Inflammation and trajectory of renal function in community-dwelling older adults

Introduction

Chronic kidney disease (CKD) is a major health burden worldwide.¹ Diabetes mellitus (DM) and essential hypertension (HTN) are established risk factors for CKD,² yet they do not fully explain the increased risk of renal function (RF) decline and CKD with aging.^{3, 4} The role of high concentrations of pro-inflammatory cytokines has drawn considerable attention because studies have shown that patients with known CKD and high inflammation experience accelerated declines in RF.^{5, 6} RF also declines with aging in the absence of major comorbidities.^7–10 Moreover, chronic inflammation is known as a feature of aging.¹¹ There is body of evidence showing associations between inflammatory biomarkers, which activate two main inflammatory pathways, and age-related diseases such as functional decline, mobility and cognititive impairments, and mortality.^11–14 Examples of such biomarkers include interleukin-6 (IL-6) and tumor necrosis factor α (TNF α) that activate nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), and IL-1β and IL-18 that are related to the nucleotide-binding oligomerization domain (NOD)-like receptor containing pyrin domain 3 (NLRP3) pathways.

To date, the results of studies that show inflammation predicts incident CKD are still controversial because of methodological limitations, including limited number of inflammatory biomarkers, not considering potentially informative censoring and loss to follow-up, or limited follow-up time.^14–18 Moreover, most of these studies were focused on patients with DM. In this study we focused on IL-6 and two soluble TNFα receptors (sTNFα-R1 and sTNFα-R2) from NF-κB pathway, IL-18 and IL-1β from NLRP3 pathway, IL-1 receptor antagonist, and HS-CRP.¹⁴

Using InCHIANTI data,¹⁹ we tested if baseline levels of these inflammatory biomarkers and their increased levels over 9 years were prospectively associated with accelerated RF decline and CKD development in nondiabetic participants who initially had normal RF.

Methods

Measurements

Statistical analyses

Participants’ characteristics are reported as means±standard deviation (SD) or median and intesrquartile range (IQR), percentages of total study population, and by incident CKD status (Table 1).

The skewed inflammatory biomarkers were natural log-transformed. Then, the biomarkers, SBP, muscle cross-sectional area, and eGFR were standardized to mean =0 and SD=1 to permit comparison of their incremental predictive associations. We used inverse probability weighted generalized estimating equations (W-GEE) to address data missing at random, after accounting for measured covariates and truncation by death.³¹ We assessed the relationship between standardized baseline inflammatory biomarkers and standardized eGFR by separate W-GEE models adjusted for muscle cross-sectional area, SBP, new-onset DM, and other covariates. Results are regression coefficients relating baseline biomarkers with eGFR at each follow-up. To assess which inflammatory biomarker had a stronger association with time-dependent eGFR or rate of change of eGFR over time, we—after evaluating collinearity (Pearson correlation, data not shown)—included all significant biomarkers in a single model adjusting for demographics, CHF, SBP, new-onset DM, and other covariates (see Methods). Then, a parsimonious model was generated by applying backward stepwise elimination of variables with p-value>0.1. We also examined possible threshold effects by comparing quartiles 2, 3, and 4 against quartile 1(reference group).

We computed the rates of change inflammatory biomarkers over time using mixed-effect models and operationalized time as a continuous variable obtaining individual time-slopes as annual rate of change. Similarly, we modeled eGFR change using a mixed-effect model and operationalized time as a continuous variable to obtain annual rate of eGFR (slope), a common RF outcome used in longitudinal studies. We reported the mean±SD of slope in incident CKD versus non-CKD. We determined if baseline inflammatory biomarkers or their standardized slopes predicted the rate of RF change after accounting for time varying conventional risk factors (SBP and DM) and baseline RF. We adjusted for baseline eGFR to account for inter-individual variability.

The slopes of biomarkers and eGFR were normally distributed. We used standardized slopes (using Z-transformed slopes) for all continuous measures in the models.

