Comparative Safety of Antidepressants in Adults with CKD

Visual Abstract

Abstract

Background

Depression is prevalent in patients with CKD and is related to poor prognosis. Despite the widespread use of antidepressants in the CKD population, their safety remains unclear.

Methods

We identified adults with CKD stages G3–5 (eGFR <60 ml/min per 1.73 m² not treated with dialysis) and incident depression diagnosis during 2007–2019 from the Stockholm Creatinine Measurements project. Using the target trial emulation framework, we compared the following treatment strategies: (1) initiating versus not initiating antidepressants, (2) initiating mirtazapine versus selective serotonin reuptake inhibitors (SSRIs), and (3) initiating SSRIs with a lower dose versus a standard dose.

Results

Of 7798 eligible individuals, 5743 (74%) initiated antidepressant treatment. Compared with noninitiation, initiation of antidepressants was associated with higher hazards of short-term outcomes, including hip fracture (hazard ratio [HR], 1.23; 95% confidence interval [CI], 0.88 to 1.74) and upper gastrointestinal bleeding (HR, 1.38; 95% CI, 0.82 to 2.31), although not statistically significant. Initiation of antidepressants was not associated with long-term outcomes, including all-cause mortality, major adverse cardiovascular event, CKD progression, and suicidal behavior. Compared with SSRIs, initiation of mirtazapine was associated with a lower hazard of upper gastrointestinal bleeding (HR, 0.52; 95% CI, 0.29 to 0.96), but a higher hazard of mortality (HR, 1.11; 95% CI, 1.00 to 1.22). Compared with the standard dose, initiation of SSRIs with a lower dose was associated with nonstatistically significantly lower hazards of upper gastrointestinal bleeding (HR, 0.68; 95% CI, 0.35 to 1.34) and CKD progression (HR, 0.80; 95% CI, 0.63 to 1.02), but a higher hazard of cardiac arrest (HR, 2.34; 95% CI, 1.02 to 5.40).

Conclusions

Antidepressant treatment was associated with short-term adverse outcomes but not long-term outcomes in people with CKD and depression.

Introduction

Depression affects 20%–25% of patients with CKD¹ and is associated with adverse health outcomes, including progression of CKD, cardiovascular events, and mortality.^2–4 Research suggests that negative behaviors (e.g., unhealthy lifestyle and medication nonadherence)^5,6 and pathophysiological disturbances (e.g., increased inflammation and metabolic dysregulation)^7,8 could contribute to the association.

Antidepressants are the main pharmacological treatment for depression, and some agents may be used to treat other conditions, such as anxiety disorders and chronic pain. However, clinical trials in the population with CKD are inconclusive on the efficacy and safety of selective serotonin reuptake inhibitors (SSRIs).^9–11 Notably, the largest randomized controlled trial to date—the CKD Antidepressant Sertraline Trial that included 201 participants—found no benefit of sertraline over placebo in improving depressive symptoms after 12-week treatment.¹² The lack of evidence leaves clinicians in a dilemma regarding the management of depression in patients with CKD, potentially resulting in suboptimal care.¹³ Nevertheless, 10%–20% of patients with CKD received antidepressants, irrespective of underlying indications.^14–16 Given the substantial burden related to depression and frequent use of antidepressants in people with CKD, it is imperative to investigate the safety of antidepressant treatment.

Existing clinical trials are underpowered to identify rare but severe safety outcomes.^9–12 Nonetheless, large observational studies have documented some serious adverse events associated with SSRI use in patients with CKD, including hip fracture, gastrointestinal bleeding, and sudden cardiac death.^17–19 To avoid adverse drug reactions, clinical guidelines and expert opinions suggest reducing the initial dose when prescribing SSRIs to patients with CKD,^20,21 but whether this dose adjustment enhances drug safety is unclear. Furthermore, there are few data on the safety profile of other commonly used non-SSRI antidepressants, such as mirtazapine.

In this study, we aimed to evaluate the comparative safety of antidepressant treatment (including drug initiation, type, and dosage) in patients with non–dialysis-dependent CKD stages G3–5, using the target trial emulation approach.^22,23

Methods

Data Sources

This study used data from the Stockholm CREAtinine Measurements project,²⁴ a health care utilization cohort that included all residents in the region of Stockholm during 2006–2019. Using the unique personal identification number, laboratory tests from routine clinical care were linked to several regional and national health registers, including the Stockholm regional health care data warehouse, the Prescribed Drug Register, the Cause of Death Register, and the Swedish Renal Register, for complete information on demographics, clinical diagnoses, prescribed drugs, vital status, and KRT end points (full description in Supplemental Methods). The study was approved by the Regional Ethics Review Board in Stockholm (Dnr: 2017/793-31) and adhered to the Declaration of Helsinki.

