ORIGINAL RESEARCH
published: 25 April 2022
doi: 10.3389/fpsyt.2022.816339
What Is the Minimum Clinically
Important Change in Negative
Symptoms of Schizophrenia? PANSS
Based Post-hoc Analyses of a Phase
III Clinical Trial
Pál Czobor 1*, Barbara Sebe 2 , Károly Acsai 2 , Ágota Barabássy 2 , István Laszlovszky 2 ,
György Németh 2 , Toshi A. Furukawa 3 and Stefan Leucht 4
Department of Psychiatry and Psychotherapy, Semmelweis University, Budapest, Hungary, 2 Global Medical Division,
Gedeon Richter Plc, Budapest, Hungary, 3 Department of Health Promotion and Human Behavior, Kyoto University School of
Public Health, Kyoto, Japan, 4 Department of Psychiatry and Psychotherapy, School of Medicine, Technical University of
Munich, Munich, Germany
1
Edited by:
Marijn Lijffijt,
Baylor College of Medicine,
United States
Reviewed by:
Qijing Bo,
Capital Medical University, China
Ravi Philip Rajkumar,
Jawaharlal Institute of Postgraduate
Medical Education and Research
(JIPMER), India
*Correspondence:
Pál Czobor
czobor.pal@med.semmelweis-univ.hu
Specialty section:
This article was submitted to
Psychopharmacology,
a section of the journal
Frontiers in Psychiatry
Received: 16 November 2021
Accepted: 23 February 2022
Published: 25 April 2022
Citation:
Czobor P, Sebe B, Acsai K,
Barabássy Á, Laszlovszky I,
Németh G, Furukawa TA and Leucht S
(2022) What Is the Minimum Clinically
Important Change in Negative
Symptoms of Schizophrenia? PANSS
Based Post-hoc Analyses of a Phase
III Clinical Trial.
Front. Psychiatry 13:816339.
doi: 10.3389/fpsyt.2022.816339
Frontiers in Psychiatry | www.frontiersin.org
Introduction: Minimum clinically important difference (MCID) is a measure that defines
the minimum amount of change in an objective score of a clinical test that must be
reached for that change to be clinically noticeable. We aimed to find the MCID for patients
with predominantly negative symptoms of schizophrenia at its earliest occurrence.
Methods: Data of a 26-week long, double-blind study with 454 patients [Positive
and Negative Symptom Scale Negative Factor Score (PANSS-FSNS) ≥24, Positive
and Negative Symptom Scale Positive Factor Score (PANSS-FSPS) ≤19] treated
with cariprazine 4.5 mg/d or risperidone 4 mg/d were analyzed. The Clinical Global
Impression—Improvement scale was used to quantify minimum improvement (CGI-I = 3)
and no clinical change (CGI-I = 4) on the PANSS-FSNS, and the MCID was estimated
with the following methods: as the mean PANSS-FSNS changes corresponding to the
first instance of minimal improvement across all visits (MCID1 ); as the difference between
the PANSS-FSNS change associated with the first instance and the PANSS-FSNS
changes associated with the last recorded clinically unchanged status across all visits
(MCID2 ); with the effect size approach (MCID3 ); as the Youden Index based cut-off value
between no clinical change and minimal improvement (MCID4 ); as the relative likelihood
of minimal improvement (MCID5 ).
Results: The MCID1 and MCID2 resulted in, respectively, a 3.8-point (18.5%) and a
1.5-point (7.3%) decrease from baseline severity on the PANSS-FSNS. Greater values
were required for the MCID at later evaluation times. The cut-off between minimum
improvement and no clinical change defined by the Youden Index was a−3-point (15%)
change in the PANSS-FSNS. The effect size approach indicated the 1.5-point difference
between minimally improved and unchanged patients to be a medium effect (ES = 0.6).
1
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
Conclusion: Applying different methods led to different results, ranging between 7.3
and 18.5% improvement from the baseline for the MCID at its earliest occurrence in
patients with predominantly negative symptoms of schizophrenia.
Keywords: minimum clinically important difference, negative symptoms, schizophrenia, MCID, clinical trial,
cariprazine
INTRODUCTION
with florid positive symptoms, a 19–28% decrease in the PANSS
total score was necessary to reach “minimal improvement” on
the CGI-Improvement scale (CGI-I) (17). To specifically assess
negative symptoms of schizophrenia, the PANSS factor score
for negative symptoms (PANSS-FSNS) has been widely used in
clinical trials (19). Leucht et al. found that minimal improvement
corresponded to a change from baseline in PANSS-FSNS scores
of −27 and −41%, as measured by the CGI-I and CGI-S (20).
Depending on the diagnostic criteria applied, negative
symptoms of schizophrenia are present in 5–60% of patients
with schizophrenia (21). These symptoms significantly affect a
patient’s quality of life as they limit functional recovery and are
associated with poor functional outcomes (22). Patients with
negative symptoms use more healthcare resources than patients
with positive symptoms (such as: primary care, emergency
care, laboratory and radiology tests, and prescription drugs),
and their treatment is usually not simple, causing a clinical
challenge for physicians. In contrast, positive symptoms have
remarkably little association with real-life functioning and are
easier to treat (23–25).
Finding the minimum clinically important change for
negative symptoms may help physicians in better assessing
treatment results as well as fostering the development of new
instruments. In this paper, we further analyse the MCID in
negative symptoms of schizophrenia, hypothesizing that as
patients get better by taking their medication, more extensive
changes in the PANSS-FSNS are needed to be considered
clinically relevant. The previous estimation by Leucht et al. took
all PANSS-FSNS changes associated with minimal improvement
into account, regardless of their timepoint (20) meaning that
the 27 and 41% improvement in the PANSS-FSNS associated
with minimal clinical changes represent weighted averages from
the first to the last instance of minimal improvement. Thus,
those percentages may have overestimated the MCID in patients
with predominantly negative symptoms of schizophrenia. In this
work, we focus on the first instance of the MCID in patients with
predominantly negative symptoms. Furthermore, to date, there is
no consensus on the best method to calculate the MCID, and we
apply both anchor- and distribution-based methods to get a more
comprehensive picture.
The efficacy of various treatment interventions can be assessed
in clinical trials by testing for statistical significance, yet a
statistically significant change on a symptom scale score does
not necessarily indicate a clinically relevant improvement (1).
Thus, various approaches have been developed across different
diseases to define the smallest beneficial effect for patients. One
of the first attempts to obtain the slightest empirically observed,
“clinically important” effects of intervention was published in
1989 by Jaeschke et al. (2), who defined the Minimum Clinically
Important Difference (MCID) as “the smallest difference in score
in the domain of interest which patients perceive as beneficial
and which would mandate, in the absence of troublesome side
effects and excessive cost, a change in the patient’s management.”
It is, therefore, a within-person, “before-after” change and
conceptually distinct from minimum between-group differences
that can be expected between two different treatments. The latter
would be called the smallest worthwhile difference (SWD) (3).
While the MCID is scale specific and assumes no substantial
adverse effects or costs, the SWD represents the ratio of benefits
and negative effects of two alternative treatments (4). Although
other reports described the MCID with very similar names, such
as the Minimal Clinically Important Difference (5, 6) Minimum
Important Difference (7) or Minimal Important Change (6),
the intended meaning is the same: MCID is a measure that
defines the minimum amount of change in the objective score
of a clinical test that must be reached for that change to be
clinically noticeable.
Calculations of the MCID are usually divided into two groups:
anchor- and distribution-based approaches. In anchor-based
methods, an objective outcome measure of change is linked to
a clinically meaningful external anchor, largely corresponding
to patient perception (3, 6, 7) or in case of impaired insight,
e.g., in dementia or schizophrenia, to clinical opinion (8–10).
Distribution-based methods use statistical properties of study
results, e.g., effect size or standard error of measurement, to
calibrate the MCID (11–15).
In schizophrenia, the Positive and Negative Syndrome
Scale (PANSS) measuring positive, negative, and general
psychopathology, is the gold-standard instrument for assessing
symptom severity (16). The clinical relevance of changes in
the PANSS total score has been previously evaluated using the
Clinical Global Impression (CGI) rating scale as an anchor
(17, 18). It largely varies across different patient populations.
Based on the CATIE study, where a very heterogeneous patient
population was analyzed, a change of a 34% decrease on the
PANSS total score was established as necessary to improve one
category on the CGI-Severity scale (CGI-S) (18). For patients
Frontiers in Psychiatry | www.frontiersin.org
METHODS
Study Design
Data were analyzed from a large randomized, doubleblind clinical trial treating patients with schizophrenia with
predominantly negative symptoms; the study’s methods and
results have been previously published (26). The study was
2
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
correlation analysis in the same population on which this work
is based (20).
According to the definition of the MCID, a within-subject
design was applied to estimate the difference between no clinical
change and minimal improvement. Our observation was that
PANSS-FSNS changes corresponding to minimal improvement
get higher and higher over time, and thus, in order to capture the
minimum clinically important difference, the MCID should be
calculated on the basis of PANSS-FSNS changes associated with
the earliest instance of minimal improvement. The MCID was
estimated with the following methods.
conducted at 66 study centers in 11 European countries from
May 2013 to November 2014. The clinical study was approved
by local independent ethics committees and was completed
following clinical practice guidelines by the International
Conference of Harmonization. All patients provided written
informed consent. The study consisted of a 4-week lead-in
period, a 26-week double-blind treatment period, and a 2-week
safety follow-up with a total of 14 visits. The primary aim of
the study was to assess the efficacy and safety of cariprazine
treatment vs. risperidone treatment in primary, persistent,
predominantly negative symptoms of schizophrenia. The
primary efficacy outcome was the change from baseline in the
PANSS-FSNS score at the end of the double-blind period (26
weeks). The secondary outcome parameter was change from
the baseline on the Personal and Social Performance (PSP)
scale. The study was well-controlled for secondary negative
symptoms, assessing depression (Calgary Depression Scale for
Schizophrenia), movement disorders (Simpson Angus Scale,
Abnormal Involuntary Movement Scale, and Barnes Akathisia
Rating Scale) and positive symptoms [PANSS factor score for
positive symptoms (PANSS-FSPS)] throughout the study. Safety
and tolerability were also assessed including adverse event
reports, laboratory assessment, vital signs, and EEG. Patients
were randomized to cariprazine 4.5 mg/d or risperidone 4 mg/d
(1:1) with 2 weeks of up-titration (26).
Anchor-Based Methods
• MCID1 : The mean PANSS-FSNS changes corresponding to
the first instance of minimal improvement (CGI-I = 3) across
all visits. In other words, MCID1 is based on the original
definition by Jaeschke et al. (2) for MCID, as it represents
a change from baseline in the original score units of the
PANSS-FSNS scale.
• MCID2 : The difference between the PANSS-FSNS change
associated with the first instance of improvement (i.e., CGII = 3) and the PANSS-FSNS changes associated with the last
recorded clinically unchanged status (CGI-I = 4) across all
visits. Thus, MCID2 represents the mean score method of
Redelmeier and Lorig (27) i.e., it shows the score difference
of the “slightly better” group minus that of the “about the
same” group (28). Accordingly, MCID2 is also expressed in the
original scale units.
Patients
Male and female patients with schizophrenia and predominantly
negative symptoms, between 18 and 65 years of age, who were
diagnosed with schizophrenia (as defined by DSM-IV-TR) and
an onset of illness ≥2 years prior, were included in the study.
Patients also needed to be in stable condition for at least 6
months with no hospitalisations. For study inclusion, patients
must have presented with predominantly negative symptoms for
≥6 months, a PANSS-FSNS ≥24, and a score ≥4 on at least 2 of
the following 3 PANSS negative symptom items: blunted affect,
passive/apathetic social withdrawal, and lack of spontaneity and
flow of conversation. The presence of a current DSM-IV-TR axis
I disorder other than schizophrenia, and an unstable condition
(such as a hospital admission in the previous 6 months), a PANSS
factor score for positive symptoms (PANSS-FSPS) >19, or a
PANSS-FSPS score increase of ≥25% during study lead-in were
grounds for study exclusion. Other clinical exclusion criteria
included substance abuse/dependence, treatment with clozapine
during the 12 months before the study, a history of non-response
to an adequate trial of risperidone for a psychotic episode, or
treatment with risperidone within 6 weeks of screening (26).
To test the statistical significance of the difference between
unchanged and minimally improved PANSS-FSNS values, we
applied a mixed model for repeated measures (MMRM) analysis
with improvement and visit as fixed effects. The subject was
used as random effect in the model. To avoid losing cases with
minimal improvement at the first visit after baseline (Week 1),
zero PANSS-FSNS change was imputed for the baseline visit
(where no improvement can be present by definition).
Distribution-Based Methods
• MCID3 : the effect size approach, based on the standardized
response mean difference, a widely used distribution-based
method to estimate the MCID, where MCID is the mean
difference between the last unchanged (CGI-I = 4) and the first
minimally improved (CGI-I = 3) PANSS-FSNS values divided
by the pooled standard deviation (SD) of the two. It formally
corresponds to the effect size calculation, making it possible to
interpret the MCID in terms of the effect sizes (28).
• MCID4 :
dichotomous
variable
indexing
minimal
improvement (not obtained = 0, obtained = 1) based
on the cut-off value between no clinical change (CGI-I = 4)
and minimal improvement (CGI = 3). A logistic regression
model, with CGI-I as dependent and PANSS-FSNS as
independent variables as well as baseline PANSS-FSNS as a
covariate, was fitted to the data. To examine the accuracy of
predicting improvement based on the PANSS-FSNS change,
a receiver operating characteristics (ROC) curve was derived.
The strategy used in the ROC analysis was to maximize both
MCID Analyses
The clinician-rated CGI-I scale was used to quantify minimum
improvement (CGI-I = 3) and no clinical change (CGI-I = 4)
on the PANSS-FSNS. To demonstrate any meaningful results
by linking an objective scale measuring symptom severity to
a subjective scale that estimates the clinical state correlation
between the two scales must be demonstrated in the target
population. This was done by Leucht et al., who performed the
Frontiers in Psychiatry | www.frontiersin.org
3
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
FIGURE 1 | Change from baseline in PANSS-FSNS (Positive- and Negative Syndrome Scale—Factor Score for Negative Symptoms) as a function of minimal change
vs. no change.
groups; the mean factor score on the PANSS-FSNS at baseline
was 27.6, with points decreasing to 19.0 points at week 26.
sensitivity and specificity, and the MCID was estimated as a
cut-off value corresponding to the maximal Youden’s index
(29, 30).
