Schizophrenia Research 197 (2018) 274–280

Contents lists available at ScienceDirect

Schizophrenia Research

journal homepage: www.elsevier.com/locate/schres

Antipsychotics and mortality in a nationwide cohort of 29,823 patients

with schizophrenia
Heidi Taipale a,b, Ellenor Mittendorfer-Rutz a, Kristina Alexanderson a, Maila Majak c, Juha Mehtälä c,
Fabian Hoti c, Erik Jedenius d, Dana Enkusson d, Amy Leval d, Jan Sermon e, Antti Tanskanen a,f,g, Jari Tiihonen a,g,⁎
a
Karolinska Institutet, Department of Clinical Neuroscience, Stockholm, Sweden
b
University of Eastern Finland, School of Pharmacy, Kuopio, Finland
c
EPID Research Oy, Espoo, Finland
d
Janssen Cilag, Solna, Sweden
e
Janssen Cilag, Beerse, Belgium
f
National Institute for Health and Welfare, The Impact Assessment Unit, Helsinki, Finland
g
University of Eastern Finland, Department of Forensic Psychiatry, Niuvanniemi Hospital, Kuopio, Finland

a r t i c l e i n f o a b s t r a c t

Article history: Introduction: It has remained controversial if antipsychotic treatment is associated with increased or decreased
Received 25 September 2017 mortality among patients with schizophrenia, and if there are any clinically meaningful differences between spe-
Received in revised form 14 December 2017 cific agents and routes of administration.
Accepted 16 December 2017 Methods: We linked prospectively gathered nationwide register-based data during 2006–2013 to study all-cause
Available online 21 December 2017
mortality among all patients aged 16–64 years with schizophrenia in Sweden (N = 29,823 in total; N = 4603 in
the incident cohort). Multivariate Cox regression models were adjusted for clinical and sociodemographic covar-
Keywords:
Schizophrenia
iates. Sensitivity analyses with the incident cohort were conducted to control for survival bias.
Antipsychotic Results: During the mean follow-up of 5.7 years, 2515 patients (8.4%) died. During the maximum follow-up
Long-acting injection (7.5 years), the lowest cumulative mortality was observed for second generation (SG) long-acting injection
Treatment (LAI) use (7.5%). Adjusted hazard ratios (aHRs) compared to SG LAI use were 1.37 (95%CI 1.01–1.86) for first gen-
Death eration (FG) LAIs, 1.52 (1.13–2.05) for SG orals, 1.83 (1.33–2.50) for FG orals, and 3.39 (2.53–4.56) for nonuse of
Prescription register antipsychotics. Concerning specific agents, the lowest mortality was observed for once-monthly paliperidone LAI
(0.11, 0.03–0.43), oral aripiprazole (0.22, 0.15–0.34), and risperidone LAI (0.31, 0.23–0.43). In pairwise compar-
ison, LAIs were associated with 33% lower mortality than equivalent orals (0.67, 0.56–0.80). The results with in-
cident cohort were consistent with the primary analyses.
Conclusions: Among patients with schizophrenia, LAI use is associated with an approximately 30% lower risk of
death compared with oral agents. SG LAIs and oral aripiprazole are associated with the lowest mortality.
© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction
Patients with schizophrenia have a 15–20 year shorter life expec-
tancy than the general population (Laursen et al., 2014), and side effects
of antipsychotic medications are considered a putative cause for the ex-
cess mortality (Glassman and Bigger Jr., 2001; Liebzeit et al., 2001; Ray
et al., 2001; Cheeta et al., 2004; Fergusson et al., 2005; Mackin et al.,
2007; Ray et al., 2009; Stahl et al., 2009; Stone et al., 2009). A systematic
review suggested that gap in mortality compared with general popula-
tion is even worsening and may be related to second generation
⁎ Corresponding author at: Karolinska Institutet, Department of Clinical Neuroscience,
Byggnad R5, S-17176 Stockholm, Sweden.
E-mail address: jari.tiihonen@ki.se (J. Tiihonen).
antipsychotic use (Saha et al., 2007). Meta-analysis and systematic re-
views of randomized controlled trials (RCTs) suggest that this is not
the case, since mortality is lower during use of antipsychotics than dur-
ing placebo (Baxter et al., 2016; Khan et al., 2007, 2013). However, these
trial results have been criticized because the duration of treatments is
usually substantially longer for active than placebo arms. Also, trials
lasting a few months are too short to assess fatal adverse events related
to cumulative drug exposure leading to health problems such as weight
gain or diabetes.
Several observational studies on large unselected cohorts have
shown that mortality is lower during use of antipsychotic compared
with no use (Tiihonen et al., 2006, 2009, 2011, 2012, 2016; Baandrup
et al., 2010; Crump et al., 2013; Vanasse et al., 2016). However, these
studies either did not control for survival bias or had short follow-up pe-
riods which made it difficult to evaluate the comparative effectiveness