Eligibility Criteria

The protocol of the target trial is outlined in Supplemental Table 1. We included all adults (≥18 years) with stages G3–5 CKD (eGFR <60 ml/min per 1.73 m² not treated with dialysis) during January 2007 to December 2019, who received an incident diagnosis of depression (International Classification of Diseases Tenth Revision code F32–F33). CKD was defined as at least one outpatient eGFR <60 ml/min per 1.73 m².²⁵ Individuals with a single eGFR <60 ml/min per 1.73 m² but all subsequent eGFR ≥60 ml/min per 1.73 m² were excluded. eGFR was calculated with the Chronic Kidney Disease Epidemiology Collaboration 2009 equation without the race coefficient,²⁶ using isotope dilution mass spectrometry standardized serum/plasma creatinine tests. Baseline was defined as the date of depression diagnosis. To identify new users of antidepressants, we required a washout period without any previous antidepressant prescription for at least 18 months. We excluded individuals with a history of KRT (dialysis or kidney transplantation), those prescribed multiple antidepressants at initiation, and those who emigrated before baseline (Figure 1). Each eligible individual was followed until death, emigration, or end of study (December 31, 2019), with virtually no loss to follow-up.

Figure 1.

Flow chart of the cohort selection. SSRI, selective serotonin reuptake inhibitor.

Treatment Strategies

We emulated three target trials to compare the following treatment strategies: (1) initiating versus not initiating antidepressant treatment, (2) initiating mirtazapine versus SSRI treatment, and (3) initiating SSRI treatment with a lower dose versus a standard dose. To reflect actual clinical practice, we allowed a grace period²² of 3 months from baseline to initiation of antidepressant treatment. Starting dose of SSRIs was estimated from free-text drug prescriptions using a machine learning algorithm²⁷ and converted to defined daily dose. According to existing recommendations,²⁸ a threshold of <1 defined daily dose per day was used to define a lower dose of SSRIs (0.5 defined daily dose accounted for 96%), and the rest was defined as the standard dose (1 defined daily dose accounted for 98%).

Study Outcomes

The short-term (1-year) outcomes included hip fracture, upper gastrointestinal bleeding, and sudden cardiac arrest, on the basis of previous findings.^17–19 The long-term (5-year) outcomes included all-cause mortality, major adverse cardiovascular event (MACE), CKD progression, and suicidal behavior. MACE was defined as the composite of myocardial infarction, stroke, or cardiovascular death. CKD progression was defined as the composite of >40% decline in eGFR, initiation of KRT, or death due to CKD (with International Classification of Disease-10 codes N18–N19 as the primary cause of death recorded in the Cause of Death Register). To define a sustained 40% decline in eGFR,²⁵ we estimated the eGFR slope with linear regression using all outpatient eGFR values from baseline to the end of follow-up.²⁹ Suicidal behavior included suicide attempts and deaths from suicide. Detailed definitions of the outcomes are listed in Supplemental Table 2.

Target Trial Emulation

We used the cloning, censoring, and weighting method³⁰ to emulate a target trial comparing the strategies “initiating antidepressants within 3 months” versus “refraining from taking antidepressants” in patients with CKD with incident depression (Supplemental Methods and Supplemental Figure 1). In brief, we created a dataset with two copies of each eligible individual (cloning step) and assigned each copy to a different strategy. Then duplicates were artificially censored if they deviated from their assigned treatment strategy (censoring step). In the main analysis, we did not impose any specific requirement on the treatment pattern after the grace period, analogous to an intention-to-treat analysis (i.e., the effect of initially starting antidepressant treatment irrespective of treatment nonadherence).³¹

The artificial censoring is informative and therefore introduces selection bias, which was adjusted for using inverse probability weighting (weighting step).³² To estimate the weights, we fitted a Cox regression model for each treatment arm separately,³⁰ including all potential time-fixed and time-varying confounders. Time-fixed confounders included age, sex, education, marital status, disposable income, calendar year, duration of CKD, nursing home care, and previous health care utilization. Time-varying confounders included eGFR, comorbidities, and concurrent use of medications (see covariates in Table 1). When analyzing the short-term outcomes and suicidal behavior, a history of these conditions was also included in the corresponding weighting model. Definitions of comorbidities and medications are detailed in Supplemental Table 3. All covariates had complete data except for education (3% missing), wherein a missing indicator was used.

Table 1.

Baseline and follow-up characteristics of patients with CKD with incident depression

Characteristics	Overall (n=7798)	Noninitiators of Antidepressants (n=2055)	Initiators of Antidepressants (n=5743)
Baseline characteristics
Age, yr, median (IQR)	80 (72–86)	76 (66–84)	81 (74–86)
Female, n (%)	4765 (61)	1252 (61)	3513 (61)
Calendar year, n (%)
2007–2011	2328 (30)	555 (27)	1773 (31)
2012–2015	2808 (36)	718 (35)	2090 (36)
2016–2019	2662 (34)	782 (38)	1880 (33)
Time interval between CKD and depression, yr, n (%)
<1	1883 (24)	533 (26)	1350 (24)
1–2	2147 (28)	579 (28)	1568 (27)
3–4	1554 (20)	404 (20)	1150 (20)
≥5	2214 (28)	539 (26)	1675 (29)
Educational attainment, n (%)
Compulsory education	2504 (32)	593 (29)	1911 (33)
Secondary education	3096 (40)	816 (40)	2280 (40)
College/university	1977 (25)	580 (28)	1397 (24)
Missing	221 (3)	66 (3)	155 (3)
Marital status, n (%)
Single	821 (11)	289 (14)	532 (9)
Married	3092 (40)	792 (39)	2300 (40)
Divorced	1648 (21)	491 (24)	1157 (20)
Widowed	2237 (29)	483 (24)	1754 (31)
Household disposable income, ×100 SEK, median (IQR)	2262 (1517–3548)	2280 (1490–3733)	2256 (1522–3489)
eGFR, ml/min per 1.73 m2, median (IQR)	57 (49–67)	58 (50–68)	57 (48–66)
Health care use in the previous year, n (%)
Any hospitalization	3722 (48)	945 (46)	2777 (48)
Any emergency department visit	1625 (21)	486 (24)	1139 (20)
Nursing home care, n (%)	290 (4)	79 (4)	211 (4)
Physical comorbidities, n (%)
Obesity	696 (9)	215 (10)	481 (8)
Hypertension	6557 (84)	1652 (80)	4905 (85)
Diabetes mellitus	1828 (23)	478 (23)	1350 (24)
Myocardial infarction	1145 (15)	319 (16)	826 (14)
Stroke	2026 (26)	515 (25)	1511 (26)
Congestive heart failure	867 (11)	190 (9)	677 (12)
Peripheral vascular disease	1739 (22)	379 (18)	1360 (24)
Cancer	1930 (25)	476 (23)	1454 (25)
Lung disease	1723 (22)	461 (22)	1262 (22)
Liver disease	355 (5)	120 (6)	235 (4)
Thyroid disease	1334 (17)	342 (17)	992 (17)
Pruritus	657 (8)	179 (9)	478 (8)
Neuropsychiatric comorbidities, n (%)
Dementia	533 (7)	112 (5)	421 (7)
Substance use disorders	709 (9)	218 (11)	491 (9)
Psychotic disorders	109 (1)	49 (2)	60 (1)
Manic episode/bipolar disorder	56 (1)	39 (2)	17 (0.3)
Anxiety disorders	1056 (14)	288 (14)	768 (13)
Insomnia	1285 (16)	302 (15)	983 (17)
Chronic neuropathic pain	809 (10)	206 (10)	603 (10)
Tobacco-related disorders, n (%)	375 (5)	109 (5)	266 (5)
Alcohol-related disorders, n (%)	489 (6)	150 (7)	339 (6)
Medications, n (%)
Renin–angiotensin system inhibitors	3927 (50)	953 (46)	2974 (52)
β-Blockers	3642 (47)	888 (43)	2754 (48)
Calcium channel blockers	1992 (26)	500 (24)	1492 (26)
Diuretics	2979 (38)	689 (34)	2290 (40)
Statins	2512 (32)	594 (29)	1918 (33)
Antiplatelet drugs	2722 (35)	612 (30)	2110 (37)
Proton-pump inhibitors	2165 (28)	570 (28)	1595 (28)
Nonsteroidal anti-inflammatory drugs	1033 (13)	281 (14)	752 (13)
Opioids and pain medications	1875 (24)	454 (22)	1421 (25)
Antiepileptic drugs	338 (4)	88 (4)	250 (4)
Antipsychotics	196 (3)	70 (3)	126 (2)
Anxiolytics, hypnotics, and sedatives	3922 (50)	824 (40)	3098 (54)
Antidementia drugs	196 (3)	27 (1)	169 (3)
Drugs used in addictive disorders	79 (1)	24 (1)	55 (1)
Follow-up characteristics
Hip fracture, n (%)	218 (3)	41 (2)	177 (3)
Follow-up, yr, mean (SD)	0.9 (0.3)	0.8 (0.3)	0.9 (0.3)
Upper gastrointestinal bleeding, n (%)	87 (1)	18 (1)	69 (1)
Follow-up, yr, mean (SD)	0.9 (0.3)	0.8 (0.3)	0.9 (0.2)
Cardiac arrest, n (%)	83 (1)	19 (1)	64 (1)
Follow-up, yr, mean (SD)	0.9 (0.3)	0.8 (0.3)	0.9 (0.2)
All-cause mortality, n (%)	2616 (34)	629 (31)	1987 (35)
Follow-up, yr, mean (SD)	3.0 (1.8)	2.8 (1.9)	3.1 (1.8)
MACEs, n (%)	1948 (25)	433 (21)	1515 (26)
Follow-up, yr, mean (SD)	2.7 (1.8)	2.6 (1.9)	2.8 (1.8)
CKD progression, n (%)	655 (8)	161 (8)	494 (9)
Follow-up, yr, mean (SD)	2.9 (1.8)	2.8 (1.9)	3.0 (1.8)
Suicidal behavior, n (%)	137 (2)	44 (2)	93 (2)
Follow-up, yr, mean (SD)	3.0 (1.8)	2.8 (1.9)	3.1 (1.8)

IQR, interquartile range; MACE, major adverse cardiovascular event; SEK, Swedish Krona.