• MCID5 as expressed in terms of ratio of odds values (p/1p) of being in improved vs. unchanged state at a certain
degree of FSNS decrease. This estimation is based on
the logistic regression model as above and expresses the
strength of the predictive power of unity change in FSNS for
clinical improvement.
By Visit Analyses
Minimal improvement on the clinical global impression scale
(CGI-I = 3) was associated with PANSS-FSNS changes ranging
from −2.5 points (Week 1) to −7.1 points (Week 26), consistent
with our hypothesis of the MCID to be smaller at its earliest
occurrence (Figure 1).
Quantifying the MCID by Anchor-Based
Methods
RESULTS
MCID1 and 2
The primary results of the study were previously published
(26): change from baseline to week 26 in PANSS-FSNS was
significantly greater with cariprazine than with risperidone [least
squares mean difference (LSMD) −1.46, 95% CI −2.39 to −0.53;
p = 0.0022; effect size = 0.31]. Also, for the secondary efficacy
parameter, least squares mean change from baseline to endpoint
in PSP total score, was greater for cariprazine than risperidone
(LSMD 4.63, 2.71–6.56; p < 0.0001; statistical effect size = 0.48).
In the parameters controlling for secondary negative symptoms,
least squares mean changes from baseline for PANSS-FSPS, CDSS
total score, and movement scales were small and similar for
cariprazine and risperidone.
A total of 454 patients from the intent-to-treat population
with at least one post-baseline PANSS assessment were pooled for
this analysis from both the cariprazine and risperidone treatment
Frontiers in Psychiatry | www.frontiersin.org
The mean PANSS-FSNS change from baseline corresponding
to the first occurrence of minimal improvement (CGI-I = 3)
and the last recorded unchanged status across all visits were
−3.8 and −2.3 points, respectively. The statistical analysis of
the two arithmetic PANSS-FSNS means showed a significant
difference (Table 1).
Quantifying the MCID by
Distribution-Based Methods
MCID3
Based on the observed PANSS-FSNS changes the standardized
effect size was determined according to the following formula:
(PANSS-FSNS change from baseline at the first instance minimal
improvement) − (Mean PANSS-FSNS Change from baseline at
4
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
TABLE 1 | Anchor based calculations of the MCID.
Visit
Overall
Mean PANSS-FSNS
change from baseline (n) at the
1st instance of minimal
improvement [1]
Mean PANSS-FSNS
change from baseline (n) at the
last recorded unchanged status
[2]
MCID1
[1]
MCID2
[1]-[2]
SD of
[1] and [2]
LSMD
(95% CI)
P-value
−3.8
(365)
−2.3
−3.8
(18.5%)
−1.5
(7.3%)
2.5
1.7
(1.1, 2.3)
<0.0001
PANSS-FSNS, Positive and Negative Syndrome Scale-Factor Score for Negative Symptoms; n, number of events; CGI-I, Clinical Global Impression—Improvement; MCID1 and MCID2 ,
Minimum clinically important differences according to the definitions in the text; SD, Standard deviation; LSMD, Least-square mean difference between [1] and [2], as estimated by
MMRM, with corresponding 95% confidence limits. Please also note that for the computation of % changes from baseline the value of 7 was subtracted from the observed baseline
severity (i.e., 27.6) since the minimum value of the PANSS-FSNS factor is 7 (i.e., the symptoms on all seven constituting items of the factor are rated as “Absent”).
FIGURE 2 | Receiver Operating Characteristic (ROC) curve: predictive accuracy of the PANSS-FSNS (Positive and Negative Syndrome Scale Factor Score for
Negative Symptoms) scale for differentiating minimally improved vs. clinically unchanged status. The values on the vertical and horizontal axis, respectively, depict the
sensitivity (“true positive rate”) and 1− specificity (“false positive rate”) values for the differentiation as a function of change from baseline in the PANSS-FSNS scale.
The leftmost part of the ROC curve represents the highest empirically observed improvements in the sample as compared to baseline while the rightmost part
represents no improvement (or even deterioration). Please note that the ROC curve for differentiating the minimally improved from the clinically unchanged status
based on the PANSS-FSNS (ROC model, depicted in blue in the figure) significantly outperforms the random classification (ROC1 model, in red), with an area under
the curve (AUC, labeled as “Area”) value of 0.7232 vs. 0.5000 (p < 0.0001).
identified a −3 point decrease from the PANSS-FSNS at baseline
as the cut-off value most effectively differentiated between true
positive and false positive classifications, i.e., minimally improved
and unchanged statuses (Figure 3).
the last recorded unchanged status)/pooled SD. Our computation
resulted in a standardized effect size for the improvement with a
value of 0.6 [i.e., −3.8 – (−2.3)/2.5 = −1.5/ 2.5 = −0.6].
MCID4
The ROC curve indicated statistically robust predictive values of
PANSS-FSNS changes, as the model fitted to our data (Model)
was highly significantly different from the reference line (ROC1)
(Figure 2). Based on the maximal Youden’s index method (29) we
Frontiers in Psychiatry | www.frontiersin.org
MCID5 as Odds Ratio
The odds ratio (OR) indicates the strength of association between
the decrease in FSNS and the improved state based on CGI-I. For
the −1.7 shift in FSNS obtained above as estimated MCID based
5
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
FIGURE 3 | Cut-off values (Youden’s indices) for predicting improvement from no clinical change to minimal improvement. To differentiate minimal improvement from
no clinical change, Youden’s J indices were computed at different cut-off points based on the PANSS-FSNS change. The sensitivity (vertical axis) and 1- specificity
(horizontal axis) values for the differentiation of minimal improvement from no clinical change are depicted in the figure for various values of change from baseline in the
PANSS-FSNS (labeled as “Cutoff”). Please note that the Youden’s J index, which shows the efficiency of differentiation based on the combination of sensitivity and
specificity, first increases then decreases with increasingly greater improvements (i.e., with greater negative values) as compared to baseline. The highest value of the
Youden’s J Index is reached at the cut-off value of −3 (i.e., at a 3 point reduction of symptom severity from baseline in the PANSS-FSNS), which identifies the optimal
change value that maximizes sensitivity and specificity simultaneously.
time effect of improvement is considered as well. The current
analyses estimated the MCID with several methods by looking
at it at its earliest occurrence.
It is important to note that the five different approaches that
we adopted in the current investigation to characterize MCID
complement each other and delineate the MCID from various
vantage points. The distribution-free approaches characterize
change over time in terms of the original units on the scale of
interest (PANSS-FSNS), either as a baseline-end point difference
associated with the minimal improvement on the CGI-I (score =
3) (MCID1) or the difference between unchanged and minimally
improved PANSS-FSNS values (MCID2). The first one of the
distribution-based methods that we adopted (MCID3) expresses
the difference between unchanged and minimally improved
PANSS-FSNS values (measured in the original scale units) in
terms of standard deviation (statistical) units. The additional
on the Least Squares Mean Difference (LSMD), the OR is 1.86
(95% CI 1.62, 2.13; p < 0.05) favoring CGI-I = 3 vs. 4, The logistic
regression analysis described above yielded an estimated OR of
2.07 (95% CI = 1.76 – 2.44; p < 0.05) for a 2-point decrease in the
PANSS-FSNS factor score during the study. Thus, this analysis
indicates that such an improvement (i.e., 2 points) in the PANSSFSNS factor score more than doubles the likelihood of achieving
minimal clinical improvement in the study.
DISCUSSION
Previous work has established the minimal improvement (CGI-I
= 3) as being associated with a 27% decrease in the PANSSFSNS for patients with predominantly negative symptoms of
schizophrenia (20) which may still be an overestimation if the
Frontiers in Psychiatry | www.frontiersin.org
6
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
two distribution-based approaches employ logistic regression
modeling in order to predict the minimally improved status on
the basis of the PANSS-FSNS change over time. They express
MCID either as measures of the ROC Curve (AUC, Youden’s
index in case of MCID4) or an OR (MCID5).
Applying these approaches, we confirmed that over time,
more and more prominent symptom changes were needed
to achieve minimal clinical improvement. In view of this
finding, we conclude that the absolute minimum clinically
meaningful difference should be considered at the earliest
instance, since this approach provides the highest assay
sensitivity to detect clinically important changes of symptom
severity over time (i.e., it allows to capture the lowest symptom
change threshold for minimal clinical improvement). We note
that the PANSS-FSNS of the patients whose clinical status
remained the same slightly decreased as well, although to
a much smaller extent. This slight symptom reduction of
the clinically not improving patients may be attributable to
unspecific changes, such as the regression to the mean effect,
a phenomenon often seen when applying strict inclusion
criteria for the patients regarding their symptom severity
(31). The presence of such unspecific changes, which evolve
gradually with time, may make it more difficult to establish a
clinically important difference at later times in a trial, thereby
providing additional rationale for focusing on the earlier time
points. Additionally, because improvement is rated against
the initial baseline, anchoring changes of symptom severity
to clinical improvement can be more and more difficult and
accompanied by greater variation as the patients progress over
time during the study.
Overall, applying various underlying concepts to calculate
the MCID, different methods led to different estimates with
respect to the PANSS-FSNS change that need to be considered
as minimally clinically significant. Our estimates from the
anchor-based analyses, ranging from 7.3 to 18.5% for patients
with predominant negative symptoms of schizophrenia, were
below the estimates reported from studies that relied on more
heterogeneous patient populations and did not take the time
effect into consideration. However, it is important to bear in
mind that the distribution-based statistical approaches showed
a marked separation (MCID3 , i.e., statistical effect size in terms
of standardized mean difference = 0.6); and highly significant
predictive power for the PANSS-FSNS scale for differentiating
Minimally Improved from Clinically Unchanged status in terms
of ROC measures (MCID4 ; e.g., AUC = 0.7232) and the odds
ratio (MCID5 ; OR = 2.07).
