Long-term intra-individual variability of albuminuria in type 2 diabetes mellitus: implications for categorization of albumin excretion rate

Diabetic kidney disease (DKD) is the leading cause of end-stage renal disease (ESRD) in the Western world. Current interventions for DKD do not arrest, but only delay the progression to ESRD [1]. The early and accurate identification of DKD, followed by early interventions may therefore offer the best chance of slowing the progression of kidney disease. One important method for evaluating the risk of progressive DKD involves identifying abnormal albumin excretion rate (AER), however the limitations of relying solely on AER as a marker of DKD are being increasingly recognised [2, 3].

The variability of 24 h urinary AER has been a topic for discussion due to its unpredictable nature. Rather than an abrupt transition from normal to abnormal values, albumin excretion often increases slowly over several years [4, 5]. The average increase in AER ranges from 10% to 30% per year until overt nephropathy develops, with some subjects showing slower rates of increase in AER after the development of microalbuminuria [6]. Regression from microalbuminuria to normoalbuminuria may also occur due to tight glycaemic control or the use of Renin-Angiotensin system inhibitors (RASi) [7]. Furthermore, the phenomenon of spontaneous regression from microalbuminuria to normoalbuminuria is now well recognised [5, 8‐10]. Despite the above, the strong relationship of progression of albuminuria in type 2 diabetes mellitus (T2DM) to declining glomerular filtration rate (GFR), ESRD and cardiovascular (CV) disease emphasises the importance of accurately classifying AER patterns [11‐13].

Several studies have demonstrated a wide range of intra-individual variability of albuminuria in diabetes, with the majority having a coefficient of variation in the range of 28% to 47% [5, 14]. Factors which affect the wide variation in AER include the type of urine sample analysed (e.g. 24 h, timed overnight, first morning, random), the concentration of urinary albumin, the time period over which the samples were collected (days, weeks, months), the clinical characteristics of participants, as well as the pre-analytical handling and storage of the urine samples [9, 14, 15]. The inherent variability of AER in people with diabetes also needs to be considered. Most previous studies investigating AER variability have been small and with a short follow up period in subjects with type 1 diabetes mellitus (T1DM) [16‐18]. Only a few, small studies of the variability of albuminuria in T2DM have been previously reported [19, 20]. Furthermore, there is a deficiency of studies which have documented the variability of albumin excretion over a prolonged period, with the follow-up period in most studies being less than 1 year.

Due to the high variability in AER, most guidelines [14, 21] recommend the use of more than or equal to two out of three AER measurements within a 3- to 6-month period to categorise AER. Even with this recommendation, misclassification of AER categories can occur due to a potentially greater long-term variability of AER than that reported in shorter-term studies. The aim of this study was therefore to investigate the long-term intra-individual variability of albuminuria over several years. Furthermore, we sought to identify the relationship of the variability of AER with various clinical and biochemical parameters.

The major finding of this study is that the long-term intra-individual coefficient of variation of AER is high, implying that more than three AER measurements may be necessary to accurately categorise albuminuria. In this study, we categorized albuminuria into normo-, micro- or macroalbuminuria groups according to participants’ initial three AER measurements and then subsequently into groups according to four patterns of AER trajectories: persistent, intermittent, progressing and regressing. The coefficient of variation of AER was 37.7%, 66% and 94.8% for subjects with normoalbuminuria who had persistent, intermittent and progressing AER pattern. It is recognised that by definition, patients with persistent AER measurements within the normo-, micro- and macro-albuminuria groups will have a lower co-efficient of variation than patients with intermittent, progressing or regressing patterns of AER. The coefficient of variation of AER for micro- and macroalbuminuria patients, regardless of the subsequent AER pattern fell within the above boundaries. In addition, out of all the baseline clinical and biochemical characteristics analysed, only the use of RASi agents significantly influenced the variability of AER. However, this effect was only evident for baseline AER variability and the data were insufficient to estimate the influence of RASi on the long term temporal patterns of AER used in this study. Kropelin et al. [24] evaluated data from three randomized intervention trials (BENEDICT, DIRECT, ALTITUDE and the Irbestartan in Patients with Type 2 Diabetes and Microalbuminuria (IRMA-2)) study and demonstrated that increasing the number of urine collections per study visit and the number of visits does not change the average drug effect estimate. It was therefore suggested that using a single urine collection per study visit was sufficient to define transition of albuminuria as an end-point in clinical trials [24].

