Keywords antipsychotics - prescribing - schizophrenia - pharmacoepidemiology
Introduction
Antipsychotic (AP) long-term prescription is one of the cornerstones of treatment
for
psychotic disorders. Patients on AP medications do better overall; early cessation
often results in higher relapse rates, hospitalization, and risk of death and
violence [1 ]
[2 ]
[3 ]. Originally for psychosis, the use of APs has expanded to address
psychotic symptoms in various mental disorders, behavioral disturbances in dementia,
pervasive developmental disorders, and as adjuncts to antidepressants or mood
stabilizers in major affective disorders [4 ].
First-generation antipsychotics (FGAs) are effective against psychotic symptoms but
often cause significant neuroleptic side effects [5 ]
[6 ]. Second-generation APs (SGAs) were introduced as more efficacious and
better tolerated but can lead to serious metabolic side effects like weight gain and
diabetes [7 ]. A meta-analyses by Leucht et
al. highlighted efficacy differences among APs, with the SGAs clozapine,
amisulpride, olanzapine, and risperidone showing superior performance, despite their
distinct side-effect profiles [8 ]
[9 ]. Similarly, Huhn et al. found clozapine,
amisulpride, zotepine, olanzapine, and risperidone to be superior to the other APs
in terms of efficacy measures [10 ]. In
terms of effectiveness, the meta-analyses by Kishimoto et al. showed that SGAs,
particularly clozapine, olanzapine, and risperidone, were superior in reducing
all-cause discontinuation compared to FGAs and other SGAs in direct comparisons
[11 ]
[12 ].
Treatment discontinuation is a broad measure of effectiveness, safety, and
tolerability, often linked to efficacy issues, adverse events, or patient
unwillingness to continue treatment. This pragmatic endpoint is used as the primary
outcome in many non-industry-funded RCTs evaluating APs in schizophrenia [13 ]
[14 ]
[15 ]
[16 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]
[22 ]. Independent real-world trials have
shown that older APs outperform expectations set by earlier efficacy trials and
meta-analyses, suggesting that while newer APs are an advancement, they do not mark
a revolutionary breakthrough in effectiveness [23 ]. Industry-sponsored RCTs and related meta-analyses, often influenced
by financial conflicts of interest, face scrutiny over potential biases and
methodological issues, casting doubt on their reliability and impartiality [24 ]
[25 ]
[26 ]. These caveats
highlight the utility of the integration of both experimental and observational
evidence to inform treatment decision-making [27 ]
[28 ].
Pharmaco-epidemiological data offers a potent lens to substantiate meta-analytic
findings on drug efficacy through real-world effectiveness data. Our previous
investigation into reboxetine, extensively prescribed in Europe for depression,
underscores this point. Eyding et al. (2010) determined it to be ineffective and
potentially harmful, findings that our pharmaco-epidemiological research supported,
highlighting low treatment retention [29 ]
[30 ]. This case illustrates
how pharmaco-epidemiological data can not only reinforce but also anticipate issues
with drug responses before a comprehensive experimental dataset is available.
Our study aims to compare oral AP dispensing trends in a large population sample from
2000 to 2021, with a primary focus on evidencing the superiority of SGAs over FGAs,
in line with prior experimental research [8 ]
[9 ]
[10 ]
[11 ]
[12 ]. To better
contextualize AP persistence within broader epidemiological shifts, we initially
examined 2000-2021 general epidemiological trends in AP use. We then targeted our
primary objective by directly comparing time-to-treatment discontinuation (TTD) of
FGAs and SGAs and conducted detailed comparisons between individual drugs.
Materials and Methods
Administrative healthcare database variables
This study utilized the computerized Healthcare Utilization database of Lombardy
(Lombardy DB), part of an automated system implemented across Italian regions
for healthcare service management. The Lombardy DB systematically collects
information on residents who are beneficiaries of the National Health Service
(SSN), encompassing socio-demographic details (such as gender, age, education
level, etc.), outpatient drug prescriptions, hospital discharge diagnoses from
both public and private facilities, specialist visits, and diagnostic tests
covered fully or partially by the SSN. The Lombardy system of automated
Healthcare Utilization databases comprises: 1. an archive of all residents
receiving SSN assistance, virtually the entire resident population, with
demographic and administrative data; 2. an archive including all certifications
of chronic diseases for exemption from co-payment; 3. an archive of all hospital
discharge forms from public or private hospitals, detailing all diagnoses
related to hospitalization; 4. an archive of all outpatient drug prescriptions
reimbursable by the SSN. Through a record-linkage procedure facilitated by a
unique individual identification code, which is used across all databases for
each SSN beneficiary and automatically anonymized for privacy, it is possible to
interconnect these regional HCU databases. This linkage enables the tracking of
complete healthcare journey of each patient. Data were registered and stored in
compliance with both Italian and European General Data Protection Regulations.
The Lombardy DB contains information about the number of dispensed Defined Daily
Doses (DDD) of all outpatient prescription-based medications. DDDs are assigned
and reviewed by researchers of the World Health Organization Collaborating
Centre of Drug Statistics Methodology [31 ]. Importantly, registered dispensations correspond to a variable
number of drugʼs boxes/packages, ranging from one to two, based on the specific
AP and prescribing practices. Additionally, for each dispensation, both the
number of boxes/packages and the number of DDDs generated are recorded,
providing a comprehensive view of the medication dispensed [31 ].
Lombardy DB is routinely updated for administrative reasons. Briefly, for a drug
to be reimbursed by the Italian National Health Service (Servizio Sanitario
Nazionale, SSN), the patients need a prescription from their general
practitioner (GP) or a SSN specialist and then get the medicines free of charge
from retail pharmacies. Private psychiatrists or neurologists usually send their
recommendations for drug treatment to GPs, who then complete the prescription
forms. Each local pharmacy provides these prescriptions to the Regional Health
Authority to get reimbursed; incomplete or incorrect reporting leads to legal
consequences. The Regional Health Authority electronically stores these
prescriptions in the Lombardy DB. Information on age and gender is available,
but diagnoses related to prescriptions are not, as only those linked to hospital
discharge records, as previously mentioned, are accessible. The database
collects only community prescriptions and no information on drugs prescribed
during admission to hospital or stay in nursing homes is available.
AP use prevalence and incidence were calculated as percentages of the resident
population, considering those with three or more AP dispensations yearly for
prevalence and the subgroup of those who did not have any prescription for the
same AP in the preceding year for incidence. All measures of oral APs dispensing
were assessed based on the Anatomical Therapeutic Chemical classification (ATC)
code [31 ] (ATC code: N05, Lithium
excluded). Data on patients’ sex, age, place of residence, date of purchase,
drug information were extracted. For the primary analysis, the following two
groups were compared: 1) SGAs (amisulpride-N05AL05, clozapine-N05AH02,
olanzapine-N05AH03, risperidone/paliperidone-N05AX08/13, aripiprazole-N05AX12);
2) FGAs (haloperidol-N05AD01, other FGAs:
N05AA/N05AB/N05AC/N05AD/N05AF/N05AG/N05AL).
Study setting and population
Prescriptions of APs dispensed to the adult population of four provinces in the
Lombardy region, Northern Italy, were analyzed on the basis of a
population-based database of dispensing records from January 1, 2000 to December
31, 2021. The study provinces directly align with the catchment areas of four
Local Health Units (LHUs): the Metropolitan City of Milan, Sondrio, Brescia, and
Pavia. Lombardy is the largest Italian region, with a population of about 10
million inhabitants. Adults living in the study areas remained stable in number
across the years and totaled around 5 million in both 2000 and 2021, accounting
for 50% of the adult population of the region at the end of the study
period.
By virtue of the universal coverage provided by the SSN, all residents aged 18–65
years were eligible for inclusion in the study, thereby encompassing the entire
resident population within this age range. For each AP analyzed annually, the
study included: (a) all individuals initiating AP treatment, defined as those
who had not received any AP prescriptions in the year preceding the start of the
observation period and (b) all individuals who were dispensed the specified AP
at least three times, were identified as continuous users.
Personalized daily dosage (PDD) and time-to-treatment discontinuation
Analysis
For the analysis, we identified continuous AP treatment periods by linking the
DDDs from the first two dispensations of each prescription event in the dataset.
This method aggregates the DDDs of each dispensed medication package within
these initial dispensations to form a continuous timeline of AP use. If the
second dispensing occurred at or before the end of the initial dispensationʼs
DDDs, the two dispensations were directly linked ([Fig. 1 ], Case A). However, in cases
where gaps between dispensations were observed, we addressed these by utilizing
two previously validated corrections: the ‘one-sixth ruleʼ and a 30-day time
margin ([Fig. 1 ], Cases B and C,
respectively) [32 ]
[33 ]. Under these conditions,
dispensations were linked into continuous treatment periods if either the DDDs
from the first dispensation, plus a period equal to one-sixth of the total DDDs
dispensed, overlapped with the date of the second dispensation ([Fig. 1 ], Case B), or if the second
dispensation occurred within a 30-day interval following the first ([Fig. 1 ], Case C). Prescription events
that did not meet the study criteria due to excessive DDD gaps between the first
two dispensations were excluded from the analysis, as it was not feasible to
calculate PDDs for these instances.
Fig. 1 The one-sixth and the one-month’s gap rules for linkage of
APs’ dispensations into treatments’ periods.
To calculate the PDD, we aggregated all DDDs received within the initial
treatment periods determined by linking the first two AP dispensations for each
patient [34 ]. This total was then
divided by the duration of these linked treatment periods (refer to [Fig. 1 ], Cases A, B, and C.). The PDDs
were then applied to assess time to treatment discontinuation (TTD) for each
participant, utilizing the same ‘one-sixth ruleʼ and a 30-day margin to manage
gaps between dispensations and ensure a continuous treatment timeline. To
prevent overestimation of treatment durations and dose accumulation, PDDs from
preceding dispensations were adjusted and reset at the date of each new
dispensing event.
For each drug, our analysis included only the first continuous prescribing event
for each individual. Thus, a subject could appear multiple times in the analysis
only if they had other prescribing events with different drugs. Additionally,
only the first prescribing event for the same individual within the same class
(SGA or FGA) was included.
For additional details on the studyʼs population selection, treatment duration,
and discontinuation algorithm, see Supplementary Document S1 .
The primary analysis of TTD for the selected oral APs was conducted over a
3-month period. This duration has been chosen to enable the identification of
significant differences in treatment outcomes, should they exist. This
relatively brief follow-up period reduces the likelihood of inaccurately
classifying cases as treatment failures—which could occur if patients
discontinue their medication upon completing their prescribed treatment or due
to sufficient symptom improvement. Such an approach promotes a more precise and
meaningful evaluation of treatment persistence and discontinuation rates. We
also conducted secondary analyses on TTD over the entire 20-year study,
particularly assessing at 2, 5, and 20 years to determine if initial differences
persisted over time.
Discontinuation of the drug and cessation of AP therapy, as well as switching to
another AP or a long-acting formulation of the same drug, were considered
discontinuations of oral treatment. This reflected a deliberate prescribing
decision, likely due to insufficient clinical response or clinically significant
side effects, poor medication adherence, or the need for more stable drug plasma
levels.
To assess the primary objective, Kaplan-Meier survival analysis was employed to
compare the duration of AP treatment among different APs. This approach
estimated the distribution of treatment duration, adjusting survival functions
for age and sex. Cases when individuals passed away, relocated to another region
or reached the end of the study period were treated as censored instances. We
utilized the Log-Rank test for comparing survival curves among different AP
treatments at the 3-month benchmark, aiming to identify any significant
disparities in treatment duration. The Cox regression model was used to
calculate both unadjusted and adjusted hazard ratios (HRs and HRad,
respectively), with 95% confidence intervals (CIs). This adjustment accounted
for age and sex as primary variables, and in separate models, it further
adjusted for the degree of AP polypharmacy at 3 months and for the Study Period
(Year of Study Entry). Using separate models helps assess the impact of each
variable set independently, to the advantage of easier interpretation of the
results. The level of AP polypharmacy was defined based on the occurrence of at
least three concurrent dispensations of distinct APs (ATC code: N05, Lithium
excluded) during any analyzed treatment period or the overlap of more than one
AP prescribing event. To account for the multiple comparisons and control the
false discovery rate, we applied the Benjamini-Hochberg method. This approach
sequentially adjusts p-values, thereby reducing the likelihood of type I errors
by considering the rank and quantity of comparisons made.
All statistical tests were performed using SAS software, SAS Institute Inc.,
Cary, NC, USA.
Data ownership, study funding, and reporting
This study was performed on behalf of the EPIFARM-Elderly project on drug
prescription in Lombardy which is a large pharmacoepidemiological collaborative
project on drug prescription in the elderly living in the Lombardy Region [35 ]. The present study had not yet
received any specific grant or support from any funding agency. Study data
belong to the Regional Health Ministry and are stored and managed by the
Istituto di Ricerche Farmacologiche ‘Mario Negri’ IRCCS. An exemption from
obtaining patientsʼ informed consent due to organizational constraints is
allowed by Italian privacy regulations for scientific research.
The study’s planning, conducting, and reporting align with the National Institute
for Health and Care Excellence (NICE) Real-World Evidence (RWE) framework, and
incorporating a detailed protocol developed prior to analysis [36 ]. We also adhered to the RECORD-PE
guidelines for reporting our results, with the RECORD-PE checklist provided in
Supplementary Document S1 for transparency [37 ].
Results
[Fig. 2 ] and [Fig. 3 ], respectively, depict the annual
AP use prevalence and incidence from 2001 to 2021. The prevalent graph demonstrates
a pronounced increase in SGA dispensations, dominating by 2021, while FGAs maintain
a relatively consistent but lesser prevalence throughout the period. Among
individual drugs, risperidone and aripiprazole both exhibit a gradual increase in
their dispensation rates. Olanzapine, notably, sees a decline in prescriptions
between 2006 and 2011, followed by a resurgence. After 2011, quetiapine also saw a
significant increase in its use, becoming the most dispensed AP drug by 2021.
Clozapine and amisulpride have remained relatively stable over the years, but with
prescription rates lower and comparable to those of FGAs. The incidence graph
illustrates a notable spike for drugs such as olanzapine, quetiapine, risperidone,
and aripiprazole between 2011 and 2013, followed by a partial decline, stabilizing
to some extent by 2021 ([Fig. 3 ]).
Fig. 2 2001–2021 Antipsychotics (APs) Dispensation Prevalence: Annual
dispensation of three or more APs (FGA= First Generation Antipsychotic; SGA=
Second Generation Antipsychotic).
Fig. 3 2001–2021 Antipsychotics (APs) Dispensation Incidence: Annual
new dispensation of three or more APs (FGA= First Generation Antipsychotic;
SGA= Second Generation Antipsychotic).
