Albuminuria, Kidney Function, and Cancer Risk in the Community

Yejin Mok; Shoshana H. Ballew; Yingying Sang; Josef Coresh; Corinne E. Joshu; Elizabeth A. Platz; Kunihiro Matsushita

Disclosures

Am J Epidemiol. 2020;189(9):942-950. 

In This Article

Results

Baseline Characteristics

The mean age of the study participants at visit 4 was 63 (standard deviation, 6) years, with 56% being women and 22% being Black (Web Table 1, available at https://academic.oup.com/aje). Compared with those who had higher eGFR levels, those who had lower eGFR levels were more likely to be older, have worse cardiovascular and cancer risk profiles (e.g., higher prevalence of hypertension, history of cardiovascular disease, higher high-sensitivity C-reactive protein levels, and higher pack-years of smoking), and take cardioprotective medication such as a statin and aspirin, but they were less likely to drink alcohol. The proportion of Black and diabetic participants demonstrated a J-shaped relationship across eGFR categories. Overall, similar patterns were observed among participants who had higher ACR levels compared with those who had lower ACR levels. A few exceptions were a monotonically higher proportion of Black and diabetic participants with increasing ACR level.

Kidney Disease Measures and Overall Cancer Risk

Over a median follow-up of 14.7 years (a total of 113,597 person-years), 2,030 incident cancer cases and 692 cancer deaths occurred (age/sex/race-standardized, in the US population, incidence rate (per 100,000 person-years) of 1,872 and mortality rate of 575). Overall cancer incidence rates did not show a clear pattern for eGFR, with the highest rate in those having eGFR values of 45–59, followed by eGFR of ≥90, and eGFR of <45 mL/minute/1.73 m2. In contrast, there was a dose-response relationship between ACR categories and overall cancer incidence rates. Largely similar patterns were seen for overall nonprostate cancer incidence rates, overall cancer mortality, and overall nonprostate cancer mortality (Table 1).

When eGFR and ACR were modeled continuously with splines, we confirmed that low eGFR (33–51 compared with 95 mL/minute/1.73 m2) and high ACR (≥78 compared with 5 mg/g) were associated with overall cancer incidence in demographically adjusted model (Web Figure 1A and 1B). However, these associations were no longer significant after adjustment for other potential confounders (Figure 1A and 1B). After excluding prostate cancer, in the demographically adjusted model, low eGFR (30–56 compared with 95 mL/minute/1.73 m2) and high ACR (≥13 compared with 5mg/g) were associated with overall nonprostate cancer incidence (Web Figure 1C and 1D). Although the associations were attenuated after adjustment for other potential confounders (Figure 1C and1D), all levels of ACR ≥103 mg/g showed significant associations with overall nonprostate cancer compared with ACR 5 mg/g (Figure 1D).

Figure 1.

Adjusted hazard ratios and 95% confidence intervals for overall cancer and overall nonprostate cancer incidence, Atherosclerosis Risk in Communities (ARIC), United States, 1996–2012. Linear splines of estimated glomerular filtration rate (eGFR) with 3 knots (45, 60, and 90 mL/minute/1.73 m2) with the referent at 95 mL/minute/1.73 m2 and linear splines of albumin-to-creatinine ratio (ACR) with 3 knots (10, 30, and 300 mg/g) with the referent at 5 mg/g. A) Overall cancer incidence for eGFR. B) Overall nonprostate cancer incidence for eGFR. C) Overall cancer incidence for ACR. D) Overall nonprostate cancer incidence for ACR. Adjustment was based on model 2 (age, sex, race–ARIC center, body mass index, hypertension, diabetes status, total cholesterol, statin use, aspirin use, high-sensitivity C-reactive protein, history of cardiovascular disease, family history of cancer, alcohol consumption amount, smoking status (never vs. <25 and ≥25 pack-years of smoking in former vs. <25 and ≥25 pack-years of smoking in current), and eGFR or ACR, as appropriate).

In sensitivity analyses, results were not materially changed with the following exceptions (Web Figure 2). Once we restricted to those who had the opportunity for cancer screening or censored cancer cases diagnosed within the first 3 years, the association between ACR and overall nonprostate cancer incidence was no longer statistically significant (Web Figure 2K and 2L). Similarly, low eGFR and high ACR were not associated with overall and overall nonprostate cancer risk in competing risk analysis (Web Figure 3). We did not observe significant interactions between demographic subgroups (Web Table 2). Cystatin C–based eGFR showed linear associations with overall and overall nonprostate cancer incidence in demographically adjusted models (Web Figure 4). However, once we adjusted for other potential confounding factors, these associations were no longer statistically significant. In general, we observed similar or slightly weaker associations for cancer mortality (Web Figure 5).

For cross-categories of eGFR and ACR, higher ACR was associated with higher risk for overall nonprostate cancer incidence irrespective of eGFR levels (P for interaction = 0.47) (Table 2), but none of the joint categories was significantly associated with overall cancer risk. We observed similar or slightly weaker associations for cancer mortality (Web Table 3).

Kidney Disease Measures and Site-specific Cancer Risk

Low eGFR was not associated with site-specific cancer incidence (Figure 2A). ACR showed significantly positive associations with lung and urinary tract cancer incidence (hazard ratio = 1.25 (95% confidence interval: 1.05, 1.49) and hazard ratio = 1.47 (95% confidence interval: 1.07, 2.03) per 8-fold increment, respectively), but it was inversely associated with prostate cancer incidence (hazard ratio = 0.79, 95% confidence interval: 0.68, 0.93) (Figure 2B). These associations remained significant in some sensitivity analyses (Web Figure 6). However, restricting to those who had the opportunity for cancer screening, the association for ACR and lung cancer incidence was attenuated (Web Figure 7E). When stratifying by smoking status, the association of ACR with lung cancer incidence was mainly observed in former and current smokers (Web Figure 7). However, of these site-specific cancers, only the relationship between ACR and lung cancer mortality was significant or borderline significant in several sensitivity analyses (Web Figure 8–10).

Figure 2.

Adjusted hazard ratios of overall and site-specific cancer incidence, Atherosclerosis Risk in Communities (ARIC), United States, 1996–2012. A) 15-mL/minute/1.73 m2 decrement of estimated glomerular filtration rate (eGFR). B) 8-fold increment of albumin-to-creatinine ratio (ACR). Adjustment was based on model 2 (age, sex, race–ARIC center, body mass index, hypertension, diabetes status, total cholesterol, statin use, aspirin use, high-sensitivity C-reactive protein, history of cardiovascular disease, family history of cancer, alcohol consumption amount, smoking status (never vs. <25 and ≥25 pack-years of smoking in former vs. <25 and ≥25 pack-years of smoking in current), and eGFR or ACR, as appropriate).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....