Risk Factors, Incidence, and Outcomes of Neuroleptic Malignant Syndrome on Long-Acting Injectable vs Oral Antipsychotics in a Nationwide Schizophrenia Cohort

Abstract

Introduction

Long-acting injectable antipsychotics (LAIs) are associated with multiple positive outcomes in psychosis, but it is unclear whether LAIs are associated with worse outcomes if neuroleptic malignant syndrome (NMS), a potentially lethal adverse effect, occurs.

Methods

We used nationwide and nationally representative databases of healthcare encounters in Finland to study the incidence and outcome predictors of NMS in patients diagnosed with schizophrenia/schizoaffective disorder between January 01, 1972 and December 31, 2017. Using a nested case-control design, we also explored differences by antipsychotic formulation (LAI vs oral antipsychotic [OAP]) and class (first-generation antipsychotic [FGA] vs second-generation antipsychotic [SGA]).

Results

One hundred seventy-two NMS cases and 1441 sex-, age-, and diagnosis-matched controls were included (age = 58.8 ± 13.1 years, males = 59.9%). Incidence of NMS was 1.99 (1.98–2.00) per 10 000 person-years. The likelihood of developing NMS did not differ by antipsychotic formulation (adjusted odds ratio [aOR]: 0.89, 95% confidence intervals [95% CI]: 0.59–1.33, for LAIs vs OAPs) or class (FGA-OAP vs SGA-OAP [aOR: 1.08, 95% CI: 0.66–1.76], FGA-LAI [aOR: 0.89, 95% CI: 0.52–1.53], SGA-LAI [aOR: 1.35, 95% CI: 0.58–3.12]). NMS risk factors included antipsychotic treatment change: increased number (odds ratios [OR]: 5.00, 95% CI: 2.56–9.73); decreased number/switch (OR: 2.43, 95% CI: 1.19–4.96); higher antipsychotic dose (>2DDDs–OR: 3.15, 95% CI: 1.61–6.18); co-treatment with anticholinergics (OR: 2.26, 95% CI: 1.57–3.24), lithium (OR: 2.16, 95% CI: 1.30–3.58), benzodiazepines (OR: 2.02, 95% CI: 1.44–3.58); and comorbid cardiovascular disease (OR: 1.73, 95% CI: 1.22–2.45). Within 30 days, 4.7% of cases with NMS died (15.1% within 1 year) without differences by antipsychotic formulation. NMS reoccurred in 5 of 119 subjects (4.2%), after a median = 795 (range = 77–839) days after rechallenge with antipsychotics.

Conclusion

NMS remains a potentially life-threatening risk, yet these results should further contribute to mitigate concerns about LAI safety regarding NMS onset or outcomes, including mortality.

Introduction

Neuroleptic Malignant Syndrome (NMS) is a potentially lethal idiosyncratic reaction to dopamine blocking agents, such as antipsychotics, characterized by fever, muscle rigidity, altered mental status, and autonomic dysfunction.^1–3 Leukocytosis, tremor, electrolyte abnormalities, and elevated creatinine kinase (CK) are also common features.^3,4 Management of NMS is supportive and immediate cessation of antipsychotics or any other precipitating agent is recommended.³ In addition, some treatments including dantrolene, bromocriptine, lorazepam, and electroconvulsive therapy (ECT) have been proposed, albeit randomized controlled trials are lacking.³ Immediate cessation of antipsychotics is not possible in the case of long-acting injectable antipsychotics (LAIs) due to their prolonged medication release, requiring 5 half-lives, ie, 10–60 days for antipsychotic blood levels return to zero.^5–7 Conversely, in the case of oral antipsychotics (OAPs) discontinuation is generally achieved after 3–5 days.⁸

LAIs have shown superiority vs OAPs for relapse prevention, all-cause discontinuation, hospitalization and mortality,^5,9–14 thus becoming a valuable treatment alternative to address non-adherence and to reduce the burden associated with daily medication intake,^11,15–17 with safety data being generally equivalent to those of OAPs.^14,18,19 The inability to stop LAIs immediately in case of a serious adverse event, however, has been cited as an argument against LAI use^5,20 and could be a deterrent for some clinicians, as the advantage of LAIs derived precisely from the longer-term medication release, may be offset by an increased risk of more severe and prolonged NMS. This question is becoming particularly relevant, given the expansion of the injection interval to currently 3 months,^21,22 with longer intervals under investigation,²³ and the expansion of LAI indications and use besides schizophrenia.²⁴

