Inflammation and Response to Sertraline Treatment in Patients With CKD and Major Depression

To the Editor:

Depression affects 20% of patients with chronic kidney disease (CKD) and is associated with dialysis initiation, hospitalization, and death.^1–3 In the Chronic Kidney Disease Antidepressant Sertraline Trial (CAST), treatment with sertraline vs placebo did not improve depressive symptoms in patients with CKD not requiring kidney replacement therapy and major depressive disorder (MDD).⁴ Patients with CKD have enhanced systemic inflammation, which has been proposed as a mediator of depression in chronic medical illnesses.^1,5–7 It is unknown whether inflammatory biomarkers predict sertraline response in patients with CKD.

We investigated whether baseline inflammatory biomarkers predicted a differential decrease in depressive symptoms after 12 weeks of randomly assigned treatment with sertraline or placebo in 193 individuals with CKD stages 3 to 5 enrolled in CAST (Item S1). Plasma hsCRP, IL-6, albumin, and pre-albumin were measured at baseline. Depressive symptoms were quantified using the clinician-scored QIDS-C₁₆. Prespecified outcomes included ≥3-point decrease in QIDS-C₁₆ (improvement), ≥50% decrease in QIDS-C₁₆ (response), and QIDS-C₁₆ < 5 at exit (remission). Logistic regression assessed inflammatory biomarkers and their interactions with treatment group as predictors of outcomes, adjusting for baseline QIDS-C₁₆, study site, age, eGFR, and diabetes. Bootstrap samples were generated to assess how these models would perform as applied to new participants. In an exploratory analysis, inflammatory biomarkers were dichotomized using the highest (hsCRP, IL-6) or lowest (albumin, pre-albumin) tertile (Fig 1 legend), and a base model included baseline QIDS-C₁₆, site, age, eGFR, diabetes, and high hsCRP. The AUC was compared for nested models with χ² tests after adding each biomarker to assess for improvement in fit.

Figure 1.

Prediction of sertraline treatment response by baseline QIDS-C₁₆ score, study site, age, eGFR, diabetes mellitus, and high hsCRP (base model) with and without addition of other inflammatory biomarkers. Inflammatory biomarkers refer to the highest tertile of hsCRP (>6.0 mg/L) and IL-6 (>5.37 pg/mL) and the lowest tertile of albumin (<3.6 g/dL) and pre-albumin (<27.0 mg/dL). P is for comparison with base model.

Baseline characteristics were balanced between treatment groups, as previously published.⁴ With median values of 5.0 (IQR, 2.0–14.6) and 2.7 (IQR, 0.8–6.0) mg/dL, sertraline responders had higher hsCRP than non-responders (P = 0.03). Greater hsCRP was associated with response in the sertraline group (Table S1). There was no statistically significant difference between groups in the proportion achieving outcomes (Table 1). Greater hsCRP was associated with increased odds of improvement and response with sertraline, with P = 0.05 for the hsCRP–treatment group interaction, such that no association was observed with placebo (Table 1). The AUCs for all models were significantly greater than chance (0.5) as measured by the lower limit of the 95% confidence interval. High hsCRP (base model) predicted response in the sertraline group, AUC = 0.712 (95% CI, 0.579–0.845). Addition of all biomarkers led to the most accurate prediction, AUC = 0.821 (95% CI, 0.721–0.921), P = 0.02 vs the base model (Fig 1). No change in AUC was noted when adding biomarkers to the base model in the placebo group (Table S2). Added biomarkers did not improve prediction for improvement or remission in either group (Tables S3–S4).

Table 1.

Association of Baseline Inflammatory Biomarkers on Improvement, Response, and Remission of Depression