The frequently-used single measure of eGFR<60 was considered as an operational definition of CKD. For individuals who had eGFR<60 at one visit and eGFR≥60 at the next visit, we allowed them to re-enter the analyses in the next visit to calculate incidence rate for the subsequent round of analysis. Therefore, we treated CKD as a recurrent event. We then examined if the inflammatory biomarkers significantly predicted incident CKD within the subsequent follow-up visits using weighted generalized linear models (Poisson distribution model, lag-time) accounting for missing values, with results expressed as relative risk (RR) and 95%CI. All models accounted for covariates described above. We performed all statistical analysis in STATA 14.1(College Station, Texas 77845 USA).

Results

Discussion

Our study shows that in the group of nondiabetic community-dwelling older adults with initially normal RF, higher baseline sTNFα receptors are associated with accelerated RF decline, and a greater likelihood of incident CKD. This association remained strong over the 9-year follow-up and was independent of SBP and new-onset DM. We also showed that greater increases of these inflammatory biomarkers were associated with faster decline of RF, independent of baseline RF, SBP, new-onset DM, and CHF.

Our results are similar to other studies showing the predictive roles of TNFα receptors for incident CKD. In contrast, a study conducted in a Norwegian population showed CRP as the inflammatory biomarker most strongly associated with incident CKD.¹⁸ That study, however, was limited to a younger population (middle age) with shorter follow-up time. Additionally, the data were collected from only two follow-up visits, which might have resulted in regression toward mean error.¹⁸ These authors also suggested that the previously reported association of TNFα-R2 with incident CKD could be secondary to decline in renal excretion. Our findings, however, are not consistent with this interpretation, as we evaluated the association of the baseline TNFα receptors and the rate of RF decline, independent of baseline RF. Moreover, our results were confirmed in lag-time models where the incident CKD was estimated independent of baseline RF and based only on high inflammation states in the previous visits. These analytical methods should minimize the interference of renal excretion of inflammatory biomarkers.

Of note, most longitudinal studies that reported associations of TNFα-R1 or TNFα-R2 with the rate of RF decline were performed in participants who were significantly older than the Norwegian participants.^15–18 While CRP is a marker of general inflammatory status, sTNFα-R2 is a tissue-specific biomarker usually expressed in endothelial cells and leukocytes³² while its expression in other tissues is increased in disease states.³³

Some studies have demonstrated that TNF-α and/or its circulating receptors are associated with reduced RF, but these studies have focused mainly on diabetic patients. One such investigation demonstrated that sTNFα receptors strongly predict stage 3 CKD in individuals with type-1 DM without proteinuria.³⁴ There is also evidence that sTNFα receptors predict end-stage renal disease in individuals with type 2 DM.³⁵ Furthermore, previous studies have found an association between high levels of sTNFα-R1 and diabetic nephropathy.^{34, 36, 37} Our results suggest that, independent of other risk factors, micro-inflammation emerges at a very early stage of RF decline and should be considered as a potential independent risk factor in older adults.^{15–17, 38} Among the inflammatory biomarkers tested, we found that sTNFα-R1 and sTNFα-R2 were the strongest and most consistent predictors of CKD. Reports indicate that TNFα induces cleavage of the TNFα receptors from the membrane receptor.³⁹ TNFα-R1 activates pathways downstream to NF-κB, such as kinases of mitogen-activated protein kinase, or “death domain molecules,” resulting in IL-6 synthesis or apoptosis.^40–42 sTNFα-R1 is considered an indicator of long-term activation of TNFα.³⁴ It has been shown that sTNFα-R1also activates the apoptosis pathway through transforming growth factor beta-1.⁴³ Allugor et al. have reported that the R92Q mutation (rs4149584) increases the affinity of TNF to its receptor. In our study, all individulas were normal homozygotes for this variant.²³

Consistent with previous research, we found that increased levels of HS-CRP were associated with the prospective, accelerated decline of RF and CKD development, and increased CRP at the 6-year follow-up predicted incident CKD at the 9-year.¹⁸ These associations, however, were attenuated after adjustment for sTNFα-R1.