https://doi.org/10.1016/j.schres.2017.12.010
0920-9964/© 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280
between specific antipsychotics. Further, data on novel agents have
been limited, and it is not known whether the route of administration
[long-acting injection (LAI) vs. oral] modifies mortality. We aimed to
study mortality during specific antipsychotic treatments in a nation-
wide cohort, also including a large number of first-episode patients to
control for survival bias.
2. Materials and methods
This study was based on nationwide data, derived from the Swedish
population-based registers. The Regional Ethics Board of Stockholm ap-
proved this research project (decision 2007/762–31).
2.1. Study population
All residents aged 16–64 (at year 2006) living in Sweden with regis-
tered schizophrenia treatment contact between July 1, 2006 until De-
cember 31, 2013 were included in this study. The flow chart of the
cohort is shown in Supplementary Fig. 1. In addition to this prevalent
cohort, an incident cohort with individuals newly diagnosed with
schizophrenia were identified. Schizophrenia diagnosis was based on
four registers: the National Patient Register (maintained by the National
Board of Health and Welfare) regarding inpatient care since 1988 and
specialized outpatient care since 2001, data on disability pension since
1994 and sickness absence since 2005 from the MiDAS register (main-
tained by the Swedish Social Insurance Agency). All Swedish residents
have been assigned a unique personal identification number which en-
abled linkage between various registers (no missing data). Drug use
data since July 2005 was gathered from the Prescribed Drug Register
(maintained by the National Board of Health and Welfare) and dates
of death were obtained from the Causes of Death Register (maintained
by the National Board of Health and Welfare). Demographic character-
istics were based on data in the LISA register (maintained by Statistics
Sweden).
All individuals with a diagnosis of schizophrenia, schizotypal and de-
lusional disorders [F20–F29 according to the International Classification
of Diseases version 10 (ICD-10) classification] were identified from in-
patient, specialized outpatient, sickness absence and disability pension
(MiDAS) registers and formed the source population (N = 57,256). An
inclusion criterion was diagnosis of schizophrenia (schizophrenia F20
or schizoaffective disorder F25) as main diagnoses in the registers dur-
ing July 1, 2006 until December 31, 2013 (N = 33,940 fulfilled this
criteria). Based on the exclusion criteria, those aged b 16 at cohort
entry or over age 64 in 2006 were excluded, leading to the study cohort
of 29,823 individuals (prevalent cohort). The incident cohort (N =
4603) was defined from the study cohort based on not having a previ-
ous main or contributory diagnosis of F20–29 (ICD-10) or 295 (ICD-9)
before July 1, 2006 in any of the four databases, and not using antipsy-
chotics between July 1, 2005 and July 1, 2006 according to the Pre-
scribed Drug Register. The cohort entry date was defined as the first
diagnosis fulfilling the inclusion criteria (starting from July 1, 2006 for
prevalent cases), and individuals were followed up until death or De-
cember 31, 2013 (which ever occurred first). This cohort has been
used also to study the risk of re-hospitalization and all-cause discontin-
uation of antipsychotic treatment (Tiihonen et al., 2017).
2.2. Exposure
Antipsychotic use was derived from the Prescribed Drug Register
which includes all prescribed dispensed drugs during 2005–2013.
Drugs administrated in by healthcare, e.g., during hospitalization are
not recorded in the register. Antipsychotics were identified according
to the Anatomical Therapeutic Chemical (ATC) classification (WHO)
code N05A, excluding lithium (N05AN01). Regarding the package infor-
mation, antipsychotics were categorized according to drug formulation
into oral antipsychotics and long-acting injections (LAI). Further
275
categorization was made into second-generation (SGA) and first-gener-
ation (FGA) antipsychotics.
The PRE2DUP method was utilized to model drug use periods from
prescription drug purchases (Tanskanen et al., 2015). This method is
based on mathematical modelling of drug purchasing behavior for
each individual and for each drug substance (ATC code). The method
takes into account stockpiling of drugs, dose changes, and periods of
hospitalization when drugs are provided by the hospital and not re-
corded in the drug register. In this method, drug use is controlled with
restriction parameters defining the minimum and maximum daily
dose for each package (Nordic product number, vnr). When modelling
antipsychotics, each drug substance was coded according to drug for-
mulation as oral or LAI, and drug use periods were constructed sepa-
rately for oral and LAI use. The PRE2DUP method has been utilized
previously in studies of antipsychotics (Tiihonen et al., 2009; Taipale
et al., 2014; Tolppanen et al., 2016) and validated by expert-opinion
on drug use period formation and by comparing it with interview-
based medication use data (Taipale et al., 2016).
2.3. Outcomes
The main outcome measure was all-cause mortality.
2.4. Covariates
The multivariate Cox regression models were adjusted for
sociodemographic factors, antipsychotic medication use and schizo-
phrenia related factors, other medication use in dependent manner
and comorbidities. Comorbid conditions were identified from the Na-
tional Patient Register (inpatient care and specialized outpatient care)
and drug use from the Prescribed Drug Register. For some variables
(such as substance abuse), combination of these data sources was
used. The exact definitions are provided in the Supplementary Table 1.
2.5. Statistical analyses
We used multivariate-adjusted Cox regression in the analyses. The
risk of mortality was compared through the use of two approaches con-
sidering time i) on antipsychotic monotherapy only, and ii) on any ther-
apy. In approach i), treatment periods were comprised into a single
factor variable indicating either monotherapy of a specified antipsy-
chotic, polytherapy if any two or more antipsychotics were used at the
same time, or no use of any antipsychotics. Events and risk time were
accounted for a specific antipsychotic only if they occurred during
monotherapy of that particular antipsychotic or for polytherapy, if two
or more antipsychotics were used at the same time. In approach ii),
treatment periods were defined by separate variables for each specific
antipsychotic indicating either ongoing treatment or no use of that par-
ticular antipsychotic. In this analysis, events and risk time were
accounted for a specific antipsychotic whenever that antipsychotic
treatment was ongoing (also when used in polytherapy). The difference
between these two approaches is described in Supplementary Fig. 2. In
these analyses, all deaths were included and deaths in hospitals were
considered attributable to the last exposure period in outpatient care.
In addition, using otherwise similar approach as in ii), we conducted
oral vs. LAI analyses, in which exposure of each antipsychotic with
both oral and LAI formulation was comprised into a factor variable
with status either no use, oral use or LAI use depending on whether
that particular antipsychotic was not used, used orally, or used as LAI,
respectively. In these analyses, simultaneous use of oral and LAI was
accounted as LAI use (because in pairwise comparisons, oral use was
the reference), and polytherapy was a separate variable that was ad-
justed for when two or more antipsychotics were used simultaneously.
For comparison between specific antipsychotics, oral olanzapine was
used as a reference drug as it was the most often used drug in the
study population.
276 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280
Time dependent Kaplan-Meier curves were constructed for FG-SG
and oral-LAI groups. In these curves, patients may contribute to several
groups. We conducted sensitivity analyses among the incident cohort in
order to control for survival bias. In the comparison of specific antipsy-
chotics (n = 20), the level of significance was set at p b 0.0025.
3. Results
The clinical and sociodemographic characteristics of the prevalent
and incident cohorts are described in Supplementary Table 2. During
the follow-up (mean 5.7 years, median 6.9 years), 2515 (8.4%) of pa-
tients died. The proportions and mean daily doses of used antipsy-
chotics are shown in Supplementary Table 3. Oral olanzapine was
most frequently used antipsychotic, followed by oral aripiprazole and
oral risperidone in both prevalent and incident cohorts. Of LAIs,
zuclopenthixol LAI was most commonly used in prevalent cohort and
risperidone LAI in incident cohort.
The adjusted risk of death was 56% lower during use of any antipsy-
chotic compared with no use of antipsychotic (1811 deaths/
13,8451 person years versus 704 deaths/32,793 person years (unad-
justed HR 0.60, 95% CI 0.55–0.66), adjusted HR 0.44, 95% CI 0.39–0.49,
p b 0.0001). Fig. 1 demonstrates the Kaplan-Meier curves of mortality
in first (FG) and second (SG) generation oral and long-acting injection
(LAI) users compared with no antipsychotic use. Cumulative mortality
rates during maximum follow-up of 7.5 years were 7.5% during SG LAI
use, 8.5% during SG oral, 12.2% during FG oral, 12.3% during FG LAI,
and 15.2% during no use of antipsychotics. Similarly, adjusted risk of
death was the lowest during SG-LAI use (Fig. 2).
The mortality rates and data on previous hospitalizations, suicide at-
tempts, and substance abuse for these drug groups are shown in Supple-
mentary Table 4. FG-LAI and SG-LAI groups were more likely to have
previous hospitalizations, 66.1% and 72.0% had at least two previous
hospitalizations, respectively. They also had higher frequency of previ-
ous suicide attempts and substance abuse than oral users.
When compared with SG LAI use, the aHRs (95% CI) were 1.37
(1.01–1.86) for FG LAIs, 1.51 (1.13–2.05) for SG orals, 1.83 (1.33–2.50)
for FG orals, and 3.39 (2.53–4.56) for no use of antipsychotics. The re-
sults for pairwise comparisons for LAI versus corresponding oral use
Fig. 1. Kaplan-Meier curve on risk of mortality in first (FG) and second (SG) generation oral and long-acting injection (LAI) use compared with no antipsychotic use. The same patients
contributed to several groups if they switched their medication. During maximum follow-up of 7.5 years, the lowest cumulative mortality rate was observed for SG LAI use (7.5%).
SG−LAI
0.38 (0.3, 0.47)
●
Treatment
0.52 (0.45, 0.6)
FG−LAI ●
0.58 (0.52, 0.65)
SG−Oral ●
0.75 (0.66, 0.85)
FG−Oral ●
0.0 0.5 1.0 1.5
Hazard Ratio
Fig. 2. The adjusted Hazard Ratios (aHRs) and 95% Confidence Intervals (CI) for first (FG)
and second (SG) generation long-acting injections (LAIs) and orals compared to no use of
antipsychotic (any therapy, including polypharmacy).
are shown in Fig. 3. aHRs were favorable to LAI formulations in all
pairwise comparison although did not reach statistical significance.
The overall risk of death was 33% lower during LAI use compared with
equivalent oral use (0.67, 0.56–0.80, p b 0.0001).
Mortality rates and person-years of use for specific antipsychotics in
monotherapy are shown in Table 1 for prevalent and incident cohort.
The adjusted risks of mortality during monotherapy of specific antipsy-
chotic use in the prevalent cohort (compared with non-use) are shown
in Fig. 4. The lowest mortality was found for once-monthly paliperidone
LAI (0.11, 0.03–0.43), followed by oral aripiprazole (0.23, 0.15–0.34)
and risperidone LAI (0.31, 0.23–0.43). The corresponding adjusted HRs
during any therapy (concomitant use of other antipsychotics allowed)
showed similar results as the monotherapy analysis (Supplementary
Fig. 3).
Results compared to oral olanzapine as the reference drug (most fre-
quently used medication) in the prevalent cohort are shown in Supple-
mentary Fig. 4. Aripiprazole was the only antipsychotic which differed
significantly from oral olanzapine when Bonferroni correction was ap-
plied (decreased mortality, p = 0.0003), in addition to no use of
 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280 277

Fig. 3. The adjusted Hazard Ratios (aHRs) and 95% Confidence Intervals results for pairwise comparisons for specific LAI (blue) versus equivalent oral (reference). aHRs for no use of
antipsychotic are shown in red.

antipsychotic which was associated with increased risk of death (p b 4. Discussion

0.0001).
In the incident cohort, the number of deaths (n = 152 total) during
specific antipsychotic treatments were too low for meaningful analysis
for several specific agents when compared with no use of antipsychotic.
The adjusted HRs were 0.15 (0.04–0.53) for SG LAIs, 0.53 (0.35–0.80) for
SG orals, 0.64 (0.34–1.21) for FG LAIs, and 0.66 (0.34–1.29) for FG orals.
Use of any antipsychotic was also associated with substantially lower
risk of death (0.54, 0.36–0.80) when compared with no use.
Our results showed that antipsychotic use was associated with an
approximately 50% lower risk of death when compared with no use,
which strongly suggests that the net effect on mortality is beneficial.
Since medications are started when symptoms re-appear or get worse,
it is obvious that the results would be even more favorable for antipsy-
chotic treatment if this bias could be eliminated. Our results are in line
with all previous large observational cohort studies (Tiihonen et al.,

Table 1
The mortality rates during monotherapy of specific antipsychotics among the prevalent and incident cohorts.

Prevalent (n = 29,823) Incident (n = 4603)

Person years Events Incidence rate/100 person years Person years Events Incidence rate/100 person years

First generation LAIs

Fluphenazine LAI 335 15 4.48 (2.70–7.44) 3 0 na
Flupentixol LAI 1619 32 1.98 (1.40–2.80) 48 0 na
Haloperidol LAI 2709 41 1.51 (1.11–2.06) 79 3 3.78 (1.22–11.72)
Perphenazine LAI 5061 71 1.40 (1.11–1.77) 277 2 0.72 (0.18–2.89)
Zuclopenthixol LAI 6326 101 1.60 (1.31–1.94) 167 5 2.99 (1.24–7.18)

First generation orals

Flupentixol oral 2435 29 1.19 (0.83–1.71) 90 2 2.22 (0.56–8.88)
Haloperidol oral 3388 62 1.83 (1.43–2.35) 189 3 1.58 (0.51–4.91)
Levomepromazine oral 1168 36 3.08 (2.22–4.27) 95 4 4.21 (1.58–11.21)
Perphenazine oral 3430 34 0.99 (0.71–1.39) 191 0 na
Zuclopenthixol oral 3436 50 1.46 (1.10–1.92) 104 1 0.97 (0.14–6.86)

Second generation LAIs

Olanzapine LAI 254 3 1.18 (0.38–3.66) 61 0 na
Paliperidone LAI 495 2 0.40 (0.10–1.62) 124 1 0.81 (0.11–5.72)
Risperidone LAI 4489 45 1.00 (0.75–1.34) 336 1 0.30 (0.04–2.12)

Second generation orals

Aripiprazole oral 5749 25 0.43 (0.29–0.64) 866 5 0.58 (0.24–1.39)
Clozapine oral 14,460 161 1.11 (0.95–1.30) 386 3 0.78 (0.25–2.41)
Olanzapine oral 19,735 226 1.14 (1.00–1.30) 1691 13 0.77 (0.45–1.32)
Quetiapine oral 4481 54 1.21 (0.92–1.57) 820 5 0.61 (0.25–1.46)
Risperidone oral 11,305 117 1.03 (0.86–1.24) 859 6 0.70 (0.31–1.56)
Other 2933 26 0.89 (0.60–1.30) 293 2 0.68 (0.17–2.73)
Polytherapy 44,643 681 1.53 (1.42–1.64) 2161 20 0.93 (0.60–1.43)
No antipsychotic 32,793 704 2.14 (1.99–2.31) 7028 76 1.08 (0.86–1.35)

LAI = long-acting injection.