A recent publication²⁵ analyzed individual patient data pooled from 662 individual NMS case reports, finding no significant differences in relevant outcomes, such as mortality, duration of hospitalization or sequelae, but a slightly longer duration of NMS on LAIs (median = 2 vs 1.4 weeks, P = .036), but only for first-generation antipsychotic (FGA) LAIs (FGA-LAIs), and only after secondary analysis. However, these results showing non-differential risk for mortality and adverse sequelae when NMS develops during LAI vs OAP treatment need replication, as published case reports are unlikely representative of the actual population at risk for the studied event. A longitudinal Danish register-based case-control study²⁶ of 224 372 patients with organic, psychotic, affective, or neurotic diagnosis, reported 83 NMS cases (0.04%), finding that NMS was associated with presence of LAIs (odds ratios [OR]: 4.53, 95% confidence intervals [95% CI]: 1.60–12.80, P < .005), similar to second-generation antipsychotic (SGA) OAPs (SGA-OAPs) (OR: 4.66, 95% CI: 1.96–11.10, P = .001) and mid-potency FGA-OAPs (OR: 4.81. 95% CI: 1.96–11.79, P = .001), but substantially lower than high-potency FGA-OAPs (OR = 23.41, 95% CI: 5.29–103.61, P < .001). However, since this study relied on data from 1996–2007, before currently available SGA-LAIs were approved, evidence for the risk of NMS associated with currently used SGA-LAIs⁷ is missing from nationwide database studies. Indeed, national databases and large case-control or cohort studies can be a suitable alternative to complement pooled, individual case report analyses,²⁵ allowing for epidemiological, long-term observations that capture a substantial proportion of the course of illness of individuals with chronic conditions, while mitigating selection biases. In this study, we used the National Finnish Registry to expand on the descriptive epidemiology of NMS, as well as the influence of treatment formulation (LAI vs OAP) and class (FGA vs SGA) on the incidence and outcomes of NMS.

The primary aims of this study were to: (1) determine the total incidence of NMS in a nationally representative sample of antipsychotic-treated individuals, (2) calculate the incidence of NMS by treatment formulation (LAI vs OAP), (3) characterize patient- and treatment-related risk factors associated with NMS using a nested case-control approach, and (4) determine survival after NMS and patient and treatment-related risk factors associated with NMS-related mortality (death during 30 days, 90 days and 1 year after NMS), using a case-control and case-only (lethal compared with non-lethal NMS) approaches. Secondary aims of the study included to (1) determine the incidence of NMS in a nationally representative sample on FGA-LAI vs SGA-LAI treatment, (2) on FGA-OAP vs SGA-OAP treatment, (3) on antipsychotic monotherapy (APMONO) vs >1 antipsychotic (APPOLY), irrespective of formulation, (4) determine the length of hospitalization after NMS as well as patient and treatment-related risk factors associated with long vs short hospitalization among NMS cases, and (5) determine the risk of second NMS (in the overall sample and among those rechallenged with antipsychotics), patient and treatment-related risk factors associated with NMS recurrence, and survival after NMS upon antipsychotic rechallenge.

Methods

We used the Finnish Care Register for Health Care and Prescription Register, nationwide and nationally representative databases of healthcare encounters in Finland, to study the incidence and outcome predictors of NMS overall and by antipsychotic formulation (ie, LAI vs OAP) and class (ie, FGA vs SGA). The structure and content of the protocol and the study followed the recommendations by the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) statement.²⁷

Study Population

The prevalent cohort from the nationwide source population included all individuals residing in Finland who were diagnosed with schizophrenia or schizoaffective disorder (ICD-10 F20, F25) between January 1, 1972 and December 31, 2014 (N = 61 889). In addition, to avoid survival bias, individuals whose first diagnoses of schizophrenia or schizoaffective disorder occurred during the observation period (ie, January 1, 1996 and December 31, 2017) and who did not use antipsychotics during a year preceding the diagnoses comprised the incident cohort (N = 8342). Finally, the last group of interest were those who developed recurrent NMS. Therefore, the populations of interest were: (1) prevalent cohort, (2) incident cohort, (3) rechallenge/reoccurrence population.

