Background
It has been long recognized that schizophrenia is a highly heterogeneous disease in terms of etiology, symptom manifestation, and treatment response [
1‐
3]. Data from both randomized controlled trials [
4‐
7] and a 30-year long observational study [
8] has shown that treatment response is heterogeneous and is typically captured by four or five trajectory groups with symptom severity defined by the aggregated symptom scores. While the core symptoms of schizophrenia are characterized by positive (e.g., delusion, hallucination) and negative (e.g., emotional withdrawal) symptoms [
9], it has been believed that currently available antipsychotics work primarily on relieving positive symptoms, whereas negative symptoms are harder to treat [
10‐
13]. The schizophrenia research community has tremendous interest in understanding the nature of negative symptoms [
8,
14,
15].
The literature, though, is limited regarding the longitudinal patterns of negative symptoms and their linkage with the positive symptoms. Previous studies on the relationship between negative and positive symptoms have been mainly conceptually driven or cross-sectional in nature [
16]. It has been hypothesized that positive symptoms are related to an increase in dopamine D2 activity, while blockade of the dopamine D2 receptors may worsen negative symptoms, thus negative and positive symptoms might be inversely correlated [
17‐
19]. It was also hypothesized that these two domains of symptomatology correspond to different etiopathogenic and pathophysiological mechanisms [
20,
21]. Concurrently, cross-sectional phenomenological studies have shown that there was no significant relationship between these two symptom domains, thus supporting the notion that these two symptom domains may be orthogonal or independent with each other [
16,
22]. On the other hand, hypotheses on the glutamatergic system suggest that glutamate is involved in the mediation of both positive and negative symptoms of schizophrenia in the same direction, although the data are inferential regarding drugs are currently in late stage clinical development [
23].
In this study, we utilized an individual-based trajectory analysis technique to characterize the trajectories of negative and positive symptoms and assessed the temporal interplay between these two symptom domains by using data from a 1-year schizophrenia trial [
24]. We seek to provide insight into the dynamic, individual-level interaction between negative and positive symptoms. Insight gained from these analyses could aid in management of patients with schizophrenia by raising awareness of the dynamic interplay between these two characteristic symptom domains and may help shed light on the pathological disease process of schizophrenia.
Discussion
This study employed an individual-based trajectory analysis method to study the temporal relationship between negative and positive symptoms. We observed that negative and positive symptom trajectories tended to move in parallel over the 1-year study for most of the 11 patterns of combined latent trajectory classes, and correlation between the changes of PANSS negative and positive symptoms was significant as demonstrated by Pearson correlation coefficients. These findings suggest that changes in negative and positive symptoms are neither inversely related nor independent with each other, at least in chronically ill patients treated in the United States, who represent the study population.
To the authors’ knowledge, the present study is the first to assess the differential symptom courses simultaneously using both negative and positive symptoms over 1 year of antipsychotic treatment. The pattern (latent trajectory classes) was mathematically derived using an algorithm designed to group patients into clusters so that the symptomatology pattern over 1 year is relatively homogeneous in each class. There was no preconceived notion (e.g., baseline characteristics) used to assign patients to one trajectory versus another. The adoption of the trajectory analysis makes possible, for the first time, the analyses and inspection of the relationship between these two symptom domains taking into account both the dynamic and heterogeneous nature of the symptom manifestation [
35,
36]. In our previous research, we applied trajectory analysis to schizophrenia clinical trial data and identified subgroups of patients with relatively homogeneous treatment-response patterns in terms of the PANSS total score or positive symptom subscale score [
3,
36,
37], but this technique had yet to be applied on negative and positive symptoms simultaneously.
The congruence observed for longitudinal change in negative and positive symptoms was not anticipated per theories that these two sets of symptom domains might be reversely correlated [
19] or be independent with each other [
20,
38]. Although the underlying mechanism of this phenomenon is not well understood, our finding might be explained by the newly reconceptualized dopamine hypothesis proposed by Howes and Kapur [
39]. Based on the emerging evidence in imaging, genetic, environmental, phenotypic, and animal studies in the past two decades, Howes and Kapur hypothesized that various symptoms in schizophrenia are systematical manifestations of the downstream neurotransmitter abnormality (i.e., the dopamine dysregulation), and different symptom domains may depend on each other through a unified upstream pathological disease process [
39]. Our findings that changes n negative and positive symptoms were positively related may be a reflection of a unified pathological disease process.
Furthermore, our findings corroborate those reported by Addington and Addington [
16] who studied 41 patients with DSM III-diagnosed schizophrenia and assessed the relationship between negative and positive symptoms measured by the Scale for the Assessment of Negative Symptoms and the Scale for the Assessment of the Positive Symptoms. Their study found positive correlations between these two symptom domains at both the acute phase and remission phase and concluded that negative and positive symptoms are not inversely related.
We observed that negative and positive symptoms moved in parallel over the 1-year period for most of the 11 patterns of combined latent trajectory classes. It could be argued that the congruence of these two symptom domains may be confounded by measurement error at the level of rating (i.e., a high score on one dimension leads to an erroneous high rating on another dimension; therefore, as the score on one dimension improves the score on the other dimension also seems to improve). While the possibility of measurement error cannot be ruled out in the context of an observational study, we think it is rather unlikely in our study as the correlation between the two dimensions was seen across time, and the trajectory subgroups have been observed to be directionally consistent with 1-year functional and health-related quality of life outcomes in a follow-up analysis (data not shown).