[Table 1 ] presents the basic demographic
characteristics and AP treatment details for the subjects in our analysis, covering
data on prescriptions, gender distribution, age specifics, and treatment persistence
across FGA and SGA classes. From January 1, 2001, to December 31, 2021, among
3,300,817 SSN beneficiaries aged 18-65, we identified 45,857 individuals with an
average age of 43.4 years (SD 12.2) who received at least three dispensations of one
of the analyzed APs without any prior consumption in the previous year (incident
cases). Of these, 42,434 subjects (93%) received at least two continuous
prescriptions and were thus included in our study cohort. These subjects were
responsible for generating a total of 58,232 prescription events that were used for
comparative assessments. Among these, 90% of the prescription events analyzed
involved polypharmacy.
Table 1 Demographics and antipsychotic treatment data of
patients included in this study (Study Cohort Subjects n=42,434). AP:
antipsychotic; FGA: first-generation antipsychotic; SGA:
second-generation antipsychotics; SD: standard deviation.
Haloperidol
Clozapine
Olanzapine
Quetiapine
Amisulpride
Risperidone
Aripriprazole
FGA*
SGA*
Incident Prescriptions, n
5210
3120
17446
16289
1541
7611
8879
12028
40112
Females, n (%)
2340 (44.91%)
1115 (35.74%)
8111 (46.5%)
8750 (53.7%)
585 (38.0%)
2984 (39.2%)
4684 (52.8%)
5575 (46.4%)
19409 (48.4%)
Age, mean (SD)
44.3 (11.5)
39.8 (11.6)
43.3 (12.3)
45.4 (12.3)
44.4 (11.6)
41.7 (12.4)
40.6 (12.4)
44.30 (11.6)
42.9 (12.8)
18–34 years, n (%)
1133 (21.8%)
1118 (35.8%)
4616 (26.5%)
3418 (21.0%)
333 (21.6%)
2332 (30.6%)
2884 (32.5%)
2616 (21.8%)
11287 (28.1%)
35–64 years, n (%)
4077 (78.3%)
2002 (64.2%)
12830 (73.5%)
12871 (79.0%)
1208 (78.4%)
5279 (69.4%)
5995 (67.5%)
9412 (78.3%)
28825 (71.9%)
Study cohort: continuative Prescriptions, n (%)
4441 (85.2%)
2852 (91.4%)
16601 (95.2%)
14322 (87.9%)
1383 (89.8%)
6788 (89.2%)
8285 (93.3%)
8001 (66.5%)
39572 (98.7%)
Females, n (%)
1939 (43.7%)
1002 (35.1%)
7683 (46.3%)
7678 (53.6%)
532 (38.5%)
2645 (39.0%)
4379 (52.9%)
3670 (45.9%)
18910 (47.8%)
Age, mean (SD)
44.1 (11.5)
39.5 (11.6)
43.2 (12.3)
45.2 (12.3)
44.2 (11.5)
41.6 (12.4)
40.5 (12.4)
44.2 (11.5)
42.5 (12.8)
18–34 years, n (%)
992 (22.3%)
1047 (36.7%)
4416 (26.6%)
3085 (21.5%)
301 (21.8%)
2095 (30.7%)
2714 (32.8%)
1751 (21.9%)
11510 (29.1%)
35–64 years, n (%)
3449 (77.7%)
1805 (63.3%)
12185 (73.4%)
11237 (78.5%)
1082 (78.2%)
4693 (69.1%)
5571 (67.2%)
6250 (78.1%)
28062 (70.9%)
Time to treatment discontinuation, mean (SD)
171.5 (306.5)
517.8 (998.3)
248.4 (627.0)
372.1 (1938.1)
338.5 (1823.7)
261.0 (991.9)
337.8 (1439.1)
751.5 (3224.9)
589.3 (2949.2)
Time to treatment discontinuation (TTD), median (IQR)
92 (53–171)
175 (77–518)
128 (70–241)
121 (58–268)
120 (66–216)
127 (66–229)
133 (69–254)
103 (54–210)
128 (66–257)
Personalised daily dose (PDD), median (IQR)
5.2 (3.33–9.4)
266.6 (155.5–466.5)
10.8 (6.7–20.0)
206.9 (68.4–514.3)
428.6 (260.9–857.1)
4.9 (3.2–8.9)
15.6 (9.7–28.0)
–
–
Chlorpromazine equivalent PDD, median (IQR)
313 (200–562.5)
399.9 (233.3–699.8)
323 (200–600)
165.5 (54.7–411.4)
195.7 (150.0–321.4)
461.5 (292.7–857.1)
311.1 (193.0–560.0)
117.7 (46.5–304.1)
300.0 (155.6–589.5)
Olanzapine equivalent PDD, median (IQR)
6.5 (4.2–11.7)
8.9 (5.2–15.6)
10.8 (6.7–20.0)
5.2 (1.7–12.7)
10.7 (6.5–21.4)
10.9 (7.1–20.0)
10.4 (6.43–18.7)
3.9 (1.6–10.1)
8.75 (4.9–16.7)
0–3 months AP polypharmacy prescriptions, n (%)
3735 (84.1)
2689 (94.3)
15447 (93.1)
12924 (90.2)
1290 (93.3)
6274 (92.4)
7427 (89.6)
6125 (76.6)
37175 (93.9)
0–3 months Number of AP polipharmacy drugs, median (IQR)
1 (1–2)
1 (1–2)
1 (1–1)
1 (1–1)
1 (1–2)
1 (1–2)
1 (1–2)
1 (1–1)
1 (1–1)
0–2 years AP polypharmacy prescriptions, n (%)
3787 (85.3)
2707 (94.9)
15490 (93.3)
12941 (90.4)
1292 (93.4)
6284 (92.6)
7434 (89.7)
6131 (76.6)
37098 (93.8)
3–5 years AP polypharmacy prescriptions, n (%)
175 (91.1)
594 (93.8)
1021 (89.0)
1062 (76.4)
62 (82.7)
311 (84.3)
527 (82.2)
356 (53.1)
2368 (68.1)
6–20 years AP polypharmacy prescriptions, n (%)
29 (85.3)
194 (92.4)
208 (87.4)
243 (62.0)
13 (54.2)
59 (67.8)
104 (67.1)
59 (18.1)
494 (39.1)
*Only the first continuous prescribing event for each subject within the same
class was included. This accounts for the lower total counts of SGA and FGA
compared to the sum of individual drugs within these classes.
Comparisons between groups revealed statistically significant differences in terms
of
PDD and TTD, with p<0.001 for all comparisons. Specifically, clozapine exhibited
a longer TTD than all other APs, with a median TTD of 175 days (IQR: 77-518),
demonstrating statistical significance in all comparisons (p<0.001).
[Table 1 ] shows that prescribed dosages of
various APs are within recommended ranges; specifically, olanzapine, risperidone,
aripiprazole, and amisulpride are on the higher end, clozapine, haloperidol, and
quetiapine are intermediate, while FGAs are on the lower end.
Globally, 38% of the prescription events were discontinued within 3 months of the
initial prescription, and 80% within 1 year. Within the first 12 months, only 0.4%
of those who discontinued resumed their original AP therapy, while 0.6% started a
different AP therapy. A substantial 99% did not resume any new continuative AP
treatment at 1-year. Over the extended period of the first 24 months, a slightly
higher proportion resumed a new continuative AP treatment: 2% returned to their
original therapy and 4% opted for a different AP. Nevertheless, a significant 94%
of
the patients did not resume any new AP continuative therapy at 2 years. It should
be
noted that 87% of the discontinuers were on AP polytherapy; thus, discontinuing the
AP medication started 3 months earlier did not imply a complete cessation of AP
therapy. Among discontinuers, both polytherapy and monotherapy groups showed
virtually identical AP therapy resumption rates (i. e., 1% at 1 year and 6% at 2
years), suggesting that the therapy modality does not significantly influence
whether patients resume a new continuative treatment after discontinuation. Notably,
among all discontinuers, 12%—specifically those in monotherapy who did not resume
any AP medications—were found not to be on continuous AP therapy within 2 years.
Meanwhile, 88% were, in fact, continuing with an AP therapy, either because they
were initially on polytherapy or because they had started a different treatment.
[Fig. 4 ], illustrates the age- and
sex-adjusted survival function estimates over a 365-day period for FGAs and SGAs.
It
shows a greater persistence in treatment for SGAs compared to FGAs, with this
disparity being particularly evident by the 90-day primary endpoint.
Fig. 4 Age and Sex-Adjusted Kaplan-Meier Survival Curves for Time to
Treatment Discontinuation by Antipsychotic Class: First-Generation
Antipsychotics (FGA) vs. Second-Generation Antipsychotics (SGA).
[Fig. 5 ] illustrates the age- and
sex-adjusted survival function estimates for various APs at 3 months. It visually
highlights marked differences in treatment persistence by the 90-day primary
endpoint: clozapine exhibits the highest persistence of use, indicating a lower
discontinuation rate compared to all other drugs; haloperidol and quetiapine show
a
more marked decrease in persistence, suggesting higher discontinuation rates among
patients.
Fig. 5 Age and Sex-Adjusted Kaplan-Meier Survival Curves for Time to
Treatment Discontinuation by Antipsychotic.
In [Fig. 6 ], we present the age- and
sex-adjusted hazard ratios (HRs) for a set of 28 AP comparisons based on the TTD at
the 3-months. The results demonstrate 24 significant differences. As a class, SGAs
demonstrated better treatment persistence than FGAs. Furthermore, all six SGAs have
HRs indicating superior persistence when compared to the class of FGAs. Clozapine
stands out for its superior persistence, surpassing the other five SGAs: olanzapine,
aripiprazole, amisulpride, risperidone, and quetiapine. Olanzapine and aripiprazole
trailed behind it, yet still showed better results than both risperidone and
quetiapine. Both amisulpride and risperidone performed only better than quetiapine.
Notably, quetiapine seems to be at a disadvantage, as it was inferior in terms of
3-month treatment discontinuation in all its pairwise comparisons against other
SGAs.
Fig. 6 Age and Sex-Adjusted Hazard Ratios (HRs) and Their 95%
Confidence Intervals (CIs) for 3-Month Time-to-Treatment Discontinuation
Across Various Antipsychotics. Note: CIs entirely to the left of the dashed
line indicate superior effectiveness of the first compared drug.
Further adjustments for polypharmacy and year yielded unwavering outcomes (see [Table 2 ]).