We emulated two more trials to examine the comparative safety of different types and dosages of antidepressants in patients with CKD.³³ Among antidepressant initiators, we compared the strategies “initiating mirtazapine versus SSRIs.” Among SSRI initiators, we further compared the strategies “initiating SSRIs with a lower dose versus a standard dose.”^20,21 For these two comparisons, stabilized inverse probability of treatment weighting was used to account for differences in baseline characteristics (at the date of drug dispensation) between the treatment groups.

Covariate balance was evaluated using the standardized mean difference between treatment arms. We estimated the hazard ratios (HRs) for the association between antidepressant treatment and outcomes using weighted Cox proportional hazards models. Adjusted survival curve was estimated in each arm using a weighted Kaplan–Meier estimator,³⁴ from which absolute risk and risk difference were obtained. For all estimates, 95% confidence intervals (CIs) were computed using nonparametric bootstrap with 500 replicates. To provide a more comprehensive interpretation of research findings, we highlighted the results that reached either the statistical significance (two-tailed P values < 0.05) or a specified magnitude of the association (HRs with ≥20% deviation from the null value).³⁵ All statistical analyses were performed using R software version 4.0.5 (R Foundation for Statistical Computing).

Sensitivity Analyses

We conducted several sensitivity analyses: (1) As switches between treatment strategies were common during follow-up, we estimated the per-protocol effect (i.e., the effect of sustained antidepressant treatment accounting for treatment nonadherence),³⁶ whereby individuals (or duplicates) were censored if they deviated from their assigned treatment strategy (after the grace period); (2) we applied an alternative grace period length of 1 or 6 months to define the initiation of antidepressant treatment; (3) we examined the comparative safety of SSRIs by comparing SSRIs with higher potential for prolonging the QT interval (citalopram and escitalopram) with SSRIs with lower QT-prolonging potential (fluoxetine, paroxetine, and sertraline)¹⁹; and (4) when examining the effect of SSRI starting dose, we excluded individuals with a daily dose other than 0.5 or 1 defined daily dose.

Results

Patient Characteristics

We identified 7798 eligible individuals (median age 80 years, 61% female) with CKD and incident depression (Figure 1). Baseline characteristics are summarized in Table 1. Hypertension (84%), stroke (26%), cancer (25%), and diabetes (23%) were the most common physical comorbidities. The prevalence of anxiety disorders, insomnia, and chronic neuropathic pain was above 10%. Concurrent use of renin–angiotensin system inhibitors (50%), β-blockers (47%), or anxiolytics, hypnotics, and sedatives (50%) was prevalent. In the original study cohort, initiators of antidepressants, compared with noninitiators, were on average older and had a higher proportion of use of anxiolytics, hypnotics, and sedatives. For short-term outcomes, we identified 218 cases (3%) of hip fracture, 87 (1%) upper gastrointestinal bleeding, and 83 (1%) cardiac arrest; for long-term outcomes, we observed 2616 (34%) deaths, 1948 (25%) MACEs, 655 (8%) CKD progression, and 137 (2%) suicidal behavior.

Initiation of Antidepressants versus Noninitiation

After cloning, 7798 duplicates were assigned to each treatment strategy. During the 3-month grace period, 5743 patients (74%) initiated and 1820 patients (23%) did not initiate antidepressant treatment. The remaining 3% died or emigrated, who contributed events to both treatment arms to avoid immortal time bias. The distribution of weights is shown in Supplemental Table 4. After weighting, the standardized mean differences for all covariates were below 0.1, indicating good covariate balance (Supplemental Table 5).

In the intention-to-treat analysis, compared with noninitiation, initiation of antidepressant treatment was associated with greater hazards of hip fracture (HR, 1.23; 95% CI, 0.88 to 1.74) and upper gastrointestinal bleeding (HR, 1.38; 95% CI, 0.82 to 2.31) within 1-year follow-up (Table 2), although not statistically significant. Antidepressant initiation was not associated with all-cause mortality (HR, 1.01; 95% CI, 0.92 to 1.11), MACE (HR, 1.08; 95% CI, 0.97 to 1.21), CKD progression (HR, 1.03; 95% CI, 0.86 to 1.25), or suicidal behavior (HR, 0.81; 95% CI, 0.56 to 1.18) over 5-year follow-up.

Table 2.

Association between antidepressant initiation and health outcomes among patients with CKD with incident depression

