A Drug-Wide Association Study on Medications Associated With Cystatin C Estimated GFR

To the Editor:

Cystatin C is increasingly used to estimate glomerular filtration rate (eGFRcys) in settings where eGFR from creatinine (eGFRcr) may be less accurate.¹ Optimal use of any eGFR equation requires understanding of the factors that influence the marker independent of changes in kidney filtration. Non-GFR determinants of cystatin C are less well understood than for creatinine. Prior studies suggest that cystatin C is generated in high cell turnover states such as inflammation, hyperthyroidism, and cancer.² Medications may also affect levels of cystatin C independent of their effect on GFR, which is important to understand for use of GFR for many treatment decisions. We leveraged a large population-based health care cohort in Sweden, where cystatin C testing has been routine for over two decades, to systematically investigate associations between medication use and eGFRcys in a hypothesis-generating drug-wide association study (DWAS).

The Stockholm Creatinine Measurements (SCREAM) project includes laboratory data, diagnoses, pharmacy dispensations, and vital status from 2006 to 2021.³ We identified 201,499 adult outpatients who had cystatin C and creatinine measured on the same day (Figure 1). We first conducted a phenotype-wide association study to identify conditions associated with eGFRcys independent of eGFRcr. For the DWAS, we considered medications other than vitamins, nutritional supplements or topical formulations that were dispensed in at least 1% of the study population. Linear regression models were used to estimate associations between medication use (grouped by ATC codes) and eGFRcys, adjusting for age, sex, eGFRcr, and identified comorbidities. ATC families associated with the clinically meaningful threshold of ≥ 5 ml/min/1.73m² difference in eGFRcys were selected for further analysis. To further minimize confounding, for each ATC family, we repeated analyses within the disease condition most commonly prescribed (e.g., persons with heart failure for high-ceiling diuretics, Item S1, Tables S1–S3).

Figure 1: Flow chart for overall study population and more in-depth analysis.

The green box depicts the overall study population that was used for identification of top ATC families. The blue boxes depict the five disease populations used for our more in-depth analyses used to identify individual medications. For patients with cancer, we explored individual medications within the ATC families of “systemic corticosteroids” and “opioids”. For patients with gout, we explored individual medications within the ATC families of “antigout preparations”. For patients with heart failure, we explored individual medications within the ATC families of “high-ceiling diuretics”. For patients with pain or disability, we explored individual medications within the ATC families of “other analgesics & antipyretics”. For patients with bipolar disorder, we explored individual medications within the ATC families of “antipsychotics”.

Median age was 61 years, and median eGFRcys was 79 ml/min/1.73 m² (Table S4). Older age and female sex were associated with lower level of eGFRcys. After adjustment for age, sex, and eGFRcr, conditions associated with lower eGFRcys included: presence of anemia, heart failure, obesity, COPD, hypertension, cancer, diabetes mellitus, pain, atrial fibrillation, hypothyroidism, and coronary heart disease (Table S5).

Of the 36 ATC families, 25 were significantly associated with eGFRcys after adjustment for comorbid conditions and multiple comparisons (Figure S1). The six ATC families with effect sizes ≥ 5 ml/min/1.73 m² were corticosteroids, high-ceiling diuretics, opioids, other analgesics, anti-gout preparations, and antipsychotics (Table S6). Table S6 lists other ATC families that were associated with eGFRcys but with effect sizes below the pre-defined threshold. In subgroups with disease conditions most common among users of each of the six ATC families, fifteen individual medications were significantly associated with lower eGFRcys (Table 1).

Table 1:

Unique medications associated with eGFRcys within cohorts of patients with their most likely indication for use

ATC Family code	ATC Family Name	Cohort with main indication (Sample size)	Individual Medications
			Unique ATC code	Drug Name	N (%)	Effect Size	P-Value
H02A	Corticosteroids for Systemic Use	Cancer (29,951)	H02AB01	betamethasone	2363 (7.9%)	−6.08	4.63×10−72
			H02AB06	prednisolone	828 (2.8%)	−6.94	3.71×10−36
			H02AB07	prednisone	81 (0.3%)	−4.55	8.97×10−3
C03C	High-Ceiling Diuretics	Heart Failure (19,873)	C03CA01	furosemide	6156 (31.0%)	−4.57	2.09×10−117
N02A	Opioids	Cancer (29,951)	N02AA01	morphine	503 (1.7%)	−7.49	3.97×10−26
			N02AA05	oxycodone	1877 (6.3%)	−7.72	6.43×10−87
			N02AA55	oxycodone and naloxone	507 (1.7%)	−3.00	3.61×10−5
			N02AE01	buprenorphine	156 (0.5%)	−5.02	6.36×10−5
			N02AB03	fentanyl	122 (0.4%)	−6.37	7.60×10−6
N02B	Other Analgesics & Antipyretics	Pain/Disability (68,786)	N02BE01	Paracetamol/ac etaminophen	10,128 (14.7%)	−4.72	7.52×10−167
M04A	Antigout Preparations	Gout (2590)	No significant medications
N05A	Antipsychotics	Bipolar disorder (6512)	N05AX08	risperidone	91 (1.5%)	−3.36	4.67×10−2
			N05AH03	olanzapine	450 (7.2%)	−3.00	1.25×10−4
			N05AN01	lithium	1997 (32.0%)	−1.37	1.80×10−3
			N05AD01	haloperidol	70 (1.1%)	−3.99	3.83×10−2
			N05AA02	levomepromazi ne	109 (1.7%)	−4.82	1.88×10−3

Multivariable analyses include age, sex, eGFRcr, comorbid conditions and all other medications within that ATC family.

We identified associations of eGFRcys with three formulations of exogenous glucocorticoids – betamethasone, prednisolone and prednisone. This is consistent with prior investigations that also found higher levels of cystatin C with exogenous glucocorticoids or excess endogenous production.^4–7 In support of direct biological mechanisms, in one study, glucocorticoids was associated with an increase in cystatin C secretion or gene expression in macrophages and cancer cells⁸. These findings were more apparent in disease than in health states, suggesting that glucocorticoid induction of cystatin C generation may be dependent upon the setting.⁸ The associations of lithium, opioids and diuretics with lower eGFRcys have not been previously discussed. One study of patients on lithium treatment undergoing iohexol clearance tests observed worse agreement between eGFRcys and measured GFR than for eGFRcr, suggesting that our results for associations of eGFRcys with lithium might be a true effect.⁹ No prior studies comment on the associations between eGFRcys to diuretics or opioids. Our findings might indicate impact of these medications on tubular handing of cystatin C or cystatin generation, respectively but requires further investigations.

This study comprehensively assessed the associations of medications with eGFRcys separate from the effect of the chronic conditions that also affect the cystatin C. The comprehensive nature of our assessment is a strength, and we are able to generate hypotheses for future research. Nevertheless, the study has limitations. Most importantly, the results are limited by observational study design. Despite adjustment, residual confounding by indication or disease severity is possible. Because GFR was not measured, we used eGFRcr as a proxy. Serious illness such as cancer, heart failure or medications such as prolonged steroids use, can affect serum creatinine independent of GFR (e.g. effects on muscle mass or tubular secretion of creatinine). In such cases, our findings might also be indicative of changes on creatinine rather than on cystatin C itself.

Our results identify drug classes that may influence cystatin C levels. Optimal study designs to investigate these results would assess both cystatin C and measured GFR before and after initiating these medications. Until then, clinicians should be cautious in interpreting eGFRcys that decreased following initiation of these medications.

Supplementary Material

Supplementary File (PDF)

Item S1, Figure S1, Tables S1–S6.

Data Sharing:

The data underlying this article cannot be shared publicly due to the privacy of individuals that participated in the study. The data may be shared on reasonable request for academic research collaborations that fulfill GDPR, as well as national and institutional ethics regulations and standards by contacting JJC (juan.jesus.carrero@ki.se).