The median intra-individual coefficients of variation of AER reported here are higher than the intra-individual variability seen in previous studies in T2DM (31–43%) [4, 20, 25]. These previous studies were typically conducted under stricter conditions (i.e., as part of a clinical investigation and not an outpatient setting). Although, one study conducted under routine clinic conditions, involving 1391 participants with T1DM and T2DM, in which an average of 2 to 3 samples per participant per visit (range 2 to 18) were collected over 2 years, reporting a coefficient of variation of 58%–82% [26]. The differences in coefficients of variation between our study and previous reports are likely to be multi-factorial. Many of the studies mentioned were conducted over shorter time intervals, on smaller numbers of patients, with smaller numbers of urine samples collected per participant and in specific study settings. By contrast, the current study investigated the intra-individual variability of AERs for patients with T2DM undergoing routine clinical assessment and a mean follow up of 7.9 years [4, 25].

There have only been a few previous reports of AER variability for patients with T2DM according to baseline AER categories of normo-, micro- and macroalbuminuria. In one study of 87 participants where only 3 overnight samples were collected per participant, the overall variability of AER was 25.7%, compared with 36.1%, 24.8% and 22.3%, for normo-, micro- and macroalbuminuric participants, respectively [27]. In contrast, the present study showed that the median coefficient of variation was significantly lower in those with normoalbuminuria (53.3%) compared to those with microalbuminuria (76%) after adjustment for the use of multiple characteristics including RASi agents (p = 0.007) at baseline. This study also showed a median coefficient of variation of 67% for those with macroalbuminuria. The higher intra-individual coefficient of variation in the albuminuria groups in the current study could be attributed to the longer duration of this study and the larger number of samples collected. Other factors that could have contributed include the progression of disease, effect of treatment on the disease or a fall in the number of participants as the years progressed.

The Renal Insufficiency and Cardiovascular Events (RIACE) study has also reported on the variability of albuminuria in T2DM [20]. The investigators determined AER from a subset of participants - 833 subjects had AER and 3229 participants had ACR measured at different laboratories. These investigators determined that the concordance rate between a single urinary albumin excretion (UAE) and geometric mean of multiple measurements depended upon the degree of albuminuria- 94.6% for normoalbuminuria, 83.5% for microalbuminuria, and 91.1% for macroalbuminuria. It is difficult to directly compare our results with the RIACE study, as the RIACE study used both ACR and AER, whereas the current study used AER. In the RIACE study, only 20.5% of participants had an AER measurement and 79.5% had an ACR measurement. Another limitation of the RIACE study is that participant characteristics were not equally distributed among albuminuria classes. In the RIACE study, the distribution of participants according to AER categories was 71.9% with normoalbuminuria, 23.2% with microalbuminuria and 4.9% with macroalbuminuria. In our study, we had 58% with normoalbuminuria, 32% with microalbuminuria and 10% with macroalbuminuria. Furthermore, compared to the RIACE study, we used AER measurements in all our participants. It is possible that using ACR can alter the number of samples required to categorise albuminuria, however in the current study our main purpose was to demonstrate AER variability.

An important aim of the present study was to observe temporal AER patterns and to determine baseline predictors of those patterns. Interestingly, in the current study, we found that only 5 out of 157 participants with microalbuminuria at baseline were persistently microalbuminuric throughout the average of 7.9 ± 3.1 years follow up period. The current study highlights the highly variable nature of microalbuminuria in patients with T2D. Historically, it has been assumed that the development of microalbuminuria signalled the inevitable progression to macroalbuminuria [14]. However, there is an increasingly recognised view that the development of microalbuminuria can no longer be viewed as a committed and irreversible stage of DKD, with spontaneous remission being frequently reported [2, 11]. Approximately 60% of patients with T1DM have displayed spontaneous remission of microalbuminuria independent of the use of RASi agents over 5–10 years of follow up [14]. However, spontaneous remission of microalbuminuria to normoalbuminuria in a cohort study of T1D patients followed up for over 30 years was not associated with a reduction in CV or renal risk compared to sustained microalbuminuria despite adjustment for RASi inhibitors [13]. However, the lack of association between CV and renal risk in the remission of microalbuminuria to normoalbuminuria may have been missed [28]. Changes in albuminuria may have been too small to detect any clinical significance despite long-term follow-up. In the present study of participants with T2DM, 28% showed remission from micro- to normoalbuminuria independent of RASi agent use. Other studies of T1DM and T2DM have reported rates of spontaneous remission from microalbuminuria to normoalbuminuria that have ranged from 39%–64% [9, 16, 17, 23].

We found that the baseline characteristics of sex, age, BMI, disease duration, HbA1c, smoking status, