Urinary albumin excretion and cancer risk: the PREVEND cohort study

ABSTRACT

Background

Chronic kidney disease (CKD) is believed to be associated with an increased risk for cancer, especially urinary tract cancer. However, previous studies predominantly focused on the association of decreased estimated glomerular filtration rate (eGFR) with cancer. In this study, we investigated the association of albuminuria with cancer incidence, adjusted for eGFR.

Methods

We included 8490 subjects in the Prevention of Renal and Vascular End-stage Disease (PREVEND) observational study. Urinary albumin excretion (UAE) was measured in two 24-hour urine specimens at baseline. Primary outcomes were the incidence of overall and urinary tract cancer. Secondary outcomes were the incidence of other site-specific cancers, and mortality due to overall, urinary tract, and other site-specific cancers.

Results

Median baseline UAE was 9.4 (IQR, 6.3–17.8) mg/24 h. During a median follow-up of 17.7 years, 1341 subjects developed cancer (of which 177 were urinary tract cancers). After multivariable adjustment including eGFR, every doubling of UAE was associated with a 6% (hazard ratios (HR), 1.06, 95% confidence intervals (CI), 1.02–1.10), and 14% (HR, 1.14, 95% CI, 1.04–1.24) higher risk of overall and urinary tract cancer incidence, respectively. Except for lung and hematological cancer, no associations were found between UAE and the incidence of other site-specific cancer. Doubling of UAE was also associated with a higher risk of mortality due to overall and lung cancer.

Conclusions

Higher albuminuria is associated with a higher incidence of overall, urinary tract, lung, and hematological cancer, and with a higher risk of mortality due to overall and lung cancers, independent of baseline eGFR.

Graphical Abstract

Graphical Abstract.

KEY LEARNING POINTS.

What was known:

• CKD is believed to be associated with a high risk of cancer, particularly of urinary tract cancer. Previous studies in this field focused on the association of eGFR with cancer risk. Studies investigating the association of albuminuria with cancer risk independent of eGFR are scarce and provide varying results.

This study adds:

• This population-based study demonstrates that when using the best estimate of albuminuria (two 24-hour UAEs) and kidney function (creatinine and cystatin C-based eGFR) higher albuminuria is associated with higher cancer risk, particularly for urinary tract cancer, independent of eGFR.

Potential impact:

• Albuminuria could be an additional marker to identify subjects eligible for screening for certain cancer types, particularly urinary tract cancer. Studies investigating the mechanism linking albuminuria to cancer risk are warranted.

INTRODUCTION

Chronic kidney disease (CKD) is defined as abnormalities of kidney structure or function, present for >3 months, with implications for health [1]. It is now generally acknowledged that CKD predisposes for shorter life expectancy, especially because of a link with cardiovascular disease [2]. Cancer is one of the other most common causes of mortality [3]. Accumulating evidence has shown that CKD is associated with a higher risk to develop cancer, especially urinary tract cancer. However, most of these studies focused on cancer risk in subjects with decreased eGFR, rather than on cancer risk in subjects with higher albuminuria [4–7].

The limited number of studies that investigated cancer risk in subjects with higher albuminuria showed inconsistent results, and often these studies did not investigate whether the reported association was independent of important covariates including eGFR. The variation in the strength of the association of albuminuria and cancer risk between the various studies may stem at least in part from the populations included, ranging from general population cohorts to high-risk cohorts, but also from different albuminuria measurement techniques that were applied. In some studies, albuminuria was assessed by semi-quantitative dipstick proteinuria, whereas in others quantitative measurement of urinary albumin concentration (UAC) or albumin–creatinine ratio (ACR) in spot urine samples was used [8–11]. These methodologies are less precise to estimate albuminuria compared to the measurement of urine albumin excretion in 24-hour urine collections (urinary albumin excretion, UAE), which is considered to be the gold standard [1].

Notably, urinary tract cancer is one of the cancer types that has consistently been shown to be associated with CKD defined by decreased eGFR or elevated albuminuria [4, 6, 9, 11, 12]. However, whether this association holds for all subtypes or only specific subtypes of urinary tract cancer is largely unknown. The data collected in the PREVEND study provides an opportunity to investigate the association of albuminuria, measured in the most optimal way, with the risk of different histological subtypes of urinary tract cancer as well as other site-specific cancers. Therefore, this study aims to examine the association of UAE, measured in 24-hour urine collections, with the incidence and mortality of overall, urinary tract, and other site-specific cancers adjusted for relevant covariates including kidney function in a population-based cohort with a long follow-up.