Outcomea	All Participants (N = 193)	Sertraline (n = 97)	Placebo (n = 96)	Biomarker × Treatment Group Interaction P	AUC (95% CI)
No. of patients
Improvement	131 (67.9%)	62 (63.9%)	69 (71.9%)
Response	55 (28.5%)	31 (32.0%)	24 (25.0%)
Remission	29 (15.0%)	15 (15.5%)	14 (14.6%)
Albumin
Improvement	aOR = 0.82 (0.44–1.54)	aOR = 1.05 (0.50–2.23)	aOR = 0.44 (0.14–1.36)	0.2	0.63 (0.59–0.66)
Response	aOR = 0.99 (0.51–1.91)	aOR = 1.31 (0.57–2.99)	aOR = 0.59 (0.20–1.73)	0.2	0.62 (0.59–0.66)
Remission	aOR = 0.73 (0.31–1.71)	aOR = 0.79 (0.28–2.22)	aOR = 0.64 (0.16–2.54)	0.5	0.68 (0.64–0.72)
Pre-albumin
Improvement	aOR = 0.93 (0.89–0.98)	aOR = 0.94 (0.88–1.00)	aOR = 0.94 (0.88–0.99)	0.4	0.67 (0.63–0.70)
Response	aOR = 0.95 (0.91–1.00)	aOR = 0.95 (0.89–1.01)	aOR = 0.95 (0.90–1.02)	0.4	0.64 (0.61–0.68)
Remission	aOR = 0.93 (0.87–0.98)	aOR = 0.93 (0.86–1.02)	aOR = 0.92 (0.84–1.00)	0.4	0.72 (0.68–0.76)
hsCRP
Improvement	aOR = 1.24 (0.94–1.63)	aOR = 1.61 (1.07–2.42)	aOR = 0.92 (0.62–1.38)	0.05	0.63 (0.59–0.67)
Response	aOR = 1.27 (0.96–1.69)	aOR = 1.57 (1.07–2.31)	aOR = 0.93 (0.59–1.47)	0.09	0.60 (0.56–0.64)
Remission	aOR = 0.83 (0.56–1.24)	aOR = 0.98 (0.59–1.66)	aOR = 0.66 (0.35–1.23)	0.3	0.66 (0.60–0.71)
IL-6
Improvement	aOR = 1.08 (0.74–1.58)	aOR = 1.15 (0.71–1.88)	aOR = 0.97 (0.52–1.80)	0.4	0.61 (0.58–0.65)
Response	aOR = 1.14 (0.76–1.70)	aOR = 1.32 (0.81–2.17)	aOR = 0.84 (0.41–1.69)	0.3	0.64 (0.60–0.67)
Remission	aOR = 1.05 (0.61–1.80)	aOR = 1.03 (0.52–2.03)	aOR = 1.08 (0.45–2.54)	0.5	0.68 (0.64–0.72)

Note: ORs and end points of the 95% CI are taken from the median of 1,000 bootstrap samples. aORs are per 1 g/dL greater albumin level, 1 mg/dL greater pre-albumin level, and 1 log-unit greater hsCRP and IL-6 levels. AUC computed by “Little bootstrap” method using 1,000 bootstrap samples. The 95% CIs for the AUC are the 2.5 and 97.5 percentiles of the distribution of AUCs. Models were adjusted for baseline QIDS-C₁₆ score, study site, age, eGFR, and diabetes mellitus.

Abbreviations: aOR, adjusted odds ratio; AUC, area under the receiver operating characteristic curve; CI, confidence interval; eGFR, estimated glomerular filtration rate; hsCRP, high-sensitivity C-reactive protein; IL-6, interleukin 6; QIDS-C16, 16-item Quick Inventory of Depressive Symptomatology.

^aImprovement, response, and remission defined as a QIDS-C₁₆ score decrease ≥ 3 points from baseline, decline ≥ 50% from baseline, or value ≤ 5 at study exit, respectively.

Thus, greater hsCRP was associated with both improvement and response in those treated with sertraline but not placebo. Inflammatory biomarkers had additive effects for sertraline response prediction, with no similar predictive capacity in those who received placebo. In vitro research and research in humans has suggested that sertraline may have anti-inflammatory effects,^8,9 which could explain these findings. Although studies in patients without CKD showed decreases in IL-6 after treatment with antidepressants,⁹ we did not find an association of IL-6 with sertraline response. Because hsCRP is a sensitive but nonspecific inflammation marker, it is possible that its elevation is due to pathways independent of IL-6 that may be influenced by sertraline. Other studies presented heterogeneous results,⁹ but suggest that a chronic inflammatory state such as CKD may affect relationships between inflammation and antidepressant treatment response.

We demonstrate that the addition of clinically useful cutoffs of inflammatory biomarkers improves prediction of sertraline response. Including all biomarkers in the model produced the strongest predictive accuracy for sertraline response (AUC, 0.821; 95% CI, 0.721–0.921; P = 0.02), but addition of IL-6 did not improve the model fit over hsCRP, albumin, and pre-albumin alone (P = 0.9) (Figure 1). The most parsimonious model for identifying individuals more likely to benefit from sertraline may not need to include IL-6, which is expensive and not readily clinically available. These findings were not replicated in the placebo group, suggesting that the predictive ability of these biomarkers may be related to sertraline treatment.

Our power was limited by few participants achieving remission to determine whether inflammatory biomarkers were associated with this important outcome. We did not measure IL-1β or TNF-α, reported to be associated with depression in patients without CKD.¹⁰ Findings due to chance related to multiple comparisons cannot be ruled out. Our novel findings are hypothesis generating and should be corroborated in larger studies to determine clinical utility.

In conclusion, among individuals with CKD and MDD, elevated plasma hsCRP independently predicted response to sertraline but not to placebo. Trials enrolling participants with CKD with elevated hsCRP may be more likely to identify a benefit of sertraline over placebo. Predictive models incorporating elevated hsCRP and low albumin and pre-albumin may be useful for identifying individuals more likely to benefit from sertraline. These findings need to be validated in future studies.