In our study, we found that sTNFα-R1 and sTNFα-R2 predict incident CKD. We used a categorical threshold of eGFR<60 ml/min/1.73m² as representing CKD, consistent with consensus definitions commonly utilized in large epidemiologic studies. We further adjusted for baseline eGFR to minimize bias from variation in initial RF.

Although IL-6 and TNFα have been frequently reported in association with age-related outcomes.^44–45 The mechanisms that link circulating inflammatory biomarkers and CKD risk, however, remain hypothetical. sTNFα receptors have been shown to play a pivotal role in the development of tubulointerstitial fibrosis and, therefore, nephropathy.⁴⁶ Moreover, inflammatory-induced atherosclerosis or oxidative stress detected by these biomarkers might mediate RF decline, with the latter linked to increased sTNFα-R2.⁴⁵ Additionally, age-related chronic inflammation might have an important role in the development of glomerulosclerosis. In a study of kidney donors, Kremers et al. reported that older age was associated with global glomerulosclerosis and higher odds of sclerosis in hypertensive donors.⁴⁷ Although these investigators did not measure inflammatory biomarkers, it is known that aging is associated with the development of inflammatory states. Mayr et al. demonstrated that angiotensin-II infusion in mice introduces a fibroinflammatory response that is followed by kidney tubulointerstitial collagen deposition mediated through the TNF-R1 signaling pathway and speculating that TNFα-R1 triggers collagen deposition.⁴⁸ Additionally, the significant association of lipocalin 2 in mesangial cells with sTNFα-R2 levels and impaired kidney function has been reported.³⁸ Therefore, a similar persistent age-related chronic inflammation might result in structural modification of glomerular and interstitial components of the kidneys, with resulting decline in RF.

Strengths and Limitations

Strengths of our study include the longitudinal design, with multiple repeated measures of RF and inflammatory biomarkers over a 9-year follow-up, and use of inverse probability weighting in the analysis to account for random missing values due to missing visits and truncation by death. Furthermore, due to the uncertainty of eGFR thresholds in aging, which identify true disease from age-related decline, we performed multiple analyses using both continuous measures of RF and an epidemiological-based threshold definition for CKD. These analyses demonstrate that the link between sTNFα receptors and RF is robust and independent of baseline RF. Lack of proteinuria measures in follow-up visits is a limitation. Consensus guidelines, however, define eGFR<60ml/min/1.73m² for more than 3 months as defining stage 3 or greater CKD, irrespective of proteinuria. In addition, albuminuria with preserved RF in older adults is primarily a manifestation of diabetic or hypertensive renal disease. Excluding the individuals with these morbidities in our study, the associations between TNFα receptors and the time-dependent slope and other RF outcomes remained significant.²⁶ One limitation of our study is that the follow-up intervals are 3 years apart. We operationalized time as a continuous variable to overcome this limitation for the annual slope of eGFR; incident CKD, though, is probably underestimated.

Conclusion

sTNFα receptors were the inflammatory biomarkers with the strongest association with the age-related decline of RF both at baseline and over time. Temporal changes in sTNFα receptors were associated with the rate of RF change. Moreover, these biomarkers predicted incident CKD independent of the main conventional risk factors and baseline RF. In contrast, in multivariate analyses, other inflammatory biomarkers, including IL-6, IL-18, and IL-1β, were not consistently significantly associated with RF outcomes. These results suggest that higher levels of sTNFα receptors, in the absence of other chronic conditions such as DM, might be an early sign of impending, accelerated RF decline. Furthermore, these biomarkers could be used to identify individuals at elevated risk of CKD who could be candidates for interventions aimed at preventing progression of RF decline.

Supplementary Material