Antipsychotic Exposure Prior to NMS

Antipsychotic exposure was considered as any outpatient dispensing of drugs categorized according to the Anatomical Therapeutic Chemical (ATC) classification with code N05A, which belong to antipsychotics. Lithium, which is also included within the same ATC code was excluded. The PRE2DUP method²⁸ was utilized for constructing antipsychotic use periods and for calculating antipsychotic dose at the time of event and during the month prior to the event. The purchased amount of treatment is recorded in defined daily dose (DDD),²⁹ together with information on drug package and formulation. Briefly, a DDD of 1 reflects the typical therapeutic dose for a particular drug. Antipsychotics used by the study cohort were further divided according to drug formulation into OAPs and LAIs. Antipsychotics were also categorized into FGAs and SGAs, resulting in the following subcategories: FGA-OAP, SGA-OAP, FGA-LAI and SGA-LAI, as well as into antipsychotic monotherapy (APMONO) vs polytherapy (APPOLY). When assessing different antipsychotic categories (OAP vs LAI), periods when both OAP and LAI are used concomitantly were coded as LAI use.

Outcomes

The primary outcome was hospital discharges with the diagnosis of NMS, ie, ICD-10 G21.0/ICD-9:333.92. For drug use analyses, outcome events recorded during the first 2 weeks after hospital admission were considered as outpatient events and assigned for exposure category, which was ongoing at the time of admission (as drugs used during hospital care are not recorded in the register data). Secondary outcomes were (1) NMS associated death (ie, lethal NMS, defined as death occurring within 30 days, 90 days and 1 year after NMS), and (2) duration of hospitalization for NMS (as a proxy for severity, categorized as long vs short by median split), and (3) NMS recurrence as second NMS event (ie, diagnosed >30 days after discharge from hospitalization of the first NMS event).

Covariates

Covariates included comorbid conditions and treatment-related characteristics since the nested case control model requires matching on sociodemographic factors. A complete list and definition of covariates can be found in supplementary table 1.

Models and Statistical Analysis

Incidence of NMS.

We estimated the absolute incidence of NMS among the whole study population, and among the antipsychotic-exposed population, as well as the incidence rate per 10 000 person-years, for each population of interest. Incidence was also calculated for secondary outcomes. For NMS recurrence, only individuals for whom there was a diagnosis of NMS and were re-exposed to antipsychotics after the hospital discharge at which NMS was diagnosed were included in these analyses. Incidence was calculated separately for each group of interest (1) prevalent cohort, (2) incident cohort, (3) by antipsychotic formulation and class, and (4) rechallenge/reoccurrence population. We also compared NMS incidence by different exposure categories by calculating the number of events per person-time of use for exposure to LAIs vs OAPs, as well as all the other predefined subgroups.

Nested Case-Control Model.

To study associations between patient and treatment characteristics and risk of NMS, we constructed a nested case-control design and analyzed it with conditional logistic regression analyses that consider the matched design. Cases of NMS were matched with controls without NMS (matched by age [±1 year], sex, time since cohort entry [±1 year], incident/ prevalent, schizophrenia/schizoaffective status), by selecting up to 10 controls for each case (1:10). Once we obtained estimates for univariate analyses, we selected variables for the adjusted multivariable model. Variables for which there was a significant association in the univariate analysis, and for whom there were at least 10 individuals within each category of the outcome (ie, present or absent) were entered in the multivariable analysis. To reduce multicollinearity, when a variable was at least moderately correlated with another one (Pearson correlation > 0.3), we included only the variable with the greatest association in the univariate analysis. We estimated the OR as well as the 95% CI. Comparison between lethal vs non-lethal NMS and between “long” vs “short” hospital stay after NMS was conducted only among cases with NMS. Multivariable models on lethal vs non-lethal NMS and long vs short stay after NMS were run with the same covariates as the main analyses.

Results

Incidence of NMS

There were 348 events per 881 696 person-years in the whole cohort, translating into an incidence rate of NMS of 3.95 (95% CI: 3.94–3.95) per 10 000 person-years. In the incident cohort there were 19 events per 95 678 person-years, translating into an incidence rate of NMS of 1.99 (95% CI: 1.98–2.00) per 10 000 person-years. Median time since diagnoses of schizophrenia to NMS was 4.8 years (interquartile range [IQR]: 1.9–12.9). Of incident cases with their first event (N = 18), N = 3 (17.7%) were diagnosed with NMS within 1 year after first diagnosis of schizophrenia/schizoaffective disorder, N = 4 (25.5%) were diagnosed with NMS within 1–3 years after diagnosis, and the rest, N = 10 (58.8%), were diagnosed with NMS during 3 years after their first schizophrenia/schizoaffective diagnoses. For the whole cohort, incidence of NMS by antipsychotic treatment formulation (LAI vs OAP), class (FGA vs SGA) and by monotherapy or polytherapy (APMONO vs APPOLY) are shown in table 1. Due to the low incidence of NMS, incidence rates for the incident cohort could not be reliably estimated for these categories.