MATERIALS AND METHODS

Study design and population

For this study we analyzed data from the Prevention of Renal and Vascular End-stage Disease (PREVEND) study, a prospective, population-based cohort of Dutch men and women aged 28 to 75 years that by design was enriched for subjects with higher levels of albuminuria. In total, 8592 individuals were enrolled. Of these, 2592 subjects had at baseline a UAC of <10 mg/L, and 6000 subjects had a UAC of ≥10 mg/L. Subjects were screened at an outpatient clinic at baseline (1997–1998) and each 3 to 4 years thereafter. Details of this study have been reported before [13, 14]. For the present analyses, we excluded those being on or with a history of dialysis (n = 22) and missing values for kidney function at baseline (n = 80), leaving 8490 participants. The PREVEND study conformed to the principles drafted in the Helsinki declaration. Local medical ethics committee approval was obtained (approval number: MEC96/01/022), and informed consent was provided by all participants.

Data collection

The procedures at each examination in the PREVEND study have been described in detail previously [13]. In brief, each examination included two visits to an outpatient clinic separated by 3 weeks. Before the first visit, all participants completed a questionnaire regarding demographic characteristics, cardiovascular and renal disease history, smoking habits, alcohol consumption, and medication use. Information on medication use was combined with information from the IADB.nl database, containing pharmacy-dispensing data from community pharmacies in the Netherlands [15]. During the first visit, participants' heights and weights were assessed. During each visit, blood pressure was measured and the average value was used. In the week before the second visit, subjects collected two 24-hour specimens after thorough oral and written instructions to avoid heavy exercise as much as possible during the urine collection and to postpone the urine collection in case of urinary tract infection, menstruation, or fever. Furthermore, fasting blood samples were drawn at the second visit.

Assessment of covariates

Body mass index (BMI) was calculated as the ratio of weight to height squared (kg/m²). Smoking status was defined as self-reported never, former, or current smoking. Alcohol consumption was evaluated as alcoholic drinks per day with one drink being considered to be equivalent to 10 g of alcohol regardless of the type of beverage [16], and was categorized in (i) no/rarely; (ii) 0.1–10 g per day (occasional to light); (iii) 10–30 g per day (moderate); and (IV) >30 g per day (heavy) [17]. Educational level was defined as low (primary education or intermediate vocational education), middle (higher secondary education), and high (higher vocational education and university) [18]. Type 2 diabetes was defined as a fasting plasma glucose level of ≥7.0 mmol/L, non-fasting glucose ≥11.1 mmol/L, or the use of glucose-lowering drugs [19]. Hyperlipidemia was defined as a total cholesterol ≥5.0 mmol/L (≥193 mg/dL) in subjects with a history of cardiovascular disease or ≥6.0 mmol/L (≥232 mg/dL) in subjects without a history of cardiovascular disease, or use of lipid-lowering medication [20]. For each of the two urine examinations, the average value of the paired 24-hour collections was calculated. Urine osmolality was calculated using the following formula: 2 × (urine Na⁺ concentration + urine K⁺ concentration) + urine urea concentration [21]. Hematuria was verified by the dipstick reading for erythrocytes of 1+ or above [22].

We estimated GFR with the 2012 combined creatinine cystatin C-based Chronic Kidney Disease Epidemiology (CKD-EPI) Collaboration equation, taking into account age, sex, and race [23]. Data of eGFR were imputed with the creatinine-based or cystatin C-based equation if measurements for one of these analytes were missing. Measurement of serum creatinine was performed by an isotope dilution mass spectrometry traceable enzymatic method on a Roche Modular analyzer using reagents and calibrators from Roche (Roche Diagnostics, Mannheim, Germany), with intra-assay and inter-assay coefficients of variation of 0.9% and 2.9%, respectively. Serum cystatin C concentrations were measured by the Gentian Cystatin C Immunoassay (Gentian AS, Moss, Norway) on a Modular analyzer (Roche Diagnostics). Cystatin C was calibrated directly using the standard supplied by the manufacturer (traceable to the International Federation of Clinical Chemistry Working Group for Standardization of Serum Cystatin C). The intra-assay and inter-assay coefficients of variation were <3.3% and <4.1%, respectively [14].