Table 2 Comparative 3-month antipsychotics treatment
discontinuation rates: Unadjusted hazard ratios (HRs), adjusted hazard
ratios (HRad) for age, sex, polypharmacy, and period/year of study
entry, and their 95% confidence intervals (CIs).
HRs (CIs)
Age and sex HRad (CIs)
Age,sex, and politherapy HRad (CIs)
age, sex, and period HRad (CIs)
Clozapine vs.
Haloperidol
0.54 (0.50; 0.59)
0.55 (0.51; 0.59)
0.55 (0.51; 0.60)
0.55 (0.51; 0.60)
Olanzapine
0.85 (0.79; 0.91)
0.85 (0.79; 0.92)
0.86 (0.80; 0.92)
0.85 (0.79; 0.92)
Quetiapine
0.71 (0.66; 0.76)
0.73 (0.68; 0.78)
0.73 (0.68; 0.79)
0.73 (0.68; 0.78)
Amisulpride
0.80 (0.71; 0.89)
0.81 (0.72; 0.90)
0.81 (0.72; 0.90)
0.81 (0.72; 0.90)
Risperidone
0.80 (0.74; 0.87)
0.80 (0.74; 0.87)
0.80 (0.74; 0.87)
0.80 (0.74; 0.87)
Aripriprazole
0.86 (0.79; 0.92)
0.87 (0.81; 0.94)
0.88 (0.81; 0.95)
0.90 (0.83; 0.97)
FGA
0.61 (0.57; 0.66)
0.61 (0.57; 0.66)
0.65 (0.60; 0.70)
0.62 (0.57; 0.66)
Olanzapine vs.
Haloperidol
0.64 (0.61; 0.67)
0.64 (0.61; 0.67)
0.63 (0.60; 0.66)
0.64 (0.61; 0.67)
Quetiapine
0.83 (0.80; 0.87)
0.84 (0.81; 0.87)
0.84 (0.81; 0.87)
0.84 (0.81; 0.87)
Amisulpride
0.94 (0.86; 1.03)
0.93 (0.85; 1.02)
0.93 (0.85; 1.02)
0.93 (0.85; 1.02)
Risperidone
0.94 (0.90; 0.99)
0.94 (0.90; 0.99)
0.94 (0.90; 0.99)
0.94 (0.90; 0.99)
Aripriprazole
1.01 (0.96; 1.05)
1.01 (0.97; 1.06)
1.01 (0.97; 1.06)
1.02 (0.98; 1.07)
FGA
0.71 (0.69; 0.75)
0.71 (0.68; 0.74)
0.72 (0.69; 0.75)
0.72 (0.69; 0.75)
Quetiapine vs.
Haloperidol
0.77 (0.73; 0.81)
0.77 (0.73; 0.81)
0.76 (0.73; 0.80)
0.77 (0.73; 0.81)
FGA
0.86 (0.83; 0.90)
0.86 (0.82; 0.89)
0.87 (0.83; 0.91)
0.87 (0.83; 0.90)
Amisulpride vs.
Haloperidol
0.68 (0.62; 0.75)
0.69 (0.62; 0.75)
0.69 (0.63; 0.76)
0.69 (0.63; 0.77)
Quetiapine
0.89 (0.81; 0.98)
0.90 (0.82; 0.98)
0.90 (0.82; 0.99)
0.90 (0.82; 0.99)
Risperidone
1.01 (0.92; 1.11)
1.01 (0.92; 1.12)
1.01 (0.92; 1.11)
1.01 (0.92; 1.12)
FGA
0.76 (0.70; 0.84)
0.77 (0.70; 0.84)
0.81 (0.74; 0.89)
0.77 (0.70; 0.85)
Risperidone vs.
Haloperidol
0.68 (0.64; 0.72)
0.68 (0.64; 0.72)
0.68 (0.64; 0.72)
0.69 (0.65; 0.73)
Quetiapine
0.89 (0.84; 0.93)
0.89 (0.85; 0.94)
0.89 (0.85; 0.94)
0.89 (0.85; 0.94)
FGA
0.76 (0.72; 0.80)
0.76 (0.72; 0.80)
0.78 (0.74; 0.82)
0.76 (0.73; 0.80)
Aripriprazole vs-
Haloperidol
0.64 (0.60; 0.67)
0.63 (0.59; 0.66)
0.63 (0.59; 0.66)
0.63 (0.60; 0.67)
Quetiapine
0.83 (0.79; 0.87)
0.83 (0.80; 0.87)
0.84 (0.80; 0.87)
0.83 (0.79; 0.87)
Amisulpride
0.93 (0.85; 1.02)
0.92 (0.84; 1.01)
0.92 (0.83; 1.01)
0.88 (0.79; 0.97)
Risperidone
0.94 (0.89; 0.99)
0.93 (0.88; 0.98)
0.93 (0.88; 0.98)
0.92 (0.87; 0.97)
FGA
0.71 (0.68; 0.75)
0.71 (0.67; 0.74)
0.73 (0.69; 0.76)
0.72 (0.68; 0.76)
SGA vs-
FGA
0.76 (0.73; 0.79)
0.76 (0.73; 0.79)
0.77 (0.74; 0.80)
0.76 (0.73; 0.79)
FGA: first-generation antipsychotics; SGA = Second-generation
antipsychotic.