Table 1.

Incidences of NMS

Category	Events (n)	Persons-Years (n)	Incidence Rate per 10 000 Person-Years (CI)
OAP	131	519 772	2.520 (2.516–2.525)
FGA-OAP	36	157 363	2.288 (2.280–2.295)
SGA-OAP	61	279 715	2.181 (2.175–2.186)
OAP MIXED	34	82 694	4.112 (4.098–4.125)
LAI	33	117 314	2.813 (2.803–2.823)
FGA-LAI	24	88 725	2.705 (2.694–2.716)
SGA-LAI	9	28 589	3.148 (3.127–3.169)
APMONO	82	367 472	2.231 (2.227–2.236)
APPOLY	82	270 393	3.033 (3.026–3.304)

Note: NMS, neuroleptic malignant syndrome; APMONO, antipsychotic monotherapy; APPOLY, antipsychotic polytherapy; CI, confidence interval; FGA-OAP, first-generation oral antipsychotic; FGA-LAI, first-generation long-acting injectable antipsychotic; LAI, long-acting injectable antipsychotic; OAP, oral antipsychotic; SGA-LAI, second-generation long-acting injectable antipsychotic; SGA-OAP, second-generation oral antipsychotic.

Patient and Treatment-Related Risk Factors Associated with NMS Onset

Our model included 172 cases of NMS and 1441 sex-, age-, and diagnosis-matched controls. Of these 172 cases treated in outpatient care before NMS, n = 103 (59.9%) were male, and the mean age was 58.8 years of age (SD = 13.1) (table 2).

Table 2.

Characteristics of NMS Cases vs Controls (1:10)

	Controls N = 1441	Cases N = 172	Unadjusted OR (95% CI)	P-value
Matching factors
Male sex % (n)	59.3% (855)	59.9% (103)	[matched]
Mean age (SD)	58.4 (12.8)	58.8 (13.1)	[matched]
Mean time since cohort entry (SD), days	3379 (2285)	3444 (2311)	[matched]
Incident	5.9% (85)	6.4%(11)	[matched]
Schizoaffective	18.3% (263)	17.4% (30)	[matched]
Exposures % (n)
OAP vs LAI				.5477
OAP	81.5% (1175)	79.7% (137)	reference
LAI	18.5% (266)	20.4% (35)	1.15 (0.77–1.71)
OAP vs LAI vs FGA vs SGA				.0235
FGA-OAP	28.8 (415)	21.5% (37)	reference
SGA-OAP	40.9 (589)	38.4 (66)	1.40 (0.86–2.27)
FGA and SGA-OAP	11.9 (171)	19.8% (34)	2.43 (1.43–4.12)
FGA-LAI	14.4 (208)	15.1 (26)	1.47 (0.86–2.52)
SGA-LAI	4.0 (58)	5.2 (9)	2.01 (0.87–4.65)
APMONO vs POLY				.0094
AP monotherapy	60.9% (877)	50.6% (87)	reference
AP polytherapy	39.1% (564)	49.4% (85)	1.53 (1.10–2.11)
Change in number of antipsychotics				<.0001
No change	94.7 (1365)	86.1 (148)	reference
Increase	2.2 (31)	8.1 (14)	4.16 (2.11–8.19)
Decrease or switch	3.1 (45)	5.8 (10)	2.38 (1.11–5.11)
AP dose at the event				<.0001
<0.5 DDDs	22.7 (327)	10.7 (18)	reference
0.5–<1.0 DDDs	23.4 (337)	18.6 (32)	2.04 (1.09–3.80)
1–<2.0 DDDs	30.1 (433)	35.5 (61)	3.15 (1.77–5.62)
≥2.0 DDDs	23.9 (344)	35.5 (61)	4.11 (2.25–7.48)
AP dose change during last month				<.0001
No change	95.6 (1377)	80.8 (139)	reference
Increase	1.7 (24)	9.9 (17)	8.05 (4.00–16.18)
Decrease	2.8 (40)	9.3 (16)	4.07 (2.15–7.69)
Other medication use
Lithium	6.2 (89)	14.5 (25)	3.13 (1.83–5.35)	<.0001
Other mood stabilizers	14.6 (210)	16.3 (28)	1.10 (0.70–1.72)	.6867
Antidepressant	25.1 (362)	24.4 (42)	0.96 (0.66–1.39)	.8406
Benzodiazepines	28.3 (408)	48.3 (83)	2.39 (1.73–3.31)	<.0001
Z-drugs	7.2 (104)	5.2 (9)	0.68 (0.34–1.37)	.3351
Anticholinergic anti-Parkinsonian drugs	10.9 (157)	26.7 (46)	2.84 (1.93–4.19)	<.0001
Comorbidities
Asthma/COPD	3.7 (53)	7.0 (12)	1.92 (0.98–4.19)	.0376
Diabetes	8.6 (124)	9.9 (17)	1.15 (0.67–1.98)	.5747
Cardiovascular disease	24.4 (352)	35.5 (61)	1.72 (1.21–2.44)	.0017
Cancer	4.8 (69)	4.7 (8)	0.94 (0.44–2.04)	.9364
Substance abuse	13.7 (197)	10.5 (18)	0.76 (0.45–1.29)	.2423