Outcome	Weighted Events	Weighted Person-Years	Incidence Rate per 1000 Person-Years	HR (95% CI)	Absolute Risk, % (95% CI)	Risk Difference, % (95% CI)
Short-term outcomes (1 yr)
Hip fracture
Noninitiation	51	1897	27.1	Reference	2.5 (1.8 to 3.5)	Reference
Initiation	186	5566	33.5	1.23 (0.88 to 1.74)	3.1 (2.7 to 3.6)	0.6 (−0.3 to 1.5)
Upper gastrointestinal bleeding
Noninitiation	19	1910	10.0	Reference	0.9 (0.6 to 1.5)	Reference
Initiation	77	5619	13.6	1.38 (0.82 to 2.31)	1.3 (1.0 to 1.6)	0.3 (−0.2 to 0.9)
Cardiac arrest
Noninitiation	24	1913	12.4	Reference	1.2 (0.7 to 1.9)	Reference
Initiation	72	5648	12.7	0.99 (0.58 to 1.69)	1.2 (0.9 to 1.5)	0.0 (−0.6 to 0.6)
Long-term outcomes (5 yr)
Death
Noninitiation	693	6227	111.4	Reference	40.8 (37.9 to 43.7)	Reference
Initiation	2067	18,137	114.0	1.01 (0.92 to 1.11)	42.1 (40.6 to 43.6)	1.3 (−1.9 to 4.6)
MACE
Noninitiation	502	5737	87.4	Reference	31.3 (28.5 to 34.3)	Reference
Initiation	1575	16,443	95.8	1.08 (0.97 to 1.21)	34.4 (33.0 to 35.9)	3.1 (−0.1 to 6.4)
CKD progression
Noninitiation	170	6111	27.9	Reference	12.4 (10.5 to 14.7)	Reference
Initiation	515	17,683	29.1	1.03 (0.86 to 1.25)	13.1 (12.0 to 14.3)	0.7 (−1.7 to 3.1)
Suicidal behavior
Noninitiation	43	6166	7.0	Reference	3.4 (2.4 to 4.7)	Reference
Initiation	103	17,972	5.7	0.81 (0.56 to 1.18)	2.6 (2.1 to 3.2)	−0.8 (−2.0 to 0.5)

Analyses were adjusted through inverse probability weighting for age, sex, education, marital status, disposable income, calendar year, duration of CKD, nursing home care, prior health care utilization, eGFR, comorbidities, and concurrent use of medications. Valid 95% confidence intervals were derived using nonparametric bootstrap with 500 replicates. CI, confidence interval; HR, hazard ratio; MACE, major adverse cardiovascular event.

Initiation of Mirtazapine versus SSRIs

Among 5743 antidepressant initiators, 1598 (28%) used mirtazapine while 3950 (69%) used SSRIs. Mirtazapine users were older and had a higher proportion of receiving prescriptions from specialist care (Supplemental Table 6). Compared with SSRIs, mirtazapine was statistically significantly associated with a lower hazard of upper gastrointestinal bleeding (HR, 0.52; 95% CI, 0.29 to 0.96) while a higher hazard of mortality (HR, 1.11; 95% CI, 1.00 to 1.22) (Table 3).

Table 3.

Association between antidepressant type and health outcomes among new users of selective serotonin reuptake inhibitors or mirtazapine

Outcome	Weighted Events	Weighted Person-Years	Incidence Rate per 1000 Person-Years	HR (95% CI)	Absolute Risk, % (95% CI)	Risk Difference, % (95% CI)
Short-term outcomes (1 yr)
Hip fracture
SSRIs	123	3502	35.1	Reference	3.4 (2.8 to 4.0)	Reference
Mirtazapine	40	1380	29.2	0.83 (0.59 to 1.17)	2.9 (2.1 to 3.7)	−0.5 (−1.5 to 0.5)
Upper gastrointestinal bleeding
SSRIs	53	3538	14.9	Reference	1.4 (1.0 to 1.9)	Reference
Mirtazapine	11	1397	7.8	0.52 (0.29 to 0.96)	0.8 (0.4 to 1.1)	−0.7 (−1.3 to −0.1)
Cardiac arrest
SSRIs	45	3559	12.6	Reference	1.2 (0.8 to 1.7)	Reference
Mirtazapine	18	1402	12.5	0.99 (0.56 to 1.75)	1.2 (0.7 to 1.8)	0.0 (−0.7 to 0.7)
Long-term outcomes (5 yr)
Death
SSRIs	1374	12,219	112.4	Reference	42.3 (40.6 to 44.0)	Reference
Mirtazapine	583	4680	124.5	1.11 (1.00 to 1.22)	46.6 (43.3 to 50.0)	4.3 (0.6 to 8.0)
MACE
SSRIs	1028	11,011	93.3	Reference	34.7 (32.9 to 36.6)	Reference
Mirtazapine	412	4173	98.7	1.05 (0.93 to 1.18)	36.8 (33.2 to 40.5)	2.1 (−1.8 to 6.0)
CKD progression
SSRIs	335	11,933	28.1	Reference	13.0 (11.6 to 14.4)	Reference
Mirtazapine	141	4538	31.0	1.10 (0.88 to 1.39)	14.6 (11.7 to 17.4)	1.5 (−1.7 to 4.7)
Suicidal behavior
SSRIs	61	12,131	5.0	Reference	2.5 (1.9 to 3.2)	Reference
Mirtazapine	26	4627	5.6	1.12 (0.59 to 2.14)	2.6 (0.8 to 4.5)	0.1 (−1.9 to 2.1)

Analyses were adjusted through inverse probability weighting for age, sex, education, marital status, disposable income, calendar year, duration of CKD, nursing home care, prior health care utilization, eGFR, comorbidities, concurrent use of medications, and source of antidepressant prescription. Valid 95% confidence intervals were derived using nonparametric bootstrap with 500 replicates. CI, confidence interval; HR, hazard ratio; MACE, major adverse cardiovascular event; SSRIs, selective serotonin reuptake inhibitors.

Initiation of SSRIs with a Lower versus Standard Dose

Among 3950 SSRI initiators, 2716 (69%) started with a lower dose, in contrast to 1234 (31%) with a standard dose. Citalopram and escitalopram users were more likely to start with a lower dose than sertraline users (Supplemental Table 7). Compared with the standard dose, initiation of SSRIs with a lower dose was associated with nonstatistically significantly lower hazards of upper gastrointestinal bleeding (HR, 0.68; 95% CI, 0.35 to 1.34) and CKD progression (HR, 0.80; 95% CI, 0.63 to 1.02), but a higher hazard of cardiac arrest (HR, 2.34; 95% CI, 1.02 to 5.40) (Table 4).

Table 4.

Association between selective serotonin reuptake inhibitor starting dose and health outcomes among new users of selective serotonin reuptake inhibitors

Outcome	Weighted Events	Weighted Person-Years	Incidence Rate per 1000 Person-Years	HR (95% CI)	Absolute Risk, % (95% CI)	Risk Difference, % (95% CI)
Short-term outcomes (1 yr)
Hip fracture
Standard dose	43	1108	39.2	Reference	3.8 (2.6 to 5.4)	Reference
Lower dose	79	2452	32.2	0.82 (0.53 to 1.26)	3.1 (2.5 to 3.9)	−0.6 (−2.2 to 0.9)
Upper gastrointestinal bleeding
Standard dose	21	1116	19.0	Reference	1.8 (1.1 to 3.1)	Reference
Lower dose	32	2475	13.0	0.68 (0.35 to 1.34)	1.3 (0.8 to 1.9)	−0.6 (−1.7 to 0.5)
Cardiac arrest
Standard dose	6	1128	4.9	Reference	0.5 (0.2 to 1.0)	Reference
Lower dose	29	2487	11.5	2.34 (1.02 to 5.40)	1.1 (0.8 to 1.6)	0.6 (0.1 to 1.2)
Long-term outcomes (5 yr)
Death
Standard dose	393	4003	98.1	Reference	38.5 (35.2 to 41.7)	Reference
Lower dose	866	8971	96.5	0.98 (0.88 to 1.10)	38.2 (36.2 to 40.3)	−0.2 (−3.9 to 3.4)
MACE
Standard dose	321	3576	89.7	Reference	34.5 (30.9 to 38.1)	Reference
Lower dose	654	8152	80.2	0.90 (0.78 to 1.03)	31.1 (29.1 to 33.0)	−3.4 (−7.4 to 0.5)
CKD progression
Standard dose	121	3912	30.8	Reference	14.0 (11.2 to 16.7)	Reference
Lower dose	217	8748	24.8	0.80 (0.63 to 1.02)	11.9 (10.3 to 13.4)	−2.1 (−5.2 to 1.0)
Suicidal behavior
Standard dose	20	3973	5.0	Reference	2.7 (1.3 to 4.0)	Reference
Lower dose	47	8893	5.3	1.05 (0.57 to 1.94)	2.7 (1.8 to 3.5)	0.0 (−1.6 to 1.6)

Analyses were adjusted through inverse probability weighting for age, sex, education, marital status, disposable income, calendar year, duration of CKD, nursing home care, prior health care utilization, eGFR, comorbidities, concurrent use of medications, source of antidepressant prescription, and type of selective serotonin reuptake inhibitors. Valid 95% confidence intervals were derived using nonparametric bootstrap with 500 replicates. CI, confidence interval; HR, hazard ratio; MACE, major adverse cardiovascular event; SSRIs, selective serotonin reuptake inhibitors.

Sensitivity Analyses

Deviations from the assigned treatment strategy were common (Supplemental Figure 2). Per-protocol analysis yielded consistent findings with the intention-to-treat analysis, but with larger point estimates of the association (Supplemental Tables 8–10). When applying the cloning, censoring, and weighting approach, the results were robust to different lengths of grace period (Supplemental Table 11). We observed similar incidences of cardiac and other outcomes for initiation of SSRIs with higher versus lower QT-prolonging potential (Supplemental Table 12). When restricting to individuals with a prescribed SSRI dose of either 0.5 or 1 defined daily dose per day, the results remained largely unchanged (Supplemental Table 13).

Discussion

Treating depression is a significant challenge in the population with CKD, and the paucity of evidence on antidepressant treatment presents a major obstacle to providing quality care.³⁷ To fill the important knowledge gap on the comparative safety of antidepressants, our study emulated a target trial using a cohort of almost 8000 individuals with non–dialysis-dependent CKD and depression under routine care. We found that initiation of antidepressants was associated with higher risks of hip fracture and upper gastrointestinal bleeding within 1 year, but not associated with all-cause mortality, MACE, or CKD progression over 5 years. Our study underscores the importance of selecting appropriate antidepressant type and dosage to improve treatment safety. Treatment with mirtazapine versus SSRIs was associated with a lower risk of upper gastrointestinal bleeding but a higher risk of mortality. SSRI treatment with a lower versus standard dose was associated with lower risks of adverse drug reactions and CKD progression.

In line with previous studies,^17,18 we found that antidepressant use in CKD was associated with greater short-term risks of hip fracture and upper gastrointestinal bleeding, albeit with a small risk difference. Moreover, while our study suggests that antidepressant treatment in patients with CKD does not alleviate depression-related poor prognosis, such as death, MACE, and CKD progression,^2–4 it also signifies the comparatively long-term safety of such treatment. A previous study among patients with CKD instead observed a significantly higher mortality rate for current users of antidepressants,³⁸ which was susceptible to prevalent user bias and confounding by indication. Current evidence from clinical trials in this area has not shown clear efficacy of antidepressants in improving depressive symptoms but indicates a higher occurrence of symptomatic adverse effects.^9–11 It is worth noting that these trials primarily focused on the use of sertraline in patients with kidney failure, typically with a sample size <100 and a follow-up period ≤12 weeks. In the meanwhile, research has highlighted potential side effects related to the prolonged use of antidepressants, particularly in the geriatric population.³⁹ Notwithstanding the widespread use of antidepressants, their relative benefits and risks in patients with CKD are remarkably understudied,^10,11 posing a barrier to making informed treatment decisions. Findings from our comparative safety analysis should caution clinicians to carefully review the appropriateness of antidepressant prescriptions on an individual basis.

To the best of our knowledge, no previous research has examined the effects of non-SSRI antidepressants in the population with CKD. Mirtazapine is the second most frequently prescribed antidepressant, after citalopram, in several European countries,⁴⁰ partly due to a faster onset of action than SSRIs and its sedative, anxiolytic, and appetite-stimulant effects.⁴¹ Compared with SSRIs, mirtazapine use was associated with a higher risk of mortality in patients with CKD, similar to findings in the general population.^42,43 Conversely, SSRI use was associated with higher risks of hip fracture and upper gastrointestinal bleeding than mirtazapine in patients with CKD, which has previously been observed^17,18 and can possibly be attributed to its effects on bone metabolism and platelet function.^44,45 The clinical guidelines recommend against combined use of SSRIs with certain medications such as nonsteroidal anti-inflammatory drugs and antiplatelet drugs because of known drug-drug interactions.^46–48 In accordance with guideline recommendations,⁴⁶ our study suggests that mirtazapine has a lower propensity for these interactions and could be considered as an alternative when a heightened risk of SSRI-related adverse effects are perceived.

Patients with CKD often require dose adjustments for medications. Our study showed that a lower starting dose of SSRIs, as compared to the standard dose, was associated with a lower risk of adverse events such as upper gastrointestinal bleeding and progression of CKD. The observed association between a lower SSRI dose and risk of cardiac arrest may reflect confounding by contraindication.⁴⁹ In 2011, the European Medicines Agency issued warnings about the potential to prolong QT interval for citalopram and escitalopram and accordingly reduced the maximum doses in older adults.⁵⁰ Therefore, patients requiring a lower dose may have underlying conditions that are associated with a higher risk of cardiac arrest. Clinical guidelines for the treatment of depression generally suggest clinicians pay careful attention to kidney function among older patients.^46,47 Nevertheless, a notable lack of information on the pharmacokinetics, effectiveness, and safety of antidepressants in the population with CKD leads to inconsistencies in dosing recommendations.²⁸ Our findings provide support for the recommendation to initiate SSRIs at a lower dose in patients with CKD.

The current study sheds light on optimizing the management of depression in patients with CKD. On the basis of existing evidence and expert opinions,¹⁰ it is reasonable to consider initiating SSRI therapy to evaluate the improvement of depressive symptoms in patients with CKD, but dose adjustment and careful monitoring are necessary to minimize adverse events. That said, treatment decisions should be made by weighing potential benefits and risks and relying on an individualized assessment of factors including drug response and safety, patients' backgrounds, kidney function, medical history, and concomitant medications. Our study also demonstrates discrepancies between clinical guidelines and current practices. First, mirtazapine was the most prescribed antidepressant among patients with CKD, and consistent with previous literature,⁴⁰ it was more likely to be prescribed to patients of older age and with more comorbidities. The decision basis of this prescribing practice needs to be justified. Second, the combined use of SSRIs and nonsteroidal anti-inflammatory drugs or antiplatelet drugs was common, which is not recommended by guidelines.^46–48 Third, over 30% of patients with CKD initiated SSRIs without dose reduction, highlighting the suboptimal implementation of SSRI dose adjustment in routine care. Addressing these issues could ultimately improve the quality of care and treatment outcomes for this vulnerable population.

This study is the first to investigate the comparative safety of antidepressant use in patients with CKD, with a specific focus on the type and dose of antidepressants. Strengths of our study include the implementation of target trial emulation, a state-of-the-art method that helps to specify clear research questions and avoid common biases in observational studies,^22,23 as well as access to a large dataset covering all residents from the region of Stockholm, which offers universal tax-funded health care. We also acknowledge several limitations. First, there were no available measures of depressive symptoms in our register data. Given the observational nature of this study, residual and unmeasured confounding cannot be ruled out. The decision to initiate or select an antidepressant was not random but influenced by multiple complex factors, including both patient-level (e.g., patients' depressive symptoms, comorbidities, and concomitant medication use) and provider-level factors (e.g., clinicians' judgments). Second, we did not have information on lifestyle factors such as smoking or alcohol use and had to rely on proxy measures instead. Third, there are unavoidable measurement errors in identifying individuals with CKD from health registers, as well as in ascertaining outcomes.²⁵ Fourth, our results apply to patients with CKD who likely had moderate to severe depression that required clinical attention; generalization to other settings should be made with caution. Finally, to better identify potential safety issues, we considered both statistical significance and the strength of association, without adjustments for multiple testing. Owing to limited statistical power, we were unable to perform additional subgroup analyses. Further studies are needed to confirm our findings and explore treatment effect heterogeneity.

In conclusion, our study found that antidepressant treatment in patients with CKD was associated with short-term fracture and bleeding risks, while not associated with long-term mortality, cardiovascular, and kidney outcomes. To enhance the treatment safety when prescribing antidepressants to patients with CKD, it is crucial to prioritize individualized drug selection and follow the principle of lower starting dose with careful titration. Further research is needed to identify subgroups of patients with different risk-benefit profiles.

Supplementary Material

cjasn-19-178-s001.pdf ^{(601.3KB, pdf)}

Disclosures

J.J. Carrero reports research funding from Amgen, Astellas, MSD, Novo Nordisk, Swedish Heart and Lung Foundation, Swedish Research Council, and Vifor Pharma; advisory or leadership role on Advisory Committees for AstraZeneca and GSK; role on Editorial Boards for American Journal of Kidney Diseases, European Heart Journal, Journal of Nephrology, and Nephrology Dialysis Transplantation; speakers bureau for Abbott Laboratories, AstraZeneca, Baxter, Fresenius Kabi, and GSK; and other interests or relationships with European Renal Nutrition working group at the ERA-EDTA and the International Society of Renal Nutrition and Metabolism. J.J. Carrero has been a consultant, speaker, or grant recipient for Abbott, Amgen, AstraZeneca, Bayer, Fresenius, Fresenius Kabi, MSD, Nestle, and Vifor Pharma, for topics unrelated to this work. All remaining authors have nothing to disclose.

Funding

H. Xu: the Strategic Research Area Neuroscience at Karolinska Institutet, Vetenskapsrådet (2022-01428), Center for Innovative Medicine (FoUI-963369). Z. Chang: Vetenskapsrådet (2018-02213) and Loo och Hans Ostermans Stiftelse för Medicinsk Forskning. J.J. Carrero: Vetenskapsrådet (2019-01059), Hjärt-Lungfonden (20190587), Stiftelsen Stig och Gunborg Westman, and Martin Rinds Stiftelse.

Author Contributions

Conceptualization: Juan Jesús Carrero, Zheng Chang, Nanbo Zhu.

Data curation: Nanbo Zhu.

Formal analysis: Nanbo Zhu.

Funding acquisition: Juan Jesús Carrero, Zheng Chang, Hong Xu.

Investigation: Zheng Chang, Tyra Lagerberg, Nanbo Zhu.

Methodology: Zheng Chang, Tyra Lagerberg, Nanbo Zhu.

Project administration: Juan Jesús Carrero.

Supervision: Juan Jesús Carrero, Zheng Chang, Hong Xu.

Writing – original draft: Nanbo Zhu.

Writing – review & editing: Juan Jesús Carrero, Zheng Chang, Kristina Johnell, Tyra Lagerberg, Hong Xu, Nanbo Zhu.

Data Sharing Statement

Data will be available for collaborative research under reasonable request and fulfillment of GDPR regulations. For inquiries, please send your proposal to the Steering Committee of the SCREAM project (email: juan.jesus.carrero@ki.se).

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/B828.

Supplemental Methods.

Supplemental Table 1. Protocol and emulation of a target trial evaluating the effects of initiating versus not initiating antidepressants in patients with CKD and depression.

Supplemental Table 2. Definition of study outcomes.

Supplemental Table 3. Definition of comorbidities and concurrent medications.

Supplemental Table 4. Distribution of the stabilized weights at the end of the grace period.

Supplemental Table 5. Characteristics at the end of the grace period by antidepressant initiation, before and after weighting.

Supplemental Table 6. Baseline characteristics by antidepressant type, before and after weighting.

Supplemental Table 7. Baseline characteristics by SSRI starting dose, before and after weighting.

Supplemental Table 8. Association between antidepressant initiation and health outcomes following a per-protocol approach.

Supplemental Table 9. Association between antidepressant type and health outcomes following a per-protocol approach.

Supplemental Table 10. Association between SSRI starting dose and health outcomes following a per-protocol approach.

Supplemental Table 11. Association between antidepressant initiation and health outcomes applying different lengths of grace period.

Supplemental Table 12. Association between SSRI type and health outcomes by QT-prolonging potential.

Supplemental Table 13. Association between SSRI starting dose and health outcomes restricting to 0.5 or 1 defined daily dose per day.

Supplemental Figure 1. Schematic representation of cloning, censoring, and weighting method.

Supplemental Figure 2. Probability of adherence to the assigned treatment during follow-up.