Assessment of urinary albumin

UAC was measured by nephelometry with a threshold of 2.3 mg/L, and intra-assay and inter-assay coefficients of variation of 2.2% and 2.6%, respectively (Dade Behring Diagnostic, Marburg, Germany). UAC was multiplied by urine volume to obtain UAE expressed as mg per 24 hours. The two 24-h UAE values of each subject per examination were averaged. ACR was calculated by dividing UAC in milligrams by urine creatinine concentration in grams. The variation in the two 24-h urine creatinine excretions was only 2.9% suggesting that on average 24-h urine collection was correct. Data on these measurements were collected during follow-up every 3 to 5 years.

Ascertainment of cancer outcomes

Primary outcomes were the incidence of overall cancer and urinary tract cancer. Secondary outcomes were the incidence of other site-specific cancers, as well as the mortality due to overall, urinary tract, and other site-specific cancers. The site-specific cancers with an event rate exceeding 0.5% in our study population during the follow-up of the study are reported [4]. Non-melanoma skin cancer was excluded from the definition of overall cancer due to its high prevalence and benign prognosis when compared to other malignancies [24]. As mentioned before, we further investigated the histological subtypes of urinary tract cancer. Subjects with a cancer diagnosis before baseline were excluded from the analyses of the corresponding cancer type and for overall cancer incidence. For cancer incidence, data were retrieved via linkage to the Dutch Nationwide Pathology Databank (PALGA). PALGA data were obtained from the period 1971–2015 [25]. In addition to PALGA data, we used the data of a self-reported questionnaire to exclude subjects with cancer before baseline. In case of multiple cancer diagnoses during follow-up, the earliest cancer diagnosis after baseline was used to index overall cancer. Cancer mortality data were ascertained by record linkage to the Dutch Central Bureau of Statistics [26]. and grouped using ICD-10 codes as: overall (C00-C97, excluding C44 non-melanoma skin cancer), urinary tract (C64-68), prostate (C61), lung (C33-34), colorectal (C18-20), and hematological cancers (C81-96). Subjects were censored at the end of follow-up (December 31, 2015) or at the date of non-cancer death, whichever occurred first.

Statistical analyses

Baseline characteristics are shown according to tertiles of UAE. Continuous data are presented as mean with SD or as median with IQR in case of skewed distribution. Categorical data are presented as percentages. Trend analysis of baseline characteristics within UAE strata was performed by linear regression or Kruskal–Wallis test for continuous variables and linear-by-linear association χ² test for categorical variables.

UAE was analyzed as a continuous term per 1-unit increment of the log₂-transformed UAE (i.e. per doubling of albumin) and per tertile. To study the associations between UAE and cancer incidence and mortality, Cox proportional hazards regression analysis was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). The percentages of missing values of smoking status, alcohol consumption, diabetes, and BMI were 0.4%, 0.5%, 0.6%, and 1.1%, respectively. Other covariates were available in all cases. We used pairwise deletion to show the baseline characteristics. We imputed the missing values of the categorical covariates by adding a category of missing and we used listwise deletion to handle the missingness of the continuous covariates when analyzing Cox models. In model 1, we first calculated HRs adjusted for age, sex, BMI, smoking status, alcohol consumption, educational level, and type 2 diabetes. In model 2 we additionally adjusted for baseline eGFR. To visualize these associations, we fit restricted cubic splines including adjustments as in model 2. We explored possible effect modification by clinically important covariates on the association of UAE with the primary outcomes by studying their cross-product term with UAE besides their main effects in model 2.

Several sensitivity analyses were performed to examine the robustness of, and the possible mechanisms underlying, the association between UAE and the primary outcomes. First, to detect potential reverse causation, we did a 1-year landmark analysis by excluding subjects with a follow-up time of <1 year and remodeled the associations. Second, to test whether the association of albuminuria with cancer incidence will differ across the various ways albuminuria can be expressed, we substituted UAE for ACR, UAC, and the average UAE of all available measurements of each examination during follow-up as an exposure variable. Third, to investigate possible bias in estimating cancer risk introduced by non-cancer mortality as a competing event, we conducted competing risk analysis using Fine and Gray subdistribution hazard regression [27]. Fourth, to investigate a possible mechanism linking UAE to urinary tract cancer, we additionally adjusted for urine osmolality in the final model. Fifth, to investigate the potential detection bias of urinary tract cancer that is relevant to hematuria, we additionally adjusted for hematuria in the final model. Sixth, to examine whether the use of 2021 CKD-EPI equation to calculate eGFR will influence the estimates of the albuminuria cancer association, we used this equation to calculate eGFR and remodeled the associations [28]. Seventh, to investigate possible confounding effects of blood pressure and the use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers (ACEi/ARBs), we additionally adjusted for these factors in the final model.

