Efficacy of Fluoxetine for Post-Ischemic Stroke Depression in Tanzania

Abstract

Objective:

Post-stroke fluoxetine trials are primarily conducted in high-income countries. We characterize post-ischemic stroke depression in fluoxetine-treated and -untreated study participants in urban Tanzania.

Methods:

Adults (>18 years old) within 14 days of CT-confirmed acute ischemic stroke onset were enrolled at Muhimbili National Hospital, Tanzania. The fluoxetine-treated group took 20mg fluoxetine daily for 90 days in a phase II trial and were compared to fluoxetine-untreated historical controls. The primary outcome was depression at 90 days, measured by the Patient Health Questionnaire-9 (PHQ-9). PHQ-9 scores were compared between fluoxetine-treated and -untreated groups. A score >=9 points was considered to reflect depression. A multivariable linear regression model assessed associations with post-stroke PHQ-9 scores.

Results:

Of the fluoxetine-treated (n=27) and -untreated (n=32) participants, the average age was 56.8 years old (39% women, 100% Black/African). The average presentation NIHSS score was 12.1 points and modified Rankin Scale (mRS) score was 3.5. The average mRS score at 90-day follow-up was 2.3. There was no significant difference between 90-day PHQ-9 scores in the fluoxetine-treated (mean=4.1 points, standard deviation=3.2; 11% depression) and untreated (mean=4.4, standard deviation=4.8; 19% depression) groups, p=.69. In the multivariable analysis, older age (β=0.08, p=.03) and higher NIHSS score (β=0.15, p=.04), but neither fluoxetine (β=0.57, p=.59) nor sex (β=−0.51, p=.63), were significantly associated with more depressive symptoms.

Conclusion:

Our findings parallel trials from higher income settings that fluoxetine does not significantly improve post-ischemic stroke depression, although our sample size was small. More work is needed to depict the longitudinal nature and treatment of post-stroke depression in Sub-Saharan Africa.

Introduction

A diagnosis of post-stroke depression is defined as a period of depressed mood, lasting two or more weeks after the occurrence of stroke, and the presence of at least four other symptoms of depression as characterized by the Diagnostic and Statistical Manual of Mental Disorders-5.¹ Post-stroke depression affects nearly one-third of patients who have experienced stroke.²

A meta-analysis exploring depression after stroke in Sub-Saharan Africa found a post-stroke depression prevalence of 31% in approximately 1500 stroke patients across 17 studies.³ The distinction between ischemic and hemorrhagic stroke was not made. Lower education level, poststroke cognitive impairment, physical disability, lower quality of life, and divorced marital status have each been associated with post-stroke depression in Sub-Saharan Africa. However, the associations with post-stroke depression in this context are not consistently measured or identified across studies. Of all studies in this meta-analysis from the African continent, none explored the impact of fluoxetine, a low-cost selective serotonin reuptake inhibitor (SSRI) typically indicated for depression and readily available on most Essential Medicines Lists.

Researchers from the AFFINITY trial,⁴ originally reporting on the lack of impact of fluoxetine on post-stroke motor recovery, recently reported on the post-depression outcomes in their participants.⁵ They concluded that daily fluoxetine, administered for 26 weeks post-stroke, did not significantly reduce depressive symptoms compared to placebo. Notably, AFFINITY enrolled <2% of participants of Black or African ancestry. To date, no trials have reported on post-stroke fluoxetine outcomes from Sub-Saharan Africa.

We recently completed a phase II trial - MAMBO (Measuring Ambulation, Motor, and Behavioral Outcomes) - of fluoxetine 20mg, administered for 90 days post-ischemic stroke to adults in urban Tanzania.⁶ Here, we aim to report on the secondary outcome of post-stroke depression in MAMBO’s fluoxetine-treated participants compared to historical ischemic stroke fluoxetine-untreated controls enrolled from the same Tanzanian hospital.⁷

Methods

Ethics Approvals

Ethics boards approvals were obtained from the Muhimbili National Hospital, the National Institute of Medical Research of Tanzania, the Tanzania Medicine and Medical Devices Authority, and the Partners Healthcare Institutional Review Board. Each participant or when necessary, a next of kin proxy, provided his or her own individual consent.