It may also be argued that the observed congruence between negative and positive symptom trajectories may be a reflection of the secondary negative symptoms, the effect on negative symptoms that associate with positive symptoms, extrapyramidal side effects, or mood [
40]. Again, we think the concept of secondary negative symptoms is not likely to explain all the congruence between these two symptom trajectories observed in our study. Tollefson and Sanger [
41] conducted a path analysis to tease out the secondary effect of positive symptoms, mood, or adverse events and found that negative symptoms of schizophrenia are directly responsive to treatment. Stauffer et al. [
42] studied primary negative symptoms using the proxy of predominant negative symptoms precluding the effect of positive symptoms, depressive symptoms, and parkinsonism, but the study showed that such segregation of patients does not suggest prognostic implications.
From a clinical perspective, negative symptoms, regardless of being primary or secondary in nature, as a whole are an indication of disease severity [
43] and are significantly related to quality of life and level of functioning [
44,
45]. Our findings agree with findings from a majority of the studies on treatment with second-generation antipsychotics in which both negative and positive symptoms improved at the population level [
28‐
30,
46‐
48]. Further, our findings suggest that, even at the individual patient level, improvements in both negative and positive symptoms are achievable, although there is still a considerable unmet medical need in term of the magnitude of the improvement in both the negative and positive symptoms. The perception that negative symptoms do not improve compared to positive symptoms may not be totally accurate.
Moreover, we observed that negative symptom subscale scores tended to be consistently higher than positive symptom subscale scores. Although the clinical meaning of a certain score on the PANSS negative or positive subscale is not clear, this observation is consistent with the previous findings that populations with chronic schizophrenia have higher negative symptoms [
45,
49] and if they improve in concert, it is understandable that their severity will continue to be higher. Exceptionally, one small group of 20 patients (5% of the overall studied population) showed higher positive than negative symptom subscale scores. It would be interesting to understand the characteristics of such a group of patients in a larger database.
Lastly, the data suggests that the majority of patients generally experience one of the three distinct trajectory patterns of treatment-response course: 1) dramatic and sustained early improvement in both negative and positive symptoms, 2) mild and sustained improvement in negative and positive symptoms, or 3) no improvement in either negative or positive symptoms. Further research on the association between these treatment response trajectory subgroups and the underlying determinants (i.e., treatment choice, pathophysiology, and etiology indicators) may help inform clinical practice or aid the effort to develop targeted treatment of schizophrenia.
This study has a number of pertinent strengths and limitations that warrant discussion. First, this study is novel; the association between the two symptom domains evidenced at the individual patient level and these analyses have not been executed by others in this fashion. The field has been accustomed to seeing data at the population level (i.e., treatment group), an approach that does not necessarily imply that negative and positive symptom response is congruent. This study shows, however, at the individual patient level, the two symptom domains in fact do appear to display a commonality of response. Second, results from this study would apply in real-world practice settings. The data source is an open-label, 1-year, pragmatic trial. The study sample included a heterogeneous patient population with a variety of comorbid conditions, such as substance abuse; physicians were allowed to adjust dosage or switch medications according to their clinical discretion. Thus, the data reflect usual clinical care.
On the other hand, the study is limited as a post hoc, exploratory study. First, this study included patients who were mostly chronically ill; therefore, the findings may not be generalizable to patients in their early stage of the illness. Second, the treatment was open-label and drug switching was allowed. As such, we did not extend this study to analysis and inference on the specific drug effect, which is a clinically important question. Thirdly, we recognize that the definition and construct of negative and positive symptoms remains a subject of debate, though PANSS negative and positive subscale scores have been extensively used in schizophrenia clinical trials. Lastly, this study was not able to differentiate primary versus secondary negative symptoms due to the operational complexity of the concept. We expect future studies employing an advanced analytical method would help shed light on this issue. Overall, replication of the finding using independent data is necessary.
Acknowledgments
This study was supported by Eli Lilly and Company, Indianapolis, Indiana, USA. The authors wish to acknowledge the advice and insight of Dr. Shitij Kapur (King’s College London), the thoughtful comments of Dr. Sara Kollack-Walker (Eli Lilly), the writing review from Dr. Kory Schuh (Eli Lilly) and the editorial assistance from Ms. Teri Tucker (PharmaNet/i3, an inVentiv Health Company).
Local IRBs approved study F1D-US-HGGD
St. Davids Human Research Review Board, One Corporate Center Suite 300, Radnor PA 19087; IRB for Human Subjects Research for Baylor College, One Baylor Plaza, Houston TX 77030; McLean IRB- McLean Hospital, 115 Mill Street, Belmont, MA 02478; UMKC Adult Health Sciences IRB, Truman Medical Center 2301 Holmes, Kansas City MO 64104. Street St.Paul, MN 55101; State of Utah, 120 North 200 West #319 Salt Lake City, Utah 84103.
Competing interests
LC, JAJ, BJK, VS, and HA are employees and current stakeholders of Eli Lilly and Company. PS is a faculty member at the University of Cincinnati. TRM is employee of King’s College London. PS and TRM were not paid for their intellectual contributions to the study and have no competing interest to report.
Authors’ contributions
LC conceived the study. LC and JAJ contributed to the initial design and coordination. LC led the statistical analysis, wrote the initial draft of the manuscript, and coordinated the development of subsequent drafts. All authors participated in the analysis and interpretation of the data, as well as revising the manuscript for critically important intellectual content. In addition, all authors read and approved the final version of the manuscript.