All P-values are two-tailed. A P-value of <0.05 is considered statistically significant. All analyses were conducted using the statistical package IBM SPSS (v.22; SPSS, Chicago, IL, USA), StataSE (v.15.0, College Station, TX, USA), and R (v.4.1.2, Boston, MA, USA).

RESULTS

Baseline characteristics

The 8490 subjects included in this study were predominantly whites, 50% were female, and their mean age was 49.8 ± 12.7 years. Median baseline UAE was 9.4 (IQR, 6.3–17.8) mg/24 h, and their mean baseline eGFR was 94.6 ± 17.3 mL/min/1.73 m². At baseline, subjects with a higher UAE were more likely to be male, to be older, to have lower eGFR, and to have hypertension and type 2 diabetes. Subjects in the highest tertile of UAE also had higher CRP levels and worse lipid profiles than subjects in the lowest tertile of UAE (Table 1).

Table 1:

Baseline characteristics of the 8490 PREVEND subjects overall and according to tertiles of UAE.

		UAE tertiles (mg/24 h)
	Total	T1: <7.1	T2: 7.1–13.5	T3: >13.5	P for trend
Participants, N	8 490	2 830	2 830	2 830
Female sex, %	50.0	59.4	50.2	40.5	<0.001
Age, y	49.8 ± 12.7	47.2 ± 11.8	48.1 ± 12.3	54.1 ± 12.8	<0.001
BMI, kg/m2	26.1 ± 4.2	25.3 ± 3.8	25.8 ± 4.0	27.2 ± 4.5	<0.001
Whites, %	95.5	94.4	96.0	96.2	<0.001
Current smoking, %	34.1	32.0	33.6	36.6	<0.001
No alcohol consumption, %	25.4	25.3	23.5	27.4	0.69
High education, %	29.4	32.4	32.6	23.4	<0.001
SBP, mmHg	129 ± 20	121 ± 16	127 ± 18	138 ± 22	<0.001
DBP, mmHg	74 ± 10	71 ± 8	73 ± 9	78 ± 11	<0.001
Use of antihypertensives, %	15.8	10.3	12.1	24.9	<0.001
Use of ACEi/ARB, %	5.8	3.0	4.6	9.5	<0.001
Hypertension, %	34.0	19.3	28.8	53.9	<0.001
Cardiovascular disease history, %	1.1	1.2	0.9	1.3	0.77
Glucose, mmol/L	4.9 ± 1.2	4.7 ± 0.8	4.8 ± 0.9	5.2 ± 1.6	<0.001
Use of antidiabetic agents, %	1.9	0.6	1.1	3.9	<0.001
Type 2 diabetes, %	3.7	1.2	2.1	7.9	<0.001
Total cholesterol, mmol/L	5.7 ± 1.1	5.5 ± 1.1	5.6 ± 1.1	5.8 ± 1.1	<0.001
Use of lipid-lowering agents, %	6.5	4.5	5.0	9.8	<0.001
Hyperlipidemia, %	4.5	3.3	3.2	6.9	<0.001
Cancer history (any type), %	3.2	2.5	2.3	4.8	<0.001
hs-CRP, mg/L	1.3 (0.6–3.0)	1.0 (0.5–2.5)	1.1 (0.5–2.7)	1.8 (0.8–3.9)	<0.001
eGFR, mL/min/1.73 m2	95 ± 17	96 ± 15	97 ± 16	90 ± 19	<0.001
No erythrocyturia, %	91.5	96.2	92.7	85.6	<0.001
Urine osmolality, mOsm/kg	541 ± 176	501 ± 170	565 ± 182	555 ± 171	<0.001
Urine creatinine excretion, mg/24h	1390 ± 401	1258 ± 347	1441 ± 398	1470 ± 421	<0.001
UAE, mg/24h	9.4 (6.3–17.8)	5.6 (4.8–6.3)	9.4 (8.1–11.0)	26.6 (17.8–54.7)	<0.001

P for trend is determined by a χ² test (categorical variables), or by linear regression or a Kruskal–Wallis test (continuous variables).

Abbreviations: ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blockers; BMI, body mass index; CRP, C-reactive protein; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; PREVEND, the Prevention of Renal and Vascular End-stage Disease study; SBP, systolic blood pressure; TC, total cholesterol; UAE, urinary albumin excretion.