An important limitation of this investigation is that,
although experienced raters met the training requirements and
qualification criteria set forth before the rater training and
administered the instruments, potential rating bias might have
occurred. For example, the increasing PANSS-FSNS changes
over time could be attributed to a possible oversight by
comparing patients’ clinical statuses to the previous visit instead
of the baseline status (32). Furthermore, one could also argue
that very early improvements might not have been real drug
effects due to the two drugs’ different pharmacokinetics and
Frontiers in Psychiatry | www.frontiersin.org
the onset of effect. Nevertheless, the MCID, by definition, is
not about treatment-effect and could be driven by complex
factors. Finally, since the CGI ratings were performed by
the clinician, a further limitation of the study is that
the minimally clinically important difference in the current
investigation was evaluated from the clinician’s perspective,
not from the patient’s. Patient-related outcomes were not
assessed to determine the minimally clinically important change.
However, we note that one study which examined patientand clinician-rated CGI assessments simultaneously in patients
with schizophrenia found only slight differences between the
two approaches (18). Nonetheless, a specifically designed study
is clearly needed to investigate this issue further. Further, an
additional limitation derives from the use of only one particular
scale. In this study we adopted one of the benchmark scales
for negative symptoms of schizophrenia, the PANSS-FSNS.
Assessing negative symptoms with different scales (such as
the SANS, BNSS, CAINS, etc.) would yield different values
as minimally important change, as scoring on these scales is
different. Consequently, further research is needed to identify
minimally clinically important changes on different negative
symptom scales.
Negative symptoms, that are not secondary to positive ones,
are known for being less responsive to antipsychotic therapy (33–
35). Their presence often challenges the therapeutic strategy and
makes clinicians switch the patients’ medication. Our findings
may help clinicians and drug developers have a more precise
idea about improving predominant negative symptoms in clinical
decision-making, or in terms of designing trials for patients with
predominant negative symptoms.
CONCLUSION
Applying different methods lead to different results, ranging
between 7.3 and 18.5% improvement from baseline for the MCID
at its earliest occurrence in patients with predominant negative
symptoms of schizophrenia, suggesting even lower thresholds
than previously thought.
DATA AVAILABILITY STATEMENT
The datasets presented in this article are not readily available
because they are part of a bigger randomized clinical study
dataset, and are owned by the company Gedeon Richter. The
datasets have been provided to authors to perform the presented
analyses only. While the full datasets can therefore not be shared,
all statistical analyses and data outputs generated for the present
study/publication can be requested by the authors.
ETHICS STATEMENT
The clinical study protocol was approved by nine central
and 37 local independent Ethics Committees in relation
to the 66 sites that recruited at least one patient; the
study was done in accordance with good clinical practice
7
April 2022 | Volume 13 | Article 816339
Czobor et al.
Clinically Important Difference, Negative Symptoms
PC, SL, and TAF. The statistical plan was outlined by PC and
KA. KA conducted the statistical analyses under the supervision
of PC. Software preparation was done by KA and PC. KA,
PC, BS, and IL had a leading role in the preparation of the
figures and table. Project administration was provided by ÁB,
BS, IL, and GN. PC and KA interpreted the statistical outputs.
Conceptual supervision for manuscript preparation and writing
was provided SL, PC, and TAF. BS, IL, ÁB, and PC performed the
writing. All authors contributed to the article and approved the
submitted version.
guidelines and the principles of the International Conference on
Harmonization. All patients provided written informed consent.
The patients/participants provided their written informed
consent to participate in this study.
AUTHOR CONTRIBUTIONS
SL, PC, ÁB, IL, and GN developed the concept of the current
investigation. Methodology contribution was provided by KA,
REFERENCES
15. Muller K, Cohen J. Statistical power analysis for the behavioral sciences.
Technometrics. (1989) 31:499–500. doi: 10.1080/00401706.1989.10488618
16. Suzuki T. Which rating scales are regarded as “the standard” in clinical trials
for schizophrenia? a critical review. Psychopharmacol Bull. (2011) 44:18–31.
17. Leucht S, Kane JM, Kissling W, Hamann J, Etschel E, Engel
RR. What does the PANSS mean? Schizophr Res. (2005)
79:231–8. doi: 10.1016/j.schres.2005.04.008
18. Hermes EDA, Sokoloff D, Stroup TS, Rosenheck RA. Minimum clinically
important difference in the positive and negative syndrome scale with data
from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE).
J Clin Psychiatry. (2012) 73:526–32. doi: 10.4088/JCP.11m07162
19. Marder SR, Daniel DG, Alphs L, Awad AG, Keefe RSE.
Methodological issues in negative symptom trials. Schizophr Bull. (2011)
37:250–4. doi: 10.1093/schbul/sbq161
20. Leucht S, Barabássy Á, Laszlovszky I, Szatmári B, Acsai K,
Szalai E, et al. Linking PANSS negative symptom scores with
the Clinical Global Impressions Scale: understanding negative
symptom scores in schizophrenia. Neuropsychopharmacology. (2019)
44:1589–96. doi: 10.1038/s41386-019-0363-2
21. Rabinowitz J, Werbeloff N, Caers I, Mandel FS, Stauffer V, Menard F, et al.
Negative symptoms in schizophrenia - the remarkable impact of inclusion
definitions in clinical trials and their consequences. Schizophr Res. (2013)
150:334–8. doi: 10.1016/j.schres.2013.06.023
22. Correll CU, Schooler NR. Negative symptoms in schizophrenia: a review and
clinical guide for recognition, assessment, and treatment. Neuropsychiatr Dis
Treat. (2020) 16:519–34. doi: 10.2147/NDT.S225643
23. Harvey PD, Bellack AS. Toward a terminology for functional recovery in
schizophrenia: is functional remission a viable concept? Schizophr Bull. (2009)
35:300–6. doi: 10.1093/schbul/sbn171
24. Harvey PD, Strassnig M. Predicting the severity of everyday functional
disability in people with schizophrenia: cognitive deficits, functional
capacity, symptoms, and health status. World Psychiatry. (2012) 11:73–
9. doi: 10.1016/j.wpsyc.2012.05.004
25. Galderisi S, Rossi A, Rocca P, Bertolino A, Mucci A, Bucci P, et al.
The influence of illness-related variables, personal resources and contextrelated factors on real-life functioning of people with schizophrenia. World
Psychiatry. (2014) 13:275–87. doi: 10.1002/wps.20167
26. Németh G, Laszlovszky I, Czobor P, Szalai E, Szatmári B, Harsányi
J, et al. Cariprazine versus risperidone monotherapy for treatment
of predominant negative symptoms in patients with schizophrenia:
a randomised, double-blind, controlled trial. Lancet. (2017)
389:1103–13. doi: 10.1016/S0140-6736(17)30060-0
27. Redelmeier DA, Lorig K. Assessing the clinical importance of
symptomatic improvements: an illustration in rheumatology. Arch
Intern Med. (1993) 153:1337–42. doi: 10.1001/archinte.1993.004101100
45008
28. Angst F, Aeschlimann A, Angst J. The minimal clinically important difference
raised the significance of outcome effects above the statistical level, with
methodological implications for future studies. J Clin Epidemiol. (2017)
82:128–36. doi: 10.1016/j.jclinepi.2016.11.016
29. Schisterman EF, Perkins NJ, Liu A, Bondell H. Optimal cut-point and its
corresponding Youden index to discriminate individuals using pooled blood
1. Correll CU, Kishimoto T, Nielsen J, Kane JM. Quantifying clinical
relevance in the treatment of schizophrenia. Clin Ther. (2011) 33:B16–
39. doi: 10.1016/j.clinthera.2011.11.016
2. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascertaining
the minimal clinically important difference. Control Clin Trials. (1989)
10:407–15. doi: 10.1016/0197-2456(89)90005-6
3. Ferreira ML, Herbert RD, Ferreira PH, Latimer J, Ostelo RW, Nascimento DP,
et al. A critical review of methods used to determine the smallest worthwhile
effect of interventions for low back pain. J Clin Epidemiol. (2012) 65:253–
61. doi: 10.1016/j.jclinepi.2011.06.018
4. Furukawa TA. Measuring clinical importance in a trial of
interventions for mixed urinary incontinence. J Am Med Assoc. (2020)
323:479. doi: 10.1001/jama.2019.19730
5. Beurskens AJHM, De Vet HCW, Köke AJA. Responsiveness of functional
status in low back pain: a comparison of different instruments. Pain. (1996)
65:71–6. doi: 10.1016/0304-3959(95)00149-2
6. Wright A, Hannon J, Hegedus EJ, Kavchak AE. Clinimetrics corner: a closer
look at the minimal clinically important difference (MCID). J Man Manip
Ther. (2012) 20:160–6. doi: 10.1179/2042618612Y.0000000001
7. Cella D, Bullinger M, Scott C, Barofsky I, Aaronson N, Berzon R, et al.
Group vs individual approaches to understanding the clinical significance
of differences or changes in quality of life. Mayo Clin Proc. (2002) 77:384–
92. doi: 10.4065/77.4.384
8. Falissard B, Sapin C, Loze JY, Landsberg W, Hansen K. Defining the
minimal clinically important difference (MCID) of the Heinrichs-carpenter
quality of life scale (QLS). Int J Methods Psychiatr Res. (2016) 25:101–
11. doi: 10.1002/mpr.1483
9. Andrews JS, Desai U, Kirson NY, Zichlin ML, Ball DE, Matthews BR.
Disease severity and minimal clinically important differences in clinical
outcome assessments for Alzheimer’s disease clinical trials. Alzheimer’s
Dement Transl Res Clin Interv. (2019) 5:354–63. doi: 10.1016/j.trci.2019.
06.005
10. Amri I, Millier A, Toumi M. Minimum clinically important difference
in the global assessment functioning. Value Health. (2014) 17:A765–
6. doi: 10.1016/j.jval.2014.08.285
11. Jacobson NS, Roberts LJ, Berns SB, McGlinchey JB. Methods for defining
and determining the clinical significance of treatment effects: description,
application, and alternatives. J Consult Clin Psychol. (1999) 67:300–
7. doi: 10.1037/0022-006X.67.3.300
12. Jacobson NS, Truax P. Clinical significance: a statistical approach to defining
meaningful change in psychotherapy research. J Consult Clin Psychol. (1991)
59:12–9. doi: 10.1037/0022-006X.59.1.12
13. de Vet HC, Terwee CB, Ostelo RW, Beckerman H, Knol DL, Bouter
LM. Minimal changes in health status questionnaires: distinction between
minimally detectable change and minimally important change. Health Qual
Life Outcomes. (2006) 4:54. doi: 10.1186/1477-7525-4-54
14. Wyrwich KW, Nienaber NA, Tierney WM, Wolinsky FD. Linking
clinical relevance and statistical significance in evaluating intra-individual
changes in health-related quality of life. Med Care. (1999) 37:469–
78. doi: 10.1097/00005650-199905000-00006
Frontiers in Psychiatry | www.frontiersin.org
8
April 2022 | Volume 13 | Article 816339
Czobor et al.
30.
31.
32.
33.
34.
35.
Clinically Important Difference, Negative Symptoms
samples. Epidemiology. (2005) 16:73–81. doi: 10.1097/01.ede.0000147512.81
966.ba
Powers DMW. Evaluation: From Precision, Recall and F-Factor. Tech Rep
SEI-07-001. Adelaide, SA: Flinders University of South Australia (2007).
Bland JM, Altman DG. Regression towards the mean. BMJ. (1994)
308:1499. doi: 10.1136/bmj.308.6942.1499
Busner J, Targum SD. The clinical global impressions scale: applying a research
tool in clinical practice. Psychiatry. (2007) 4:28–37.
Fusar-Poli P, Papanastasiou E, Stahl D, Rocchetti M, Carpenter W, Shergill
S, et al. Treatments of negative symptoms in schizophrenia: meta-analysis
of 168 randomized placebo-controlled trials. Schizophr Bull. (2015) 41:892–
9. doi: 10.1093/schbul/sbu170
Aleman A, Lincoln TM, Bruggeman R, Melle I, Arends J, Arango C,
et al. Treatment of negative symptoms: Where do we stand, and where
do we go? Schizophr Res. (2017) 186:55–62. doi: 10.1016/j.schres.2016.
05.015
Barnes TR. Evidence-based guidelines for the pharmacological
treatment of schizophrenia: recommendations from the British
Association for Psychopharmacology. J Psychopharmacol. (2011)
34:3–78. doi: 10.1177/0269881119889296
patent 2020-548587 concerning smartphone CBT apps pending, and intellectual
properties for Kokoro-app licensed to Mitsubishi-Tanabe. SL reports honoraria
as a consultant/advisor and/or for lectures from Angelini, Böhringer Ingelheim,
Geodon Richter, Janssen, Johnson & Johnson, Lundbeck, LTS Lohmann, MSD,
Otsuka, Recordati, SanofiAventis, Sandoz, Sunovion, TEVA, Eisai, Rovi, and
Medichem. GN and IL have issued patents for cariprazine.
This study was sponsored by Gedeon Richter Plc. Gedeon Richter was involved
in the study design, collection (via contracted clinical investigator sites), analysis,
and interpretation of data and decided to submit it for publication.
Publisher’s Note: All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated organizations, or those of
the publisher, the editors and the reviewers. Any product that may be evaluated in
this article, or claim that may be made by its manufacturer, is not guaranteed or
endorsed by the publisher.
Copyright © 2022 Czobor, Sebe, Acsai, Barabássy, Laszlovszky, Németh, Furukawa
and Leucht. This is an open-access article distributed under the terms of the Creative
Commons Attribution License (CC BY). The use, distribution or reproduction in
other forums is permitted, provided the original author(s) and the copyright owner(s)
are credited and that the original publication in this journal is cited, in accordance
with accepted academic practice. No use, distribution or reproduction is permitted
which does not comply with these terms.
Conflict of Interest: PC, BS, KA, ÁB, IL, and GN reports personal fees from
Gedeon Richter Plc., outside the submitted work. TAF reports grants and
personal fees from Mitsubishi-Tanabe and from Shionogi, personal fees from
MSD and from SONY, outside the submitted work. In addition, TAF has a
Frontiers in Psychiatry | www.frontiersin.org
9
April 2022 | Volume 13 | Article 816339