Optimal Frequency of Urinary Albumin Screening in Type 1 Diabetes

Patients with type 1 diabetes (T1D) have an estimated cumulative lifetime risk of chronic kidney disease of 50%, and chronic kidney disease can progress to kidney failure requiring dialysis or kidney transplantation and is associated with markedly reduced life expectancy and quality of life (1–6). The presence of moderately elevated albuminuria, also referred to as “microalbuminuria” and defined according to albumin excretion rate (AER) ≥30 to 300 mg/24 h, predicts progression to later stages of kidney disease. These later stages include severely elevated albuminuria, also referred to as “macroalbuminuria” with AER >300 mg/24 h, and reduced kidney function with estimated glomerular filtration rate (eGFR) ≤60 mL/min/1.73 m² (3). Evidence-based clinical practice recommendations call for interventions for moderately or severely elevated albuminuria, including intensified efforts to optimize glycemia (HbA_1c ≤7%) and blood pressure (<130/80 mmHg), and a comprehensive approach to reducing risk of cardiovascular disease and kidney damage including initiation of renin-angiotensin system inhibition (3,4,7,8). For these reasons, screening for albuminuria is recommended at least annually, beginning 5 years after diagnosis of T1D (4,7).

However, as incidence of elevated albuminuria is influenced by specific risk factors such as glycemic exposure (3,9), annual screening for all people with T1D may not be the most efficient strategy. While the test itself may seem simple and straightforward for patients and clinicians, abnormal results require confirmation by repeat testing (at least two of three urine samples abnormal), and false positives may arise from common, nonpathologic factors such as major exercise, menstruation, or asymptomatic infections (4,7). Cost and inconvenience of repeated testing are important considerations, and the coronavirus disease 2019 pandemic has reinforced the need to further justify procedures and costs (10).

Motivated by these considerations, we used >30 years of detailed measurements of kidney outcomes and comprehensive phenotyping in the Diabetes Control and Complications Trial (DCCT) and its observational follow-up, the Epidemiology of Diabetes Interventions and Complications (EDIC) study (9,11–14), to determine optimal, evidence-based screening frequency for albuminuria testing. Specifically, we aimed to compare routine annual screening with individualized recommendations based on covariate risk profiles, with the goals of earlier detection among those at high risk and fewer tests among those at low risk of elevated albuminuria.

We performed analyses of the DCCT cohort (randomized clinical trial, clinical trial reg. no. NCT00360893, ClinicalTrials.gov) and EDIC (follow-up observational study, clinical trial reg. no. NCT00360815, ClinicalTrials.gov) (13,15). Briefly, started in 1983, 1,441 individuals with T1D were enrolled in DCCT and randomly assigned to receive either intensive therapy (n = 711) aimed at maintaining glycemic levels as close as safely possible to the nondiabetes range or conventional therapy (n = 730) aimed at preventing hypo- and hyperglycemic events (13). Recruited participants consisted of two subcohorts, termed the primary prevention and secondary intervention cohorts. Members of the primary prevention cohort had 1–5 years’ diabetes duration and no evidence of retinopathy (by fundus photography) or kidney disease (AER <40 mg/24 h according to the definition at the time). Members of the secondary intervention cohort had 1–15 years’ diabetes duration, evidence of early retinopathy, or AER <200 mg/24 h (13). In the current analyses, 96 of the 1,441 DCCT participants were excluded for moderately elevated albuminuria (AER ≥30 mg/24 h) at DCCT baseline and 2 for insufficient follow-up data (Supplementary Fig. 1).

The DCCT demonstrated large, consistent effects of intensive therapy in preventing the development and decreasing the progression of microvascular complications, including kidney disease (16). After the end of the DCCT (1993), the conventional therapy participants were taught intensive therapy and all participants were referred back to their health care providers for subsequent diabetes care. EDIC (1994–present) had 97% of the DCCT participants enrolled, with >92% of the surviving cohort at risk for elevated albuminuria still actively participating as of 31 May 2017, the cutoff date for the current analyses. Protocols were approved by the institutional review boards of all participating clinical centers, and all participants provided written informed consent.

The primary study outcome for kidney disease during DCCT and in this report is elevated albuminuria at two consecutive DCCT/EDIC study visits, including moderately (AER ≥30 mg/24 h) or severely (AER ≥300 mg/24 h) elevated albuminuria. AER was measured annually during DCCT and semiannually during EDIC. Four-hour timed urine samples were collected from DCCT baseline through EDIC year 18 (2012), and albumin was measured with fluoroimmunoassay (12,17). From EDIC year 19 (2013) onward, single-voided morning urine samples were collected, and AER was estimated from urine albumin and creatinine concentrations with an equation developed and validated in the study population (12,17). Kidney failure occurred in 29 individuals, 27 of whom had elevated albuminuria prior to onset. Two of these cases occurred without prior elevated albuminuria and were not considered events of elevated albuminuria in the current analysis. We note that the AER in milligrams per 24 h from timed samples quantitatively approximates the albumin-to-creatinine ratio (ACR) in milligrams of albumin per grams of creatinine and this relationship has been broadly accepted in clinical practice guidelines (8). For example, an AER value ≤10 mg/24 h is approximately equivalent to an ACR value of ≤10 mg/g, and an AER ≥30 mg/24 h is approximately equal to an ACR of ≥30 mg/g.