Urinary albumin excretion and the risk of primary outcomes

During a median follow-up of 17.7 years, 1341 subjects developed a de novo malignancy, corresponding with a 10-year absolute risk (AR) for cancer of 8.1% (95% CI, 7.5%–8.6%, Table 2). Every doubling of UAE was associated with a 7% (HR, 1.07, 95% CI, 1.03–1.10) higher overall cancer incidence, when adjusted for age, sex, BMI, smoking status, alcohol consumption, educational level, and type 2 diabetes. This association remained essentially unchanged after additional adjustment for eGFR (HR, 1.06, 95% CI, 1.02–1.10). A linear association of every doubling of UAE with the risk of overall cancer incidence was observed (Fig. 1, P_{non-linearity} = 0.91).

Table 2:

Association of UAE with cancer incidence (data are shown for site-specific cancers when their event rate exceeds 0.5% in our study population during the total follow-up).

		UAE tertiles (mg/24 h)
	UAE as continuous variable (log2)	T1: <7.1	T2: 7.1–13.5	T3: >13.5
Overall cancer
No. of events/subjects	1341/8218	366/2759	397/2764	578/2695
10-year AR (%, 95% CI)	8.1 (7.5–8.6)
Crude	1.21 (1.17–1.25)	1.00 (ref.)	1.10 (0.96–1.27)	1.83 (1.60–2.08)
model 1	1.07 (1.03–1.10)	1.00 (ref.)	1.02 (0.88–1.18)	1.19 (1.03–1.37)
model 2	1.06 (1.02–1.10)	1.00 (ref.)	1.03 (0.89–1.19)	1.19 (1.04–1.37)
Urinary tract cancer
No. of events/subjects	177/8465	27/2829	54/2824	96/2812
10-year AR (%, 95% CI)	1.1 (0.8–1.3)
Crude	1.37 (1.27–1.47)	1.00 (ref.)	2.04 (1.29–3.24)	4.00 (2.61–6.13)
model 1	1.15 (1.05–1.25)	1.00 (ref.)	1.65 (1.04–2.63)	1.78 (1.14–2.77)
model 2	1.14 (1.04–1.24)	1.00 (ref.)	1.68 (1.05–2.67)	1.77 (1.13–2.78)
UCC
No. of events/subjects	135/8470	21/2830	39/2826	75/2814
10-year AR (%, 95% CI)	0.8 (0.6–1.0)
Crude	1.37 (1.26–1.48)	1.00 (ref.)	1.89 (1.11–3.22)	4.00 (2.46–6.49)
model 1	1.13 (1.02–1.25)	1.00 (ref.)	1.51 (0.89–2.58)	1.66 (1.00–2.76)
model 2	1.13 (1.02–1.25)	1.00 (ref.)	1.53 (0.90–2.61)	1.66 (1.00–2.75)
RCC
No. of events/subjects	27/8485	3/2830	11/2827	13/2828
10-year AR (%, 95% CI)	0.2 (0.1–0.2)
Crude	1.36 (1.13–1.64)	1.00 (ref.)	3.73 (1.04–13.37)	1.58 (1.39–17.10)
model 1	1.19 (0.96–1.47)	1.00 (ref.)	3.29 (0.91–11.84)	2.81 (0.77–10.30)
model 2	1.20 (0.96–1.49)	1.00 (ref.)	3.24 (0.90–11.70)	2.78 (0.76–10.22)
Lung cancer
No. of events/subjects	231/8474	53/2826	55/2828	123/2820
10-year AR (%, 95% CI)	1.3 (1.0–1.5)
Crude	1.32 (1.23–1.41)	1.00 (ref.)	1.06 (0.72–1.54)	2.61 (1.89–3.60)
model 1	1.12 (1.04–1.21)	1.00 (ref.)	0.97 (0.66–1.44)	1.40 (0.99–1.97)
model 2	1.13 (1.04–1.22)	1.00 (ref.)	0.97 (0.66–1.43)	1.39 (0.99–1.97)
Melanoma
No. of events/subjects	72/8472	28/2823	21/2825	73/2824
10-year AR (%, 95% CI)	0.3 (0.2–0.4)
Crude	1.00 (0.84–1.19)	1.00 (ref.)	0.76 (0.43–1.35)	0.94 (0.54–1.63)
model 1	0.98 (0.82–1.18)	1.00 (ref.)	0.79 (0.44–1.40)	0.90 (0.50–1.61)
model 2	0.98 (0.82–1.18)	1.00 (ref.)	0.79 (0.45–1.40)	0.90 (0.50–1.62)
Breast cancer (in female)
No. of events/subjects	213/4203	85/1665	64/1409	64/1129
10-year AR (%, 95% CI)	2.3 (1.9–2.8)
Crude	1.03 (0.93–1.14)	1.00 (ref.)	0.90 (0.65–1.24)	1.19 (0.86–1.64)
model 1	0.99 (0.89–1.11)	1.00 (ref.)	0.89 (0.64–1.23)	1.09 (0.78–1.53)
model 2	0.99 (0.89–1.10)	1.00 (ref.)	0.90 (0.65–1.25)	1.10 (0.79–1.53)
Prostate cancer (in male)
No. of events/subjects	182/4225	44/1147	48/1404	90/1674
10-year AR (%, 95% CI)	2.3 (1.9–2.8)
Crude	1.18 (1.08–1.28)	1.00 (ref.)	0.90 (0.60–1.36)	1.59 (1.11–2.29)
model 1	0.99 (0.90–1.10)	1.00 (ref.)	0.79 (0.52–1.20)	0.84 (0.57–1.24)
model 2	0.98 (0.88–1.08)	1.00 (ref.)	0.80 (0.53–1.22)	0.83 (0.57–1.22)
Colorectal cancer
No. of events/subjects	194/8469	53/2827	65/2823	76/2819
10-year AR (%, 95% CI)	1.2 (1.0–1.4)
Crude	1.10 (1.01–1.21)	1.00 (ref.)	1.25 (0.87–1.80)	1.61 (1.13–2.28)
model 1	0.94 (0.84–1.04)	1.00 (ref.)	1.17 (0.81–1.69)	1.00 (0.69–1.45)
model 2	0.93 (0.84–1.04)	1.00 (ref.)	1.17 (0.81–1.69)	1.00 (0.69–1.45)
Hematological cancer
No. of events/subjects	113/8472	26/2827	36/2827	51/2818
10-year AR (%, 95% CI)	0.6 (0.4–0.7)
Crude	1.30 (1.18–1.43)	1.00 (ref.)	1.41 (0.85–2.34)	2.21 (1.38–3.55)
model 1	1.14 (1.02–1.28)	1.00 (ref.)	1.27 (0.76–2.10)	1.32 (0.80–2.17)
model 2	1.12 (1.00–1.25)	1.00 (ref.)	1.30 (0.78–2.16)	1.32 (0.80–2.17)