278 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280
0.11 (0.03, 0.43)
LAI Paliperidone ●
0.22 (0.15, 0.34)
Oral Aripiprazole ●
0.31 (0.23, 0.43)
LAI Risperidone ●
LAI Haloperidol
LAI Perphenazine
Oral Perphenazine
LAI Olanzapine
Treatment LAI Zuclopenthixol
Other orals
Oral Risperidone
Oral Flupentixol
Oral Quetiapine
Oral Olanzapine
Oral Zuclopenthixol
LAI Flupentixol
Oral Clozapine
Oral Haloperidol
Oral Levomepromazine
LAI Fluphenazine
0.0
Fig. 4. The adjusted Hazard Ratios of mortality during exposure to antipsychotic monotherapies compared with no use in the prevalent population, including all LAIs and ten most
frequently used orals. All treatments except olanzapine LAI, levomepromazine and fluphenazine LAI survived Bonferroni correction (level of significance p b 0.0025).
2006, 2009, 2011, 2012, 2016; Baandrup et al., 2010, Crump et al., 2013:
Vanasse et al., 2016) and meta-analyses on RCTs (Baxter et al., 2016;
Khan et al., 2007, 2013). The lowest risk of death was observed during
those time periods when patients used second generation LAIs, both
in the prevalent and incident cohorts.
Oral aripiprazole and commonly used LAIs were associated with the
lowest mortality both as monotherapy and as any therapy
(polypharmacy allowed). Two previous Swedish studies have also re-
ported that aripiprazole is associated with the lowest risk of death, but
they did not study mortality for LAIs (Crump et al., 2013; Ringbäck
Weitoft et al., 2014). These studies used smaller cohorts and shorter fol-
low-up periods, resulting in low statistical power, did not apply
Bonferroni correction for multiple comparisons, and did not conduct
sensitivity analyses to control for survival bias. However, their results
on aripiprazole are well in line with our results. Previous results from
large Finnish cohorts (Tiihonen et al., 2009; Kiviniemi et al., 2013)
showed the lowest mortality for clozapine, but in the present study clo-
zapine was not among those medications with the best outcomes. This
may be related to its more restricted use in Sweden compared to in Fin-
land. It is probable that in Sweden clozapine is used only among the
most severely ill patients, and the traditional between-subject analysis
used in the comparison of specific antipsychotics may not have been
able to fully adjust for the severity of illness. On the other hand, since
aripiprazole has little metabolic adverse effects, it may have used fre-
quently among patients with high risk of cardiovascular morbidity,
but still it was associated with very low mortality compared with
most of the other antipsychotics. Among the commonly used antipsy-
chotics, levomepromazine was associated with the highest mortality,
which is in line with previous geriatric and schizophrenia cohort studies
(Kiviniemi et al., 2013; Sahlberg et al., 2015; Schmedt et al., 2016; Shi et
al., 2007). This discourages the use of levomepromazine either as mono-
therapy or as combination treatment in schizophrenia.
In the Cox regression analyses, we adjusted for 20 sociodemographic
and clinical co-variates such as number of previous hospitalizations,
previous and current use of psychotropic and somatic medications, sub-
stance abuse, and history of suicidal behavior. Unfortunately, we were
not able to adjust for smoking status which is a confounder here. How-
ever, the adjusted HR (0.44) was even lower than the unadjusted HR
(0.60) in the comparison between antipsychotic use versus no use,
which indicates that patients using more antipsychotics have a higher
intrinsic risk of death. This suggests that the beneficial effect of antipsy-
chotic use would have been even stronger if residual confounding could
0.36 (0.26, 0.51)
●
0.37 (0.29, 0.49)
●
0.38 (0.27, 0.54)
●
0.39 (0.12, 1.23)
●
0.4 (0.32, 0.5)
●
0.4 (0.27, 0.61)
●
0.41 (0.33, 0.51)
●
0.44 (0.3, 0.64)
●
0.47 (0.36, 0.63)
●
0.48 (0.41, 0.56)
●
0.48 (0.36, 0.65)
●
0.52 (0.36, 0.75)
●
0.53 (0.44, 0.64)
●
0.59 (0.45, 0.78)
●
0.77 (0.55, 1.09)
●
0.83 (0.49, 1.43)
●
0.5 1.0 1.5
Hazard Ratio
have been totally eliminated. Previous literature has reported that pa-
tients treated with LAIs more often have substance abuse, more severe
psychopathology, and more previous hospitalizations than patients
using oral antipsychotics (Shi et al., 2007). This was also seen in our re-
sults as patients using LAIs had the highest rate of previous psychiatric
hospitalizations, suicide attempts, and substance abuse, and these co-
variates were associated with higher risk of death. This implies that pa-
tients receiving LAIs might have less healthy life styles in general, and
