial eta² = 0.080, 0.087, 0.069 and 0.042, respectively.

Post-hoc analyses with Bonferroni revealed that the BD-II group had significantly higher affective lability scores compared to all of the other groups for the ALS total (p < 0.05) and depression-elation (p < 0.05) dimensions. For the ALS anxiety-depression dimension, the scores for the BD-II group were significantly higher than those of the BD-I and SZ groups (p < 0.001 for both), but not the MP group (p = 0.060). Finally, the BD-II group had significantly higher scores compared to the SZ group (p = 0.006) and the MP group (p = 0.042) groups, but not the BD-I group (p = 0.089) on the anger dimension. There were no significant differences in the scores for the SZ and BD-I groups on any dimension (p = 1.000).

The overall differences in affective lability between the groups remained statistically significant also after adjusting for the effects of sex, current symptom- and substance use status: ALS total F = 5.305, df = 3, p = 0.001; ALS anxiety-depression F = 6.139, df = 3, p < 001; ALS depression-elation F = 4.432, df = 3, p = 0.005 and ALS anger F = 4.184, df = 3, p = 0.006. Again, the effect sizes were moderate with partial eta² of 0.057, 0.067, 0.048 and 0.046. Post-hoc group comparisons with Bonferroni showed that the difference in AL between BD-II versus SZ and BD-I remained statistically significant for the ALS total (p = 0.004 for both SZ and BD-I), anxiety-depression (SZ p = 0.038, BD-I p = 0.001) and depression-elation dimensions (SZ p = 0.031, BD-I p = 0.006). The difference between BD-II and SZ on the anger domain also remained statistically significant (p = 0.004). The difference between the MP group and the BD-II group no longer remained statistically significant for the total-, depression-elation- and anger domains after adjusting for covariates.

In all of the diagnostic groups, the one-way repeated measures ANOVAs with Bonferroni post-hoc tests showed that the anxiety-depression and depression-elation dimensions contributed most to the total affective lability, with no significant differences in the level of scores between these dimensions in any group (p = 1.000 for SZ and BD-I, p = 0.225 for BD-II, p = 0.814 for MP). Across the board, the anger dimension was found to contribute significantly less to the total score compared to the anxiety-depression and depression-elation dimensions (p < 0.001).

The score distributions for all dimensions of the ALS-SF for the four diagnostic groups are shown in Fig. 2a–d. The median score is illustrated by a vertical straight line. The double generalized linear models found no significant differences (p > 0.05) in the score dispersions between any of the groups in any of the domains, indicating that despite differences in mean ALS-SF scores between groups, the dispersion in the scores was the same (for more information see Additional file 1).

The present study must be interpreted in light of some limitations. The cross-sectional nature of the study implies that we cannot make causal attributions about the association between diagnostic group and affective lability. In addition, data on comorbid anxiety disorders, personality disorders and ADHD are lacking, and the measures used for anxiety and depressive symptoms are based on a scale primarily developed for assessing psychotic symptoms. Hence, the possibility that current symptoms still could have influenced the association between affective lability and diagnostic group cannot be ruled out completely. However, we believe that the likelihood of this is limited due to the relatively low levels of anxiety and depressive symptoms. Further, the risk of recall- and response bias cannot be ruled out as the ALS-SF is a self-report instrument. The study also has several strengths; it has a large, diagnostically well-characterized sample covering the psychosis spectrum, and uses a multidimensional assessment scale of affective lability that provides a richer insight and understanding of the construct.

Our results illustrate that affective lability is more prominent in individuals with BD-II compared to SZ and BD-I, and that this is not explained by differences between the groups in sex, levels of affective-, psychotic- or anxiety symptoms or severity of substance use. BD-II thus appears to be a particularly vulnerable diagnostic group with respect to affective lability. No differences in affective lability were found between individuals with BD-I and SZ. The results further add information about the structure of affective lability in these disorders emphasizing the significance of fluctuations between depressive- and other affective states. The findings also show that there is an even dispersion of affective lability scores within each diagnostic group, and that the dispersion also appears to be largely equivalent across groups. Overall, the study provides the concept of affective lability in psychotic disorders with more granularity by showing differences and similarities between diagnostic groups that may have implications for both research and clinical practice.