HRs and 95% CIs were derived from Cox proportional hazards regression models.

Model 1: adjusted for age, sex, BMI, smoking, alcohol, educational level and type 2 diabetes; model 2: as model 1+ adjusted for baseline eGFR.

Abbreviations: AR, absolute risk; BMI, body mass index; eGFR, estimated glomerular filtration rate; RCC, renal cell carcinoma; UAE, urinary albumin excretion; UCC, urothelial cell carcinoma.

Figure 1:

Associations of UAE (in mg/24 h) with cancer incidence. The Kaplan–Meier curves present the survival function of (a) overall cancer and (b) urinary tract cancer according to UAE tertiles. Survival is defined as the fraction of subjects that did not develop cancer (or cancer-related mortality) during the total follow-up, with censoring for non-cancer mortality. The splines show the associations of UAE with the risk of the incidence of (c) overall cancer and (d) urinary tract cancer. Data were fitted by Cox proportional hazards regression models based on restricted cubic splines with three knots and adjusted for age, sex, BMI, smoking, alcohol, educational level, type 2 diabetes, and baseline eGFR. The spline curves are truncated at the 1.0th and 99.0th percentiles of the distribution curve. The reference standard for UAE was 9.4 mg/24 h. P-values for the nonlinear association are P = 0.91 for (c) and P = 0.16 for (d). Abbreviations: aHR, adjusted hazard ratio; BMI, body mass index; eGFR, estimated glomerular filtration rate; UAE, urinary albumin excretion.

Regarding the risk of urinary tract cancer, 177 out of 8465 subjects documented a new diagnosis during follow-up. The 10-year AR of incident urinary tract cancer was 1.1% (95% CI, 0.8%–1.3%, Table 2). In the multivariable model, every doubling of UAE was associated with a 15% (HR, 1.15, 95% CI, 1.05–1.25) higher risk of urinary tract cancer. This association remained significant after further adjusting for eGFR (HR, 1.14, 95% CI, 1.04–1.24). When investigating histological subtypes of urinary tract cancer, similar trends were presented in the association of UAE with the incidence of urothelial cell carcinoma (UCC) and renal cell carcinoma (Table 2), although the latter association did not reach formal statistical significance, probably due to lower power as the number of renal cell carcinoma events was relatively limited.

In addition, we did not find evidence for effect modification by age, sex, smoking status, BMI, and eGFR in the associations of UAE with the incidence of overall and urinary tract cancers (Fig. 2, all Pinteraction >0.05).

Figure 2:

Subgroup analyses investigating effect modification of the association of UAE with cancer incidence by age, sex, smoking status, BMI, and eGFR. HRs and 95% CIs were derived from Cox proportional hazards regression models. HR were adjusted for age, sex, BMI, smoking, alcohol, educational level, type 2 diabetes, and baseline eGFR. Abbreviations: BMI, body mass index; CI, confidence interval; eGFR, estimated glomerular filtration rate.

Urinary albumin excretion and the risk of secondary outcome

Melanoma, breast, prostate, lung, colorectal, and hematological cancers were included in the analyses of site-specific cancer incidence because the event rates of these cancers exceeded 0.5% in our study population. UAE was independently associated with an increased risk to develop lung cancer (HR, 1.13, 95% CI, 1.04–1.22, Table 2) and hematological cancer (HR, 1.12, 95% CI, 1.00–1.25), whereas we did not observe significant associations of UAE with the incidence of melanoma, breast, prostate, and colorectal cancers.

The associations of UAE with mortality due to overall, urinary tract, prostate, lung, colorectal, and hematological cancers were also analyzed. During follow-up, 604 deaths due to cancer occurred and the corresponding 10-year AR was 3.1% (95% CI, 2.7%–3.5%, Table S1). A 9% (HR, 1.09, 95% CI, 1.03–1.14) higher risk of cancer mortality was observed per doubling of UAE after multivariable adjustment including eGFR. Except for lung cancer, the associations of UAE with mortality due to site-specific cancers did not reach statistical significance, perhaps due to the limited power we had for these events.

Sensitivity analyses

The results of the various sensitivity analyses are described in Supplementary Tables S2–S7 and in general corroborated the results of the main analyses. The results remained essentially similar when we excluded subjects with a follow-up time of <1 year, suggesting no indications of reverse causation bias. In addition, UAE remained associated with the primary outcomes after we additionally adjusted for urine osmolality and hematuria, suggesting no role for urine osmolality nor for detection bias related to hematuria. The results were materially unchanged when we used the 2021 CKD-EPI equation to calculate eGFR, indicating the limited influence of the 2021 equation on the estimated associations. The results were also robust after we additionally adjusted for systolic blood pressure and the use of ACEi/ARBs, suggesting no confounding effects of these factors. We only found a few exceptions. First, the association of albuminuria with incident urinary tract cancer was slightly stronger when we included average UAE during total follow-up as the exposure variable instead of baseline albuminuria (HR, 1.17, 95% CI, 1.08–1.27, Table S3). Note that during the follow-up of 7.9 years subjects collected on average seven 24-hour urine samples. Second, the association of UAE with urinary tract cancer was slightly attenuated in the competing risk analyses but remained statistically significant (HR, 1.09, 95%CI, 1.00–1.19, Table S4).

DISCUSSION

In this population-based cohort study, higher UAE was associated with a higher risk of incident overall, urinary tract, lung, and hematological cancer, as well as mortality due to overall and lung cancer. These associations were independent of important covariates, including baseline eGFR. The association of UAE with the incidence of overall and urinary tract cancer was robust, as shown by the various sensitivity analyses that were performed.

To date, only a few other studies have investigated the association between albuminuria and cancer incidence and these studies reported conflicting results [8–12, 29–32]. For example, Lees et al. found a weak positive association of albuminuria with risk to develop prostate cancer, while Mok et al. found a reverse association [9, 10]. These inconsistent results may be caused by differences between studies in populations sampling strategy (a screening-based study in which everybody is invited for screening versus a registry-based study in which screening was performed only on medical indication), follow-up time, albuminuria reference group, and the method employed to quantify albuminuria (as 24-hour UAE in our study versus less precise assessment as ACR or a dipstick in a first morning void or a spot urine sample in other studies). Another explanation could be the differences between studies in unexplored confounders. For example, Ahn et al. reported an association between albuminuria and colorectal cancer, which we did not find in our study [11]. This inconsistency may at least in part be explained by the unexplored dietary confounders in these two studies. Various dietary components (e.g. processed meat) are recognized risk factors to develop colorectal cancer [33] and the intake of these components are known to be different between European and Asian populations [34].