Note: AP, antipsychotic;APMONO, antipsychotic monotherapy; APPOLY, antipsychotic polytherapy; CI, confidence interval; COPD, chronic obstructive pulmonary disease; FGA-OAP, first-generation oral antipsychotic; FGA-LAI, first-generation long-acting injectable antipsychotic; LAI, long-acting injectable antipsychotic; OAP, oral antipsychotic; SGA-LAI, second-generation long-acting injectable antipsychotic; SGA-OAP, second-generation oral antipsychotic.

Results of the multivariable final model showed no differences in the likelihood of developing NMS based on antipsychotic formulation (OR: 0.89, 95% CI: 0.59–1.33 for LAI vs OAP) or antipsychotic polytherapy (OR: 0.99, 95% CI: 0.70–1.40 vs monotherapy). Conversely, change in antipsychotic treatment (increased number of antipsychotics–OR: 5.00, 95% CI: 2.56–9.73) or decreased number/switch of antipsychotics (OR: 2.43, 95% CI: 1.19–4.96), higher antipsychotic dose (>2DDDs–OR: 3.15, 95% CI: 1.61–6.18) was related to increased likelihood of developing NMS (figure 1). Further, co-treatment with anticholinergic anti-Parkinsonian drugs (OR: 2.26, 95% CI: 1.57–3.24), lithium (OR: 2.16, 95% CI: 1.30–3.58), benzodiazepines (OR: 2.02, 95% CI: 1.44–3.58) and comorbid cardiovascular disease (OR: 1.73, 95% CI: 1.22–2.45) were also associated with increased likelihood of NMS (figure 1). In additional sensitivity analyses, excluding cases who initiated use of benzodiazepines or anticholinergics within 7, 14, 30, or 60 days before the event, the results did not change. We also performed a sensitivity analysis on the subgroup of patients receiving only LAIs vs OAP only, also showing no differences in the likelihood of developing NMS (OR: 0.84, 95% CI: 0.40–1.74).

Fig. 1.

Factors associated with neuroleptic malignant syndrome. Forest plot showing factors associated with NMS presented as adjusted odds ratios (aORs) with 95% confidence intervals (CIs) resulting from the multivariate model, which was adjusted for all factors shown, in addition to age, sex, incident, and type of diagnoses which were matched between cases and controls. The category “anticholinergic anti-Parkinsonian drugs” does not include dopamine agonists, levodopa or any drug other than biperidene (99%) and trihexyphenidyl (ATC-category N04AA01). Abbreviations: AP: antipsychotic; DDD: daily defined dose; LAI: long-acting injectable antipsychotic; OAP: oral antipsychotic; NMS: neuroleptic malignant syndrome.

A secondary model was conducted to study FGA-OAP vs SGA-OAP vs FGA-LAI vs SGA-LAI, which could not be included in the above model due to overlapping definitions with LAI vs OAP variables. Compared with FGA-OAP, SGA-OAP (adjusted OR [aOR]: 1.08, 95% CI: 0.66–1.76, P = .7555), FGA-LAI (aOR: 0.89, 95% CI: 0.52–1.53, P = .6804), SGA-LAI (aOR: 1.35, 95% CI: 0.58–3.12, P = .4889) and concomitant use of both FGA-OAP and SGA-OAP (aOR: 1.24, 95% CI: 0.73–2.08, P = .4284) were not associated with NMS.