higher levels of morbidity than other patients, which suggests that the
superiority of LAIs over orals would have been even more robust if re-
sidual confounding could have been totally eliminated.
The relative difference in mortality between any LAI versus equiva-
lent orals was 33% in the total cohort, and the mortality for SG LAIs com-
pared with other drug classes was 27–49% lower in the prevalent cohort
and 71–77% lower in the incident cohort. These figures correspond to 2–
4% difference in the absolute risk during about 6 years follow-up in the
prevalent cohort, and 2% difference during 3.5 years follow-up in the in-
cident cohort. Extrapolation of these results would correspond to about
10% difference in absolute risk in a 15–20 year time span, which sug-
gests that wider use of SG LAIs might result in substantially lower mor-
tality among patients with schizophrenia. Overall, the results show that
excess mortality among patients with schizophrenia is more likely asso-
ciated with a lack of antipsychotic therapy rather than with antipsy-
chotic treatment.
The main strength of this study is nationwide coverage of all schizo-
phrenia patients and their follow-up based on register data covering all
residents. We also used data on actually purchased medications instead
of data on prescriptions given to the patients. Our results are generaliz-
able to relatively high-income countries with Caucasian populations
with all antipsychotic treatments reimbursed by the state. To account
for survival bias, we conducted analyses within incident cohort
representing new cases. Drug use was modelled with state-of-the-art
PRE2DUP method which describes well actual drug use when compared
with interview-reported use (Taipale et al., 2016). Use of other medica-
tions was treated as continuously updated variables instead of crude
baseline estimates allowing better adjustment for time varying
conditions.
A limitation was that, despite of adjustment for covariates, all obser-
vational studies have residual confounding. However, adjustment for
the effects of known clinical and sociodemographic variables indicated
that patients using more antipsychotics have a higher intrinsic risk of
death. Therefore, it is likely that residual confounding has diluted rather
 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280
than exaggerated the observed difference between use versus no use of
antipsychotics. Although antipsychotics were compared with time
when antipsychotics were not used, the majority of persons did have
time periods of antipsychotic use and never-use was rare. Thus, non-
use time represents mostly non-treatment periods of users. Lithium
use was not considered in the analyses although it may have anti-sui-
cidal effects. The patients using LAIs had the highest rate of previous
psychiatric hospitalizations, suicide attempts, and substance abuse
which co-variates were associates with increased risk of death. This sug-
gests that the superiority of LAIs over orals would have been even more
robust if residual confounding could have been totally eliminated.
5. Conclusions
Mortality among patients with schizophrenia is over 40% lower dur-
ing those time periods when the patients use antipsychotics than when
they do not. LAI use is associated with an approximately 30% lower risk
of death compared with the oral use of the same medication. SG LAIs
and oral aripiprazole are associated with the lowest mortality.
Funding
The study was funded by Janssen Cilag (R092670SCH4045). Drs Jedenius, Sermon,
Leval and Enkusson, co-authors employed by Janssen Cilag, were involved in study con-
tent and design and critical revision of the manuscript for important intellectual content.
Their authorship roles adhere to ICMJE criteria.
Contributors
J. Tiihonen, E. Mittendorfer-Rutz, K. Alexanderson, E. Jedenius, D. Enkusson, A. Leval, J.
Sermon, A. Tanskanen, and H. Taipale were responsible for the study concept and design.
E. Mittendorfer-Rutz, K. Alexanderson, A. Tanskanen, and H. Taipale were responsible for
data extraction. M. Majak, J. Mehtälä, and F. Hoti were responsible for statistical analysis.
H. Taipale and J. Tiihonen were responsible for drafting the manuscript. All other authors
were responsible for critical revision of the manuscript and have accepted the final
version.
Conflict of interest
J. Tiihonen reports serving as a consultant to the Finnish Medicines Agency Fimea,
AstraZeneca, Bristol-Myers Squibb, Eli Lilly, F. Hoffman-La Roche, Janssen-Cilag, Lundbeck,
and Organon; receiving fees for giving expert testimonies to AstraZeneca, Bristol-Myers
Squibb, Eli Lilly, GlaxoSmithKline, Janssen-Cilag, Lundbeck, Otsuka and Pfizer; receiving