Importantly, our results support the notion that the associations of albuminuria with cancer risk are independent of eGFR. Our study documents an essentially unchanged effect size when the association of albuminuria with cancer risk is adjusted for eGFR. The mechanisms connecting CKD to cancer risk are yet unknown, but inflammation is one of the most frequently hypothesized mechanisms [4, 5, 35, 36]. However, preliminary analyses of our data set showed that adjustment for high-sensitivity C-reactive protein (hs-CRP), a cytokine reflecting the degree of systemic inflammation [37], only partially or did not influence the association of albuminuria with urinary tract and overall cancer incidence, respectively (data not shown). We cannot rule out the possibility that hs-CRP may not comprehensively capture the chronic or local inflammation; still, it may well be that other mechanisms link albuminuria to cancer risk. For instance, albuminuria per se can reflect endothelial dysfunction or abnormal activation of the renin–angiotensin system, which is associated with pro-cancerous states of immunosuppressive microenvironment and tumor vasculature formation [38, 39]. Specifically for urinary tract cancer, another hypothesis could be that urinary albumin and/or other middle- and large-sized molecules that are lost in urine irritate urothelial epithelia by direct contact and lead to oncogenesis [40]. Future, experimental studies are warranted to examine these hypotheses.

The results of our study imply that albuminuria could be a candidate marker to help increase the precision and cost-effectiveness of screening programs for certain cancer types. For example, subjects with microscopic hematuria undergo risk stratification to help decide in whom cystoscopy and imaging examinations are recommended [41, 42]. The strong association of albuminuria with incident urinary tract cancer observed in our study suggests that these CKD markers could be added to the list of covariates that is used for risk stratification.

Our study has several strengths and limitations. A strength is that we used two 24-hour UAE measurements to quantify albuminuria exposure, which is the widely accepted gold standard. Other measurements of albuminuria used in prior studies, such as dipstick and ACR in spot urine samples, are considered less accurate, particularly for the detection of lower levels of urinary albumin and variability induced by the circadian rhythm in UAE [1, 43]. Furthermore, the 24-hour urine collections are on average accurate in our study, given that the variation in the two 24-hour urine creatinine excretions was only 2.9% and that the results were robust after we reanalyzed the data excluding subjects with potential inadequate 24-hour urine collections (data not shown). This study also adds to the as-yet limited evidence to answer the question whether the association of albuminuria with cancer risk is independent of eGFR. Important in this respect is that eGFR was quantified in the best manner available, because we incorporated not only serum creatinine, but also cystatin C. Furthermore, our well-phenotyped cohort enabled us to adjust for various important lifestyle and socioeconomic covariates [44]. We further handled some of these covariates that are possibly non-linearly associated with cancer (e.g. age, BMI, and eGFR) as splines rather than continuous variables in the Cox models to address residual confounding, and the results were robust (data not shown). Last, data on cancer incidence were unbiased because these were obtained via linkage with the PALGA registry, which reached nationwide coverage in 1991 [25]. Regarding limitations, for some of the site-specific cancers, the relatively low prevalence might have resulted in limited power to detect significant associations. In addition, as this study is observational in design, it cannot disentangle causality but only provide evidence supporting associations. This study also cannot adjust for the duration of smoking because relevant data were not available. Last, the PREVEND cohort is by design enriched for subjects with higher albuminuria. This will affect the prevalence of increased albuminuria, but not the strength of the association of increased albuminuria with cancer incidence (Table S8).

In conclusion, subjects with higher albuminuria, determined as 24-hour UAE, are at increased risk to develop overall, urinary tract, lung, and hematological cancers, independent of baseline eGFR. Elevated albuminuria is also independently associated with a higher risk of mortality due to overall and lung cancer. Future studies are required to investigate the mechanisms underlying these associations.

FUNDING

The PREVEND study is supported by several grants from the Dutch Kidney Foundation (E.033) and the Dutch Heart Foundation (2001.005), the Dutch Government, the US National Institutes of Health, and the University Medical Center Groningen, the Netherlands. L.L. is supported by a scholarship from the China scholarship council (CSC number: 202008440376).

AUTHORS’ CONTRIBUTIONS

All authors conceived and designed the study. L.M.K., B.v.d.V., E.G.G., S.J.L.B., R.A.d.B., N.S., and R.T.G. contributed to data acquisition. L.L. and L.M.K. conducted data analysis. All authors contributed to the interpretation of the data. L.L. and L.M.K. drafted the manuscript. All authors revised the article. R.T.G. supervised the work. All authors approved the final version of the manuscript.

DATA AVAILABILITY STATEMENT

The data underlying this article will be shared on reasonable request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

None declared.