Setting

Tanzania has a gross national income per capita (Atlas method, 2020) of 1080 USD.⁸ All participants were prospectively recruited from the Muhimbili National Hospital in Dar es Salaam. This is the largest academic referral hospital in Tanzania with an estimated 1000–1200 patients admitted per week. The hospital has four neurologists. Stroke services available through the neurology department include basic care and thrombolysis for those who can afford it.⁹ The yearly age-standardized incidence rate of stroke in Dar es Salaam, Tanzania is presumed to be approximately 316 per 100,000 persons.¹⁰ Not all patients have health insurance and most have government funded support or family support for hospital expenses.¹¹

Participants

A flow diagram of participants is provided in Figure 1. Participants were (1) fluoxetine-treated participants in a phase II trial of fluoxetine (MAMBO; 11/2019 to 10/2020) and (2) prospectively enrolled participants from an observational cohort study (07/2016 to 03/2017) who served as historical controls.

Figure 1.

Flow of participants.

The participants were age >18 years old; admitted within 14 days of stroke symptom onset, meeting the WHO definition of acute stroke; and had a CT-confirmed ischemic stroke at the time of onset. The fluoxetine-treated group had additional exclusion criteria: dysphagia preventing pill swallowing, hyponatremia at stroke onset (<125 mmol/L), and/or hepatic impairment (serum ALT) >120 units/L). In this group, participants were administered open-label 20mg oral fluoxetine daily for 90 days. Fluoxetine for this trial was procured through the University of Iowa Pharmacy laboratory. For complete details on the inclusion and exclusion criteria for this group, see the published MAMBO protocol.⁶ Enrollment in the MAMBO trial stopped early due to COVID-19, leading to a lower than intended sample size.

Procedures

Both groups had baseline and 90-day follow up visits conducted by physicians and study staff at the Muhimbili National Hospital. NIH Stroke Scale (NIHSS) scores, sociodemographic variables, and modified Rankin Scale (mRS) scores were assessed at baseline. Adherence in the fluoxetine-treated group was assessed using a medication possession ratio, assessed in person by the investigators every 30 days. At 90-days, mRS scores were reassessed, and the Patient Health Questionnaire-9 (PHQ-9) was administered in Kiswahili or English, as preferred by the participant.

Outcome

The primary outcome reported here is the burden of depressive symptoms at the 90-day follow up visit, as measured by the PHQ-9.¹² This measure has previously been translated and validated in Kiswahili in a Sub-Saharan African setting.¹³ The PHQ-9 assesses depressive symptoms within the last two weeks with a series of 9 items, each using scores from 0 (not at all) to 3 (nearly every day) points. Overall scores range from 0–27 points. A score of >=9 points was used to classify significant depressive symptom burden, consistent with a recent large trial definition.⁵

Data Analysis

Variables were described by their means and standard deviations (SDs). Baseline characteristics between the fluoxetine-treated and -untreated groups were compared using t-tests, tests of two proportions, or chi-square tests, as appropriate. A Mann-Whitney U-test was used to compare PHQ-9 scores between fluoxetine-treated and -untreated participant groups, and a Fisher’s exact test was used to compare the proportions of participants meeting criteria for depressive symptom burden between these two groups. Ninety-five percent confidence intervals (CIs) were calculated for the proportion of participants in each group who met criteria for depressive symptom burden. A multivariable linear regression model was used to assess associations between PHQ-9 scores at 90 days as the continuous outcome variable and participant sex (binary: male or female), age (continuous), fluoxetine administration (binary: treated or untreated), and NIHSS score (continuous) as explanatory variables. No significant deviations from standard statistical assumptions were found for this model. All analyses were conducted in R (Version 1.0.153). A p-value <.05 was considered statistically significant.

Results

Participant characteristics at baseline, comparing fluoxetine-treated (n=27) and -untreated (n=32) groups, are reported in Table 1. Overall, the groups were balanced on the clinical and sociodemographic variables assessed with the exception that more fluoxetine-treated participants were employed.

Table 1.