In our secondary analyses, we further evaluated the long-term persistence of AP
treatments by comparing HRs for time-to-treatment discontinuation across several
time frames, from 0–2 years to 0-20 years. Additionally, we illustrated the
long-term time-to-treatment discontinuation over a 0-20-year period through survival
graphs, providing a visual representation of treatment persistence across the
studied APs (Supplementary Figure S1–S3 ). Notably, the head-to-head
comparisons of APs showed general consistency in their HRs over time, particularly
confirming clozapineʼs superiority across all periods (Supplementary Table
S1 ). However, some variations were observed, especially in the case of
quetiapine. Secondary analyses showed an inversion of trends in its favor,
outperforming olanzapine, risperidone, and amisulpride, despite being inferior to
them in the primary analysis.
Discussion
From 2001 to 2021, SGA prescriptions increased over FGAs, possibly due to perceived
better effectiveness and tolerability. Although the general trend for SGAs was
upward, specific drugs like olanzapine showed significant fluctuations, likely
influenced by external factors. Initially approved for schizophrenia, olanzapine
received expanded approvals for manic episodes and bipolar disorders in the early
2000s, broadening its applications. However, in 2006, lawsuits over its metabolic
side effects impacted global prescription trends, including in Italy [38 ]. This decline reversed around 2012 when
olanzapine’s patent expired, leading to increased prescriptions with the
availability of cheaper generics—a trend also seen with risperidone and quetiapine.
By 2021, the rise in SGA prescriptions, notably quetiapine and olanzapine, may have
been influenced by the mental health challenges of the COVID-19 pandemic.
We studied the persistence of newly initiated AP treatments in a naturalistic setting
where 90% of subjects were on polytherapy during the first 3 months of therapy,
highlighting challenges associated with monotherapy and underscoring the need for
personalized strategies. One possible explanation for this high polypharmacy rate
is
tapering practices, where patients are gradually transitioned from one AP to
another. However, polypharmacy remained consistently high during longer treatment
periods, particularly within the first 5 years, with a subsequent reduction only
during the 6–20-year follow-up, likely reflecting natural attrition. This trend of
later polypharmacy reduction was more pronounced for SGAs, suggesting a shift in
treatment strategies over time.
Our primary findings, consistent across therapy types even after adjusting for
polytherapy, confirm the superiority of SGAs over FGAs and, specifically, the
persistence of clozapine. Notable disparities in drug persistence evident at the
90-day endpoint persisted and remained stable over the following years.
To analyze the persistence of the classes of FGAs and SGAs, we used a method that
considers intra-class switching, where transitions between different
antipsychoticAPs within the same class are not classified as treatment
discontinuations. This approach led to particularly prolonged TTD for the class of
FGAs. Despite this, our analysis still highlights the clear superiority of SGAs over
FGAs. The superiority of SGAs aligns with the insights drawn from experimental
research [8 ]
[9 ]
[10 ]
[11 ]
[12 ]. In particular, this reaffirms the
narrative presented by Leucht et al. and Huhn et al., which emphasized the superior
performance of certain SGAs, specifically clozapine, amisulpride, olanzapine, and
risperidone [8 ]
[9 ]
[10 ]. Moreover, the finding of the superiority of clozapine is in line
with the results reported by Kishimoto et al., highlighting its dominance among
other SGAs in terms of all-cause discontinuation [11 ]
[12 ].
The mixed results for other SGAs, especially quetiapineʼs limited persistence
compared to counterparts like clozapine, olanzapine, and risperidone, require
further exploration. This aligns with Leucht et al. (2013) who found quetiapine
moderately effective and noted higher discontinuation rates in 2009 [9 ]
[39 ]. However, in our secondary analyses (Supplementary Materials),
quetiapine demonstrated higher persistence over longer follow-up periods (up to 20
years) compared to other APs except clozapine, contrasting with the primary
endpoint. This suggests that quetiapine may have a more favorable profile in
long-term treatment strategies. Nonetheless, the value of these secondary results
is
limited by the decreasing number of subjects who continue taking the same medication
over extended follow-up periods, reducing the absolute number of individuals at risk
and making the comparisons less representative.
Conversely, while olanzapine often leads in experimental studies, our data presents
a
more complex view. Similarly, amisulpride and risperidone, typically praised for
their efficacy, showed inconclusive results in our study [8 ]
[9 ]
[10 ]
[11 ]
[12 ]. Our findings highlight clozapineʼs marked persistence over
olanzapine, consistent with previous observational studies. Meta-analyses by
Soares-Weiser et al. (2013) and Masuda et al. (2019) have reported similar outcomes
[40 ]
[41 ]. Weiser et al. (2021) noted that
American veterans faced a significantly lower risk of discontinuing clozapine
compared to olanzapine [42 ]. Furthermore,
a comprehensive 7-year retrospective study in Québec found that individuals on
risperidone, clozapine, or polytherapy were significantly less likely to discontinue
treatment than those on olanzapine and quetiapine [43 ]. Additionally, Brodeur et al. (2022)
observed an exceptionally low discontinuation risk with clozapine compared to
olanzapine [44 ]. These consistent findings
affirm the pivotal role of clozapine in managing schizophrenia, underscoring its
efficacy and broad applicability across diverse patient populations. However, the
notable superiority of clozapine over olanzapine in our study, echoed in other
observational research, contrasts with results from meta-analyses of RCTs. This
discrepancy likely stems from clozapine being prescribed specifically for
treatment-resistant schizophrenia, where it demonstrates superior efficacy compared
to other APs, which are more likely to be prescribed for a broader range of
conditions in naturalistic settings as their prescribability has steadily increased
[45 ].
Our study, inherent with biases typical of observational designs such as selection
bias, endpoint misclassification, and confounding, is further limited by the absence
of psychiatric diagnoses and other clinical details. This gap prevents us from
assessing the appropriateness of AP prescriptions or distinguishing between patients
with conditions like psychosis, treatment-resistant depression, or bipolar
disorders. For example, the lower persistence observed with quetiapine may reflect
its adjunctive role in managing conditions such as anxiety or insomnia, often at
doses lower than those used for primary AP purposes, thereby influencing its 3-month
TTD rates. To mitigate bias from diagnostic variability, we focused our analysis on
adults aged 18-65 years, excluding those prescribed APs for dementia-related
behaviors, thus aiming for a more uniformly diagnosed cohort.
Utilizing observational data to infer drug effectiveness poses challenges, as the
endpoint of treatment discontinuation may not fully capture patient experiences. For
instance, discontinuations could indicate recovery or symptom control, not merely
treatment failure. Without detailed reasons for each cessation, we risk
misinterpreting positive outcomes as ineffective treatment. External factors like
patient beliefs or inadequate support could also drive discontinuations. To minimize
this bias, our study focused on a 3-month treatment duration, aligning with the
acute phase of AP treatment, which typically isn’t long enough to observe
discontinuations due to recovery.
While our research design has known limitations, its significance remains robust.
Our
large-scale epidemiological study enhances the understanding of drug efficacy in
routine clinical practice, not just in controlled environments. Additionally, the
high prevalence of polytherapy in our findings underscores the complexities and
criticalities of real-world AP prescribing [46 ]. The early predictive value of treatment discontinuation as a
reliable indicator of long-term effectiveness, highlights the utility of
observational data in clinical decision-making and research. Furthermore, the
stability of our findings, unaffected by annual prescription trends, emphasizes the
reliability of our method in evaluating pharmacological outcomes.
Despite the inevitable limitations of its observational design, the extensive
two-decade span of our study provides valuable insights for advancing AP research.
It bolsters the existing literature on the superiority of SGAs over FGAs and
highlights clozapine as a treatment option with the highest persistence rates [47 ]
[48 ].
These findings have important implications for clinical decisions and could guide
future research into broader applications and management strategies for clozapine.
Highly effective for treatment-resistant schizophrenia, clozapine offers benefits
including reduced suicide risk and improved cognitive function, thereby enhancing
quality of life. However, its usage is constrained by the need for strict monitoring
due to serious side effects such as agranulocytosis and myocarditis [45 ]. Although clozapine showed superior
persistence in our study, it remained one of the least prescribed APs. Our findings
underscore the need for further RCTs to compare clozapine with other APs, such as
olanzapine, across various diagnoses, while also exploring innovative strategies for
its utilization, management of potential side effects, and addressing polypharmacy
considerations. This could extend its benefits while optimizing its monitoring for
safer, broader applications.
Contributors
Alberto Parabiaghi formulated and wrote the study protocol, including the
methodology, and led the development of the project. Barbara D’Avanzo, and Angelo
Barbato contributed to the conception of the study and supervised the development
of
the project. Alessia Galbussera was responsible for data management and conducted
the statistical analysis. Mauro Tettamanti oversaw the statistical analysis. Alberto
Parabiaghi led the interpretation of the data. The original draft was written by
Alberto Parabiaghi. Angelo Barbato and Barbara D’Avanzo contributed to the results
validation. All authors reviewed the drafted work and approved the final
manuscript.
Data availability
Study data were obtained from the Lombardy Region under specific restrictions. These
data, used under a license for the current study, are not publicly accessible.
Access was regulated through the EPIFARM agreement between the Istituto di Ricerche
Farmacologiche "Mario Negri" IRCCS and the Regional Health Authority, with
the authorization granted by the latter. However, data can be made available by the
authors upon reasonable request and with the Lombardy Regionʼs permission.
Additional details on the results are also available from the corresponding author
upon reasonable request.
Informed Consent Statement
Informed Consent Statement
All methods were carried out in accordance with the Declaration of Helsinki. The
study was exempt from patients informed consent, and it was waived according to
General Authorization for the Processing of Personal Data for Scientific Research
Purposes Issued by the Italian Privacy Authority on August 10, 2018;
https://www.garanteprivacy.it/home/docweb/-/docweb-display/docweb/9124510#5.
Furthermore, the study provided sufficient guarantees of individual records
anonymity, with respect to Italian privacy law.