lecture fees from AstraZeneca, Bristol-Myers Squibb, Eli Lilly, GlaxoSmithKline, Janssen-
Cilag, Lundbeck, Novartis, Otsuka, Pfizer; receiving grants from Stanley Foundation and
the Sigrid Jusélius Foundation; serving as a member of the advisory boards for
AstraZeneca, Eli Lilly, Janssen-Cilag, and Otsuka, and participating in research projects
funded by Janssen-Cilag and Eli Lilly with grants paid to Karolinska Institutet. K.
Alexanderson has an unrestricted research grant from Biogen. F. Hoti, M. Majak, and J.
Mehtälä reported being employed by EPID Research, which is a contract research organi-
zation that performs commissioned pharmacoepidemiological studies, and thus its em-
ployees have been and currently are working in collaboration with several
pharmaceutical companies. A. Tanskanen and H. Taipale reported participating in research
projects funded by Janssen-Cilag and Eli Lilly with grants paid to the Karolinska Institutet.
A. Tanskanen reported serving as a member of advisory board for Janssen-Cilag. E.
Jedenius, D. Enkusson, A. Leval, and J. Sermon are employed by Janssen Cilag Pharmaceu-
ticals. E. Mittendorfer-Rutz reports no conflict of interest.
Acknowledgements
Authors thank Ms. Aija Räsänen for secretarial assistance.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.schres.2017.12.010.
References
Baandrup, L., Gasse, C., Jensen, V.D., Glenthoj, B.Y., Nordentoft, M., Lublin, H., Fink-Jensen,
A., Lindhardt, A., Mortensen, P.B., 2010. Antipsychotic polypharmacy and risk of death
from natural causes in patients with schizophrenia: a population-based nested case-
control study. J. Clin. Psychiatry 71 (2), 103–108.
Baxter, A.J., Harris, M.G., Khatib, Y., Brugha, T.S., Bien, H., Bhui, K., 2016. Reducing excess
mortality due to chronic disease in people with severe mental illness: meta-review
of health interventions. Br. J. Psychiatry 208 (4), 322–329.
Cheeta, S., Schifano, F., Oyefeso, A., Webb, L., Ghodse, A.H., 2004. Antidepressant-related
deaths and antidepressant prescriptions in England and Wales, 1998–2000. Br.
279
J. Psychiatry 184, 41–47.
Crump, C., Winkleby, M.A., Sundquist, K., Sundquist, J., 2013. Comorbidities and mortality
in persons with schizophrenia: a Swedish national cohort study. Am. J. Psychiatry 170
(3), 324–333.
Fergusson, D., Doucette, S., Glass, K.C., Shapiro, S., Healy, D., Hebert, P., Hutton, B., 2005.
Association between suicide attempts and selective serotonin reuptake inhibitors:
systematic review of randomised controlled trials. BMJ 330, 396.
Glassman, A.H., Bigger Jr., J.T., 2001. Antipsychotic drugs: prolonged QTc interval, torsade
de pointes, and sudden death. Am. J. Psychiatry 158 (11), 1774–1782.
Khan, A., Schwartz, K., Stern, C., Redding, N., Kolts, R.L., Brown, W.A., Robinson, D.S., 2007.
Mortality risk in patients with schizophrenia participating in premarketing atypical
antipsychotic clinical trials. J. Clin. Psychiatry 68 (12), 1828–1833.
Khan, A., Faucett, J., Morrison, S., Brown, W.A., 2013. Comparative mortality risk in adult
patients with schizophrenia, depression, bipolar disorder, anxiety disorders, and at-
tention-deficit/hyperactivity disorder participating in psychopharmacology clinical
trials. JAMA Psychiat. 70 (10), 1091–1099.
Kiviniemi, M., Suvisaari, J., Koivumaa-Honkanen, H., Häkkinen, U., Isohanni, M., Hakko, H.,
2013. Antipsychotics and mortality in first-onset schizophrenia: prospective Finnish
register study with 5-year follow-up. Schizophr. Res. 150 (1), 274–280.
Laursen, T.M., Nordentoft, M., Mortensen, P.B., 2014. Excess early mortality in schizophre-
nia. Annu. Rev. Clin. Psychol. 10, 425–448.
Liebzeit, K.A., Markowitz, J.S., Caley, C.F., 2001. New onset diabetes and atypical antipsy-
chotics. Eur. Neuropsychopharmacol. 11 (1), 25–32.
Mackin, P., Bishop, D., Watkinson, H., Gallagher, P., Ferrier, I.N., 2007. Metabolic disease
and cardiovascular risk in people treated with antipsychotics in the community. Br.
J. Psychiatry 191 (1), 23–29.
Ray, W.A., Meredith, S., Thapa, P.B., Meador, K.G., Hall, K., Murray, K.T., 2001. Antipsy-
chotics and the risk of sudden cardiac death. Arch. Gen. Psychiatry 58 (12),
1161–1167.
Ray, W.A., Chung, C.P., Murray, K.T., Hall, K., Stein, C.M., 2009. Atypical antipsychotic drugs
and the risk of sudden cardiac death. N. Engl. J. Med. 360 (3), 225–235.
Ringbäck Weitoft, G., Berglund, M., Lindström, E.A., Nilsson, M., Salmi, P., Rosén, M., 2014.
Mortality, attempted suicide, re-hospitalisation and prescription refill for clozapine
and other antipsychotics in Sweden-a register-based study. Pharmacoepidemiol.
Drug Saf. 23 (3), 290–298.
Saha, S., Chant, D., McGrath, J., 2007. A systematic review of mortality in schizophrenia: is
the differential mortality gap worsening overtime? Arch. Gen. Psychiatry 64 (10),
1123–1131.
Sahlberg, M., Holm, E., Gislason, G.H., Køber, L., Torp-Pedersen, C., Andersson, C., 2015. As-
sociation of selected antipsychotic agents with major adverse cardiovascular events
and noncardiovascular mortality in elderly persons. J. Am. Heart Assoc. 4 (9),
e001666.
Schmedt, N., Kollhorst, B., Enders, D., Jobski, K., Krappweis, J., Garbe, E., Schink, T.,
2016. Comparative risk of death in older adults treated with antipsychotics: a
population-based cohort study. Eur. Neuropsychopharmacol. 26 (9),
1390–1400.
Shi, L., Ascher-Svanum, H., Zhu, B., Faries, D., Montgomery, W., Marder, S.R., 2007. Charac-
teristics and use patterns of patients taking first-generation depot antipsychotics or
oral antipsychotics for schizophrenia. Psychiatr. Serv. 58 (9), 482–488.
Stahl, S.M., Mignon, L., Meyer, J.M., 2009. Which comes first: atypical antipsychotic treat-
ment or cardiometabolic risk? Acta Psychiatr. Scand. 119 (3), 171–179.
Stone, M., Laughren, T., Jones, M.L., Levenson, M., Holland, P.C., Hughes, A., Hammad, T.A.,
Temple, R., Rochester, G., 2009. Risk of suicidality in clinical trials of antidepressants
in adults: analysis of proprietary data submitted to US Food and Drug Administration.
BMJ 339, b2880.
Taipale, H., Koponen, M., Tanskanen, A., Tolppanen, A.M., Tiihonen, J., Hartikainen, S.,
2014. Antipsychotic polypharmacy among a nationwide sample of community-
dwelling persons with Alzheimer's disease. J. Alzheimers Dis. 41 (10),
1223–1228.
Taipale, H., Tanskanen, A., Koponen, M., Tolppanen, A.M., Tiihonen, J., Hartikainen, S.,
2016. Agreement between PRE2DUP register data modeling method and comprehen-
sive drug use interview among older persons. Clin. Epidemiol. 8, 363–371.
Tanskanen, A., Taipale, H., Koponen, M., Tolppanen, A.M., Hartikainen, S., Ahonen, R.,
Tiihonen, J., 2015. From prescription drug purchases to drug use periods - a second
generation method (PRE2DUP). BMC Med. Inform. Decis. Mak. 15, 21.
Tiihonen, J., Wahlbeck, K., Lönnqvist, J., Klaukka, T., Ioannidis, J.P., Volavka, J., Haukka, J.,
2006. Effectiveness of antipsychotic treatments in a nationwide cohort of patients
in community care after first hospitalisation due to schizophrenia and schizoaffective
disorder: observational follow-up study. BMJ 333 (7561), 224.
Tiihonen, J., Lönnqvist, J., Wahlbeck, K., Klaukka, T., Niskanen, L., Tanskanen, A., Haukka, J.,
2009. 11-year follow-up of mortality in patients with schizophrenia: a population-
based cohort study (FIN11 study). Lancet 374 (9690), 620–627.
Tiihonen, J., Haukka, J., Taylor, M., Haddad, P.M., Patel, M.X., Korhonen, P., 2011. A nation-
wide cohort study of oral and depot antipsychotics after first hospitalization for
schizophrenia. Am. J. Psychiatry 168 (6), 603–609.
Tiihonen, J., Suokas, J.T., Suvisaari, J.M., Haukka, J., Korhonen, P., 2012. Polypharmacy with
antipsychotics, antidepressants, or benzodiazepines and mortality in schizophrenia.
Arch. Gen. Psychiatry 69 (5), 476–483.
Tiihonen, J., Mittendorfer-Rutz, E., Torniainen, M., Alexanderson, K., Tanskanen, A., 2016.
Mortality and cumulative exposure to antipsychotics, antidepressants, and benzodi-
azepines in patients with schizophrenia: an observational follow-up study. Am.
J. Psychiatry 173 (6), 600–606.
Tiihonen, J., Mittendorfer-Rutz, E., Majak, M., Mehtälä, J., Hoti, F., Jedenius, E., Enkusson, D.,
Leval, A., Sermon, J., Tanskanen, A., Taipale, H., 2017. Real-world effectiveness of an-
tipsychotic treatments in a nationwide cohort of 29 823 patients with schizophrenia.
JAMA Psychiat. 74 (7), 686–693.
280 H. Taipale et al. / Schizophrenia Research 197 (2018) 274–280

Tolppanen, A.M., Koponen, M., Tanskanen, A., Lavikainen, P., Sund, R., Tiihonen, J.,
Hartikainen, S., Taipale, H., 2016. Antipsychotic use and risk of hospitalization or
death due to pneumonia in persons with and those without Alzheimer disease.
Chest 150 (6), 1233–1241.
Vanasse, A., Blais, L., Courteau, J., Cohen, A.A., Roberge, P., Larouche, A., Grignon, S., Fleury,
M.J., Lesage, A., Demers, M.F., Roy, M.A., Carrier, J.D., Delorme, A., 2016. Comparative
effectiveness and safety of antipsychotic drugs in schizophrenia treatment: a real-
world observational study. Acta Psychiatr. Scand. 134 (5), 374–384.
WHO Collaborating Centre for Drug Statistics Methodology, d. The Anatomical Therapeu-
tic Chemical Classification System. Structure and principles. http://www.whocc.no/
atc/structure_and_principles/, Accessed date: 30 July 2017.