Participant characteristics at enrollment

	Fluoxetine- treated (MAMBO participants)	Fluoxetine- untreated (Tanzanian controls)	Comparison between groups, p
n	27	32	-
Age in years, mean (standard deviation; SD)	53.7 (13.0)	59.5 (14.6)	.11
Sex, n (%)			.28
Female	8 (30)	15 (47)
Male	19 (70)	17 (53)
NIH Stroke Scale score, mean (SD)	10.5 (4.4)	13.5 (9.0)	.12
Modified Rankin Scale score, mean (SD)
At baseline	3.7 (0.9)	3.3 (1.1)	.10
At 90-day follow up	2.1 (1.0)	2.4 (1.6)	.32
Highest level of education, n (%)			.28
None	0	4 (13)
Primary school	16 (59)	18 (56)
Some secondary school	2 (7)	4 (13)
Complete secondary school	6 (22)	4 (13)
University	2 (7)	1 (3)
Employment status, n (%)			<.01
Employed	25 (93)	19 (59)
Unemployed	2 (7)	13 (41)
Comorbidities,† n yes (%)
BMI >30 kg/m2	1 (4)	6 (19)	.17
Cigarette smoking	6 (22)	3 (9)	.32
Diabetes	6 (22)	5 (16)	.75
History of depression	1 (4)	2 (6)	1
History of stroke	0	2 (6)	.12
HIV/AIDS	0	0	-
Hypertension	25 (93)	23 (72)	.09
Substance use, including alcohol	5 (19)	5 (16)	1
90-day Patient Health Questionnaire-9 (PHQ-9) score
Mean (SD)	4.1 (3.2)	4.4 (4.8)	.81
PHQ-9 score >=9, n (%)	3 (11)	6 (19)	.42

^†Comorbidities were primarily self-reported by participants.

Participants were on average 56.8 years old (SD=14.1 years). Thirty-nine percent were women, 100% were African Black, and most had a highest educational attainment level of primary school (58%, n=34). At enrollment, the average NIHSS score was 12.1 points (SD=7.4 points), and the average mRS was 3.5 points (SD=1.0 points), indicating that participants had mild to moderately severe strokes and experienced moderate to moderately severe disability. At 90 days, the average mRS score was 2.3 (SD=1.4), indicating mild disability. In the fluoxetine-treated group, adherence to fluoxetine was 96% at 90 days.

The distribution of PHQ-9 scores for both groups is shown in Figure 2. The average PHQ-9 score at 90-day follow up was 4.1 points (SD=3.2) for fluoxetine-treated patients and 4.4 points (SD=4.8) for fluoxetine-untreated patients. Eleven percent of fluoxetine-treated participants met criteria for significant depressive symptom burden (n=3; 95% CI [0,24]), compared to 19% (n=6; 95% CI [4,33]) in untreated participants. We did not observe a statistically significant difference in the mean PHQ-9 scores between the two groups, p=.69. Similarly, we did not observe a statistically significant difference in the proportion of participants who met criteria for depression by fluoxetine treatment versus not, p=.49.

Figure 2.

Comparison of 90-day Patient Health Questionnaire-9 scores between fluoxetine-treated and -untreated participants depicted by A) box plots and B) proportional bar charts.

In the multivariable linear regression model (Table 2) with PHQ-9 score as the outcome variable and age, sex, fluoxetine treatment, and NIHSS score as explanatory variables, only older age (β=0.08, p=.03) and higher baseline NIHSS score (β=0.15, p=.04) were statistically significantly associated with depressive symptom burden. There were no significant association with depressive symptom burden by sex or fluoxetine administration.

Table 2.

Multivariable linear regression statistics evaluating variables associated with post-stroke depression in Tanzania

	β	Standard error	95% confidence interval of β	p
Fluoxetine treated	0.57	1.06	[−1.55, 2.69]	.59
Age (continuous)	0.08	0.04	[0.01, 0.16]	.03*
Female sex	−0.51	1.05	[−2.61, 1.60]	.63
NIH Stroke Scale score (continuous)	0.15	0.07	[0.01, 0.29]	.04*

^*p < .05

Discussion

We develop an evidence base on the limited efficacy of fluoxetine for post-ischemic stroke depression in Sub-Saharan Africa. In our fluoxetine-treated MAMBO trial participants from Tanzania, 11% met criteria for significant depressive symptom burden at 90-day follow up.

Although we did not have a placebo arm in the MAMBO trial, using historical controls from the same Tanzanian hospital we found 19% of this fluoxetine-untreated sample met criteria for significant depressive symptom burden at 90 days. However, the difference in depressive symptom burden between these groups was not statistically significant. In both groups, higher stroke severity at onset and older age are associated with a higher burden of depressive symptoms by 90 days.

A network of clinical trials examined the role of post-stroke fluoxetine on post-stroke recovery and depression. The FLAME (France), EFFECTS (Sweden), and FOCUS (United Kingdom) trials^14-16 utilized different measures for depression, but the AFFINITY trial (Australasia) used the PHQ-9. We found a remarkably similar prevalence of depressive symptom burden in our fluoxetine-treated sample (11%) at 90 days compared to the 10.8% reported in the fluoxetine-treated arm of AFFINTY at 84 days. While 19% of our fluoxetine-untreated sample met criteria for depressive symptom burden at 90 days, only 9.7% of the placebo arm patients at 84 days in the AFFINITY trial met such criteria. This may relate to milder strokes in the AFFINITY trial, better stroke services, or other factors. Notably, our fluoxetine-treated and control participants were younger at enrollment (mean age 56.8 years in Tanzania versus 63.8 years in AFFINITY, 64.7 in FLAME; >56% of participants in both FOCUS and EFFECTS were >70 years of age) and experienced more severe stroke presentations as measured by the median NIHSS score (11 points in Tanzanian participants versus 6 points in AFFINITY, 6 points in FOCUS, and 3 points in EFFECTS; the mean NIHSS score was 13 points in FLAME).

However, like the AFFINITY trial, we also found no statistically significant difference between fluoxetine-treated and untreated post-stroke depression in our study participants in Tanzania. Although our sample size was much smaller, we provide class IV evidence that fluoxetine has no measurable benefit on post-stroke depression in unselected patients at 90 days in Sub-Saharan Africa.

The prevalence of post-stroke depression reported in our participants study is also notably less than the approximately one-third of post-stroke patients commonly reported in the literature.^2,3 There are several possible explanations for this. Our participants were well-evaluated and received better than usual care because of study enrollment, including access to free neuroimaging as well as 30-, 60-, and 90-day follow up visits by a physician. Participants may have had response bias, wanting to report fewer depressive symptoms to an investigator. In addition, there is a very high post-stroke mortality at the Muhimbili National Hospital, approximating 50%.¹⁷ Participants with the most severe strokes were more likely to die and could not be included in 90-day analyses. Another source of differences across post-stroke depression studies globally may be due to the lack of consensus on which scales should be used to accurately measure post-stroke depression.² The PHQ-9 is the most frequently used scale to assess post-stroke depression in Sub-Saharan Africa,³ and cutoff value of >=9 points indicating depressive symptom burden has been established prior to our report.⁵

Our data provide class IV evidence regarding the efficacy of fluoxetine for post-stroke depression in Tanzanian adults. Our findings are limited in that they use historical controls and derive from relatively small numbers of participants, subjecting our findings to a type II error. Another limitation of the current research is the greater loss to follow up and death in the historical fluoxetine-untreated controls compared to the fluoxetine-treated MAMBO group: it is possible that certain factors related to the death and loss to follow-up in the historical controls could be related to the depressive outcomes of participants.

Our results suggest that fluoxetine, a low-cost SSRI, may not be efficacious for treating poststroke depressive symptoms in this urban Tanzanian setting. Fluoxetine may still be efficacious in select patients or those with a clinical history of pre-stroke depression which we were unable to adequately measure in our study. Especially given the lower access in such settings to psychiatric and rehabilitative services and the low-cost and accessibility of fluoxetine, future large, multi-site randomized controlled trials should incorporate the Sub-Saharan African region to evaluate post-stroke depression more thoroughly.