Recognized and putative risk factors were assessed based on standardized physical examinations and the collection of blood and urine samples, all analyzed in the DCCT/EDIC central laboratory at the University of Minnesota with standard methods (18). HbA_1c was measured with high-performance liquid chromatography quarterly during DCCT and annually during EDIC (18). Fasting lipids (triglycerides and total and HDL cholesterol) were measured annually during DCCT and every other year during EDIC. LDL cholesterol was calculated with the Friedewald equation (18).

The analyses included data in the combined DCCT/EDIC period (1983–2017) up to and including the time elevated albuminuria was detected, or up to the last evaluation for participants who maintained normoalbuminuria (AER <30 mg/24 h). Data for the eligible participants were divided into follow-up intervals beginning with the time of AER <30 mg/g until the next measurement. The exposure time was reset to zero at every visit with AER <30 mg/g. Median (with first and third quartiles) for quantitative variables and percentages for binary variables were used to describe the risk profile variables in each follow-up interval, and each participant contributed multiple follow-up intervals.

We evaluated the probability of reaching albuminuria by the time of the next evaluation as a function of the risk profile at the current normoalbuminuria visit. In accounting for interval-censored data, a piecewise-exponential model assuming constant 6-month rates of elevated albuminuria was used, as it provided a fit similar to the nonparametric Turnbull estimator (19) (Supplementary Fig. 2). The cumulative incidence of elevated albuminuria was calculated stratified by levels of covariates. These risk estimates were then used to evaluate the optimal screening frequency. Specifically, risk factor profiles with 3-year incidence rates of elevated albuminuria >5% were considered high risk, indicating the need for greater screening frequency. Rates <1% were considered to indicate low risk and the need for lesser frequency.

More frequent screening schedules result in shorter length of time between the actual onset of elevated albuminuria and the next renal evaluation when the condition is detected (referred to as “time with undetected albuminuria”), while yielding a greater expected number of renal evaluations. Conversely, less frequent screening schedules result in longer time with undetected albuminuria and lower numbers of evaluations. We sought to determine whether a risk factor–based screening strategy could reduce time with undetected albuminuria and the number of tests compared with routine annual testing regardless of risk factors by examining for numerically lower time with undetected albuminuria and lower number of tests. Additional details are provided in Supplementary Material.

Data collected for DCCT/EDIC through 30 June 2017 are available to the public through the NIDDK Central Repository (https://repository.niddk.nih.gov/studies/edic/). Data collected in the current cycle (July 2017–June 2022) will be available within 2 years after the end of the funding cycle.

Characteristics of the 1,343 participants included in analyses at DCCT baseline are shown in Table 1. In brief, median age was 28 years, duration of diabetes 4 years, AER 10 mg/24 h, and HbA_1c 8.5%. Urine testing for albuminuria occurred at DCCT baseline, yearly during the trial, and biennially during the long-term EDIC follow-up (Supplementary Fig. 3 shows this bimodal distribution of testing frequency). Over follow-up, there were 17,252 kidney evaluations with normoalbuminuria (AER <30 mg/24 h), representing a mean 12.8 follow-up intervals per participant. Characteristics at the beginning of each of these intervals are also summarized in Table 1. Overall, 357 participants (27%) developed the primary outcome (elevated albuminuria), representing a 1.8% risk of developing elevated albuminuria within a year and a 3.2% risk of developing elevated albuminuria within 3 years (Supplementary Fig. 2 and first row of Supplementary Table 1, with AER density function shown in Supplementary Fig. 4).

The estimated cumulative incidence of elevated albuminuria within 1 and 3 years from prior evaluation stratified by levels of selected risk factors is shown in Fig. 1. Weak associations were seen for most risk factors (Fig. 1A–C and Supplementary Table 1). However, relative to the other risk factors, strong associations were observed for HbA_1c and AER (Fig. 1D and E). The cumulative incidence of albuminuria was positively associated with HbA_1c levels, with a 0.9% probability of elevated albuminuria within a year for HbA_1c levels ≤7.0%, 1.2% probability for HbA_1c levels 7.1–9.0%, 3% probability for HbA_1c levels 9.1–10.0%, and 3.7% probability for HbA_1c levels >10.0% (Fig. 1D and Supplementary Table 1). For AER, the probability of developing albuminuria within a year was 0.5% for AER ≤10 mg/24 h, 2.0% for AER 10.1–19.9 mg/24 h, and 7.7% for AER 20–29.2 mg/24 h (Fig. 1E and Supplementary Table 1).

The effect of the combined categories of HbA_1c and AER on incident albuminuria, created from the 12 cells in the 4 × 3 table representing four categories of HbA_1c and the three categories of AER (Supplementary Table 2), are shown in Fig. 1F and Supplementary Table 1. These groupings were categorized as follows: low risk, defined according to current AER ≤10 mg/24 h and HbA_1c ≤8%; high risk, defined according to AER 21–30 mg/24 h or HbA_1c ≥9%; and average risk, defined as all other levels. The 3-year cumulative incidence of albuminuria was 0.6% for those in the low risk category, 2.4% for those in the average risk category, and 8.9% for those in the high risk category (Fig. 2).

These results were used to determine the frequency of screening for elevated albuminuria (Table 2). For example, a personalized screening strategy that specifies AER screening in 2 years for individuals at low risk, in 3 months for individuals at high risk, and in 1 year for individuals at average risk was associated with a 52.0% reduction in time with undetected albuminuria and 13.6% reduction in the number of urine tests as compared with annual screening. Selection of screening in 6 months rather than 3 months for the high risk group was associated with 34.9% reduction in time with undetected albuminuria and 20.4% reduction in the number of urine tests as compared with annual screening. Results associated with individualizing screening intervals based on AER levels alone, or HbA_1c levels alone, are shown in Table 2.

Risk factors other than HbA_1c and AER did not yield strong risk stratification (Supplementary Table 1 and Supplementary Fig. 5), with the exception of triglyceride values >150 mg/dL, which occurred only in 4.4% of all evaluations. Personalizing the screening schedule based on AER and HbA_1c levels may also cover individuals at higher risk based on other factors. For example, ∼71% of patients with evaluations with triglyceride levels >150 mg/dL and ∼75% of patients with evaluations at age <20 years also had HbA_1c >9% or AER >10 mg/24 h and would be tested more frequently under the proposed screening strategy. Similarly, ∼60% of all smokers had HbA_1c ≥9% or AER >10 mg/24 h and therefore would also be tested more frequently under the proposed screening strategies.

We developed an incidence model for elevated albuminuria derived from >17,000 urinary albumin excretion assessments over 30 years of follow-up in 1,343 DCCT/EDIC participants with normoalbuminuria (AER <30 mg/24 h) at baseline. The findings of these analyses provide strong justification for a personalized screening schedule as an alternative to the current practice of routine annual screening for elevated albuminuria in all people with T1D. While numerous covariates were associated with greater magnitude of risk and shorter time to onset of elevated albuminuria, HbA_1c and the quantitative level of AER within the normal range had by far the largest effects. A simple screening strategy based on the current level of HbA_1c and current level of AER was associated with a 34.9% reduction in the time with undetected elevated albuminuria and a 20.4% reduction in the number of urine tests (and therefore a 20.4% reduction in direct testing costs) as compared with annual screening.

In current clinical practice, albuminuria is assessed on an annual basis after 5 years of T1D duration (12,17,20). However, despite an improved understanding of risk factors for kidney disease, in prior research investigators did not attempt to investigate alternative screening intervals. Fixed annual screening has been supported only by consensus, not high-quality observational or interventional evidence (4,7). Historically, individualized screening may not have been regarded as a priority owing to the simplicity and relatively low cost of albumin-to-creatinine ratio measurement (21). This is in contrast to research on the evidence-based screening frequency for diabetic retinopathy, for which the timing and frequency of screening procedures have greater impact on patient burden and cost (22).

The global coronavirus disease 2019 pandemic and a broader movement to provide remote health care options have provided motivation for clinicians to seek more streamlined, evidence-based approaches for routine testing (10,23). The reduction in overall clinical diabetes care that has occurred during the pandemic has been associated with worse metabolic control and greater risk of complications in some studies (24,25). These findings emphasize the need for optimal surveillance screening of diabetes-related complications, particularly in individuals at high risk. Concerns about insufficient screening (26), along with the concern of overlooking the patients at highest risk in the setting of remote and virtual care adoption (10,24,27,28), strongly justify an approach that targets those at the greatest complications risk for frequent screening. To date, an individualized approach has been studied and implemented most extensively for retinopathy screening (22,29). An individualized approach for retinopathy screening has been endorsed by many organizations as a clinical practice guideline for future efficient practice (30,31). The findings of the current analysis provide similar justification for individualizing the frequency of albuminuria screening.

The greatest reduction in time with undetected elevated albuminuria was observed with the strategy of using level of AER as a covariate risk factor (shown in Table 2), as those individuals with levels closer to the diagnostic threshold of 30 mg/24 h for AER (30 mg/g for ACR) would be screened more frequently. While a strategy that jointly considers both HbA_1c and AER levels for the determination of screening frequency provides more precision, we accept that simplified strategies with use of HbA_1c alone (Table 2) would also be associated with reductions in time with undetected elevated albuminuria and in the number of tests.

Personalized schedules have the potential to reduce the undetected time and the number of tests at lower cost. Based on prevalence of 0.5% (32), it is estimated that 1.65 million people currently live with T1D in the U.S. Considering the prevalence of ACR <30 mg/g (normoalbuminuria) and kidney disease incidence, the proposed screening schedule could result in ∼5.38 million fewer urine tests over 20 years. While we have not performed a formal cost-effectiveness analysis, the direct laboratory handling and analysis cost of ACR testing is ∼$50 (in U.S. dollars), which would translate in direct cost savings that could exceed $269 million as compared with routine annual screening over 20 years ($215 and $403 million for ACR testing of $40 and $75, respectively [see Supplementary Material]). This represents a conservative, informal estimate, as it does not include consideration of indirect patient and health care costs, the consequences of false-positive and false-negative tests, or clinical outcomes.

While the DCCT/EDIC longitudinal data set represents an unparalleled resource owing to its detailed long-term phenotyping, the current analysis has limitations. First, the highest level of evidence to support an individualized screening frequency strategy would be a future controlled trial in which individuals are randomly assigned to an individualized versus fixed annual screening (33,34). Second, scheduling screenings at fixed annual intervals may be easier to implement in clinical settings than individualized scheduling, though a simplified approach (such as the three-level approach according to HbA_1c and AER levels summarized in Supplementary Table 2) and automated scheduling may reduce this burden and enhance implementation (35). Third, we acknowledge that greater screening frequency in those at high risk may be difficult to implement owing to issues with adherence in those with features of high risk such as elevated HbA_1c. Fourth, in DCCT/EDIC a timed AER was measured, rather than ACR, for much of the combined duration. ACR is more typically assessed in practice. However, AER and its diagnostic thresholds (in milligrams per 24 h) have commonly been equated to ACR and its diagnostic thresholds (in milligrams per gram) in clinical practice guidelines and clinical care (8,17,20). Identification of kidney disease in which eGFR is abnormally low, despite lack of the development of elevated albuminuria, requires continued screening of eGFR (4,7,36–38). Finally, our findings do not apply to screening during pregnancy or in type 2 diabetes.

In conclusion, in people with T1D and normoalbuminuria (AER <30 mg/24 h), a personalized screening schedule for the identification of elevated albuminuria onset based on current level of HbA_1c and current level of urinary albumin is associated with a substantial reduction in the number of tests required (and their associated costs) for screening at the population level and is also associated with more timely identification of kidney disease onset.

Funding. DCCT/EDIC has been supported by cooperative agreement grants (1982–1993, 2012–2022) and contracts (1982–2012) with the Division of Diabetes, Endocrinology, and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases (current grant nos. U01 DK094176 and U01 DK094157) and by the National Eye Institute, the National Institute of Neurological Disorders and Stroke, the General Clinical Research Centers program (1993–2007), and Clinical Translational Science Center program (2006–present), Bethesda, MD. Industry contributors have provided free or discounted supplies or equipment to support participants’ adherence to the study: Abbott Diabetes Care (Alameda, CA), Animas (Westchester, PA), Bayer Diabetes Care (North America Headquarters, Tarrytown, NY), Becton Dickinson (Franklin Lakes, NJ), Eli Lilly (Indianapolis, IN), Extend Nutrition (St. Louis, MO), Insulet Corporation (Bedford, MA), LifeScan (Milpitas, CA), Medtronic Diabetes (Minneapolis, MN), Nipro Diagnostics (Ft. Lauderdale, FL), Nova Diabetes Care (Billerica, MA), OMRON (Shelton, CT), Perrigo Diabetes Care (Allegan, MI), Roche Diabetes Care (Indianapolis, IN), and Sanofi (Bridgewater, NJ).

Industry contributors have had no role in DCCT/EDIC.

Duality of Interest. B.A.P. has received speaker honoraria from Abbott, Medtronic, Insulet, and Novo Nordisk; has served as an advisor to Boehringer Ingelheim, Insulet, Sanofi, and Abbott; and has received research support to his research institute from Novo Nordisk and the Bank of Montreal. No other potential conflicts of interest relevant to this article were reported.

Authors Contributions. B.A.P. and I.B. designed the study with input from all co-authors. I.B. conducted the statistical analyses. B.A.P. wrote the initial draft of the manuscript. I.B., I.H.d.B., M.M., B.Z., J.B., G.M.L., D.M.N., and J.M.L. contributed revisions to the manuscript. All authors approved the final content. I.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.