Patient and Treatment-Related Risk Factors Associated with Long vs Short Hospital Stay

Duration of hospitalization due to NMS had a median of 31 days (IQR: 12–83). Hence, we categorized short duration as cases in which the length of hospitalization was <31 days and long duration as cases that had median ≥ 31 days. With this classification, n = 88 had “long” duration of hospitalization. We found no differences in predictors between the 2 groups (table 3).

Table 3.

Characteristics of NMS Cases With Long vs Short Hospital Stay After the NMS Event

	Short Stay (<31 days) N = 84	Long Stay (≥31 days) N = 88	P-value
Sample characteristics
Male sex % (n)	54.8 (46)	64.8 (57)	.1806
Median age (IQR)	62 (52–69)	57 (47–67)
Time since cohort entry (IQR), days	3478 (510–5473)	3229 (1206–4837)
Incident	6.0 (5)	6.8 (6)	.8166
Schizoaffective	13.1 (11)	21.6 (19)	.1422
Exposures % (N)
OAP vs LAI			.9719
OAP	79.8 (67)	79.6 (70)
LAI	20.2 (17)	20.5 (18)
OAP vs LAI vs FGA vs SGA			.7632
FGA-OAP	23.8 (20)	19.3 (17)
SGA-OAP	35.7 (30)	40.9 (36)
FGA and SGA-OAP	20.2 (17)	19.3 (17)
FGA-LAI	16.7 (14)	13.6 (12)
SGA-LAI	3.6 (3)	6.8 (6)
APMONO vs POLY			.4435
AP monotherapy	53.6 (45)	47.7 (42)
AP polytherapy	46.4 (39)	52.3 (46)
Change in number of antipsychotics			.2516
Increase	6.0 (5)	10.2 (9)
Decrease or switch	8.3 (7)	3.4 (3)
AP dose at the event			.0784
<0.5 DDDs	10.7 (9)	10.2 (9)
0.5–<1.0 DDDs	26.2 (22)	11.4 (10)
1–<2.0 DDDs	33.3 (28)	37.5 (33)
≥2.0 DDDs	29.8 (25)	40.9 (36)
AP dose change during last month			.2516
No change	85.7 (72)	86.4 (76)
Increase	6.0 (5)	10.2 (9)
Decrease	8.3 (7)	3.4 (3)
Other medication use
Lithium	13.1 (11)	15.9 (14)	.6007
Antidepressant	22.6 (19)	26.1 (23)	.5915
Benzodiazepines	48.8 (41)	47.7 (42)	.8871
Z-drugs	7.1 (6)	3.4 (3)	.2717
Anti-parkinson drugs	29.8 (25)	23.9 (21)	.3823
Comorbidities
Asthma/COPD	10.7 (9)	3.4 (3)	.0601
Diabetes	13.1 (11)	6.8 (6)	.1679
Cardiovascular disease	39.3 (33)	31.8 (28)	.3062
Cancer	7.1 (6)	2.3 (2)	NA
Substance abuse	11.9 (10)	9.1 (8)	.5468

Note: AP: antipsychotic; APMONO: antipsychotic monotherapy; APPOLY: antipsychotic polytherapy; CI: confidence interval; COPD: chronic obstructive pulmonary disease; FGA-OAP: first-generation oral antipsychotic; FGA-LAI: first-generation long-acting injectable antipsychotic; IQR: interquartile range; NA: Not applicable; LAI: long-acting injectable antipsychotic; OAP: oral antipsychotic; SGA-LAI: second-generation long-acting injectable antipsychotic; SGA-OAP: second-generation oral antipsychotic.

Patient and Treatment-Related Risk Factors Associated with NMS Survival and Mortality

Within 30 days of NMS diagnosis, N = 3 controls (0.2%) and N = 8 cases (4.7%) died (P < .0001), N = 9 controls (0.6%) and N = 17 cases (9.9%) died within the 90 days following NMS (P < .0001), and N = 34 (2.4%) controls and N = 26 (15.1%) cases died within 1 year after NMS (P < .0001).

Of those who died within the 30 days of NMS diagnosis (N = 8), N = 7 used OAP (5.1% of OAP users with NMS died) and N = 1 used LAI (2.9% of LAI users died). Of subjects who died within 90 days after NMS diagnosis, N = 14 (10.2% of OAP users with NMS) used OAPs, and N = 3 (8.6% of LAI users with NMS) used LAIs. Lastly, of subjects who died within 1 year following NMS diagnosis, N = 21 (15.3% of OAP users with NMS) were OAP users, and N = 5 (14.3% of LAI users with NMS) were LAI users. No differences by formulation were detected in mortality (χ ² [1, n = 172] = 0.024, P = .8778).

Incidence and Risk Factors Associated to NMS Recurrence

Of those discharged from the hospital after the first NMS episode (n = 153), n = 119 (77.8%) re-started antipsychotic treatment in outpatient care within 1 year after discharge. Antipsychotics used in outpatient care that were used as part of antipsychotic rechallenge after first NMS were mostly SGA-OAPs (84.0%), followed by FGA-OAPs (11.8%), FGA-LAIs (3.4%) and SGA-LAIs (0.8%). Of this rechallenge cohort (n = 119), N = 5 (4.2%) suffered NMS reoccurrence; N = 3 during SGA-OAP use and N = 1 during SGA-OAP+FGA-OAP use, and one additional subject had NMS reoccurrence during an inpatient care period, for which antipsychotic exposure could not be defined. No cases of NMS reoccurrence were reported during LAI use. NMS reoccurred after a median of 795 (range = 77–839) days after antipsychotic reintroduction.

Discussion

In this study, we used a large national database to assess the incidence and outcomes of NMS. Large observational studies allow investigators to dramatically increase the sample size of the population studied and thus potentially permit the study of rare syndromes that would otherwise be impossible to investigate due to the impracticality of conducting prospective studies. This point is critical for low incidence-high stakes events, such as NMS. Using this approach, we were able to identify incidence, mortality, and independent risk factors for NMS as well as to study in detail the influence of certain variables of interest, such as antipsychotic formulation and class.

As expected, the incidence of NMS was low (3.95 in the total cohort and 1.99 in the incident cohort per 10 000 person-years). Due to this relatively low incidence and its idiosyncratic nature, prospective controlled long-term studies targeting NMS are exceedingly difficult to conduct, highlighting the value of large observational cohort studies to study such infrequent events. Further, a substantial overlap with other conditions make NMS a challenging diagnosis,^30,31 and a potentially rapid clinical deterioration can make it difficult to obtain consent for prospective clinical research.

Importantly, the risk for NMS and serious outcomes following NMS did not differ between LAI vs OAP formulations. These findings, together with the results of a recently published patient-level case report/series meta-analysis²⁵ (another alternative to study rare events, such as NMS), should contribute to easing concerns about LAI-associated risk of developing NMS, further highlighting the value of this therapeutic alternative, as beyond the potential benefits already outlined in the introduction,^5,9–14 early use of LAIs in schizophrenia can also significantly delay time to hospitalization.³² However, it is also possible that patients on LAI may have received oral agents before starting LAIs, thus making the LAI cohort less susceptible to NMS. Unlike other studies,³³ we did not find an association between antipsychotic class (FGA vs SGA) and NMS, but our sample was larger, and included nationwide data from many different hospitals. Other similar studies did not show that association either.²⁶

We also found that recent antipsychotic change, both increased number and, to a lesser extent, decreased number/switch, as well as higher antipsychotic dose were related to an increased likelihood of developing NMS. Our findings are aligned with previous reports associating alteration of antipsychotic treatment and in particular escalation of antipsychotic dose as an important risk factor for the development of NMS.³⁴ The association between risk of NMS onset and antipsychotic decreased number/switch is more difficult to explain and may relate to changes in blood levels and/or receptor occupancy, particularly if occurring abruptly. Nonetheless, it is also possible that antipsychotics are being reduced or changed because of early symptoms of NMS.

Co-treatment with lithium was associated with NMS, as reported previously,^26,35 potentially exposing the involvement of drug-drug interactions and/or other neurotransmitter systems in the pathophysiology of NMS. Further, we found that benzodiazepines and anticholinergics were associated with NMS. In an additional analysis excluding cases who initiated use of these treatments within 7, 14, 30, or 60 days before the event to eliminate the possibility of reverse causality, results did not change, suggesting that these treatments had not been started shortly before NMS onset to treat early symptoms of NMS (eg, rigidity), but rather had been started earlier, probably intended to treat something else (eg, Parkinsonism). Interestingly, other studies found increased risk of NMS associated to benzodiazepine treatment present up to a 3-month period before NMS onset as well.²⁶ Hence, some adverse events and/or comorbid conditions present before NMS onset, such as akathisia, EPS, or other conditions for which these treatments could have been used, may themselves be risk factors for NMS, possibly indicating an underlying degree of antipsychotic and/or drug-drug interaction sensitivity. Unfortunately, due to the characteristics of these large national databases, we lacked reliable data on such comorbid diagnoses, specific treatments and/or adverse events to confirm these hypotheses. Further, we found no risk factors of a longer hospital stay related to NMS, confirming previously reported lack of association between antipsychotic formulation and longer hospitalization for NMS.²⁵

Our findings suggest that mortality figures attributed to NMS range from 4.7% during 30 days after NMS to 9.9% during 90 days following NMS, escalating to 15.1% during 1 year after NMS (although the increased mortality that long after NMS onset may be due to the higher frequency of cardiovascular conditions that was associated with NMS), being in all cases significantly higher than in control groups. No differences in formulation were detected when analyses were possible. Such results are relevant, highlighting the need for increased clinical awareness to maximize early detection and treatment of this potentially deadly complication. Unfortunately, our sample size did not allow for comprehensive characterization of mortality risk factors. Still, these results confirm earlier relevant findings using a case report/series patient-level meta-analysis approach showing no differences in NMS mortality by antipsychotic formulation.²⁵

Additionally, we found that reoccurrence of NMS among patients rechallenged with antipsychotics was 4.2%, none of which happened with LAI treatment. Although the small sample size precluded investigating specific risk factors associated with NMS recurrence, and it is possible that clinicians may have carefully curated the candidates for rechallenge, our results are aligned with recent reviews suggesting similar yet somewhat higher recurrence rates,³⁶ and highlight the need for prospective studies aimed to inform clinicians about the most appropriate treatments to minimize NMS recurrence. Still, since many of these patients can present very severe psychiatric symptoms without treatment, antipsychotic rechallenge should be seriously considered after an adequate risk/benefit analysis. In fact, although based on a small number of cases, a recent systematic review of case reports found that, at least when rechallenging patients who had developed NMS during clozapine treatment, 7 out of 7 cases were rechallenged successfully (100%, 95% CI, 56.1%–100%).³⁷ In this sense, the lack of recurrence on LAIs could be related to the fact that LAIs may possess additional safety features compared to OAPs, as antipsychotic levels and thus receptor occupancy are less prone to fluctuation compared to OAP, whose blood levels can fluctuate dramatically based on bioavailability and rely on regular daily intake.

This study has several limitations. First, coding a diagnosis of NMS depends on the accuracy of the discharging physician and/or abstractors in using the correct and appropriate diagnostic codes, which may constitute a sampling bias, often an inherent limitation of retrospective database studies. We would argue, though, that this is a conservative bias, since only the clearest cases may have been diagnosed by the providers. Second, we only included NMS resulting in hospitalization, so milder cases may have been overlooked. However, current NMS treatment guidelines, albeit diverse, suggest either hospital admission or treatments that would usually require hospital admission.³⁸ Third, although it is unrealistic to be able to randomize for a condition with such low incidence as NMS, it is possible that unmeasured prognostic factors were unevenly distributed between cases and controls, which was mitigated by nested matching with a 10:1 ratio. Fourth, since the total number of cases with NMS was small, we lacked power for certain analyses, including for rechallenge analyses. Fifth, as drugs used exclusively during inpatient care are not available in the register data, events happening during long inpatient care were not included in these analyses. Finally, national databases are limited to a specific population group, and data may not be generalizable to other national groups with, for instance, a different racial or socioeconomic profile.

Conclusion

The incidence of NMS in antipsychotic-treated individuals is low, and risk for NMS was not different between OAP and LAI formulations, confirming previous findings using an alternative research methodology. Risk factors for NMS included antipsychotic number increase and decrease/switch, albeit to a lesser extent. Other factors included higher antipsychotic dose, and co-treatment with lithium, benzodiazepines, and anticholinergics. No differences in NMS-related mortality or length of hospitalization were detected between LAI vs OAP formulations. Frequency of mortality due to NMS was 5% at 30 days and 10% and 90 days, significantly higher than in controls. NMS reoccurred in 4.2% of rechallenged individuals. These data can inform clinical decision-making and education of prescribers, patients, and families about the potential risk of antipsychotics of different formulations and classes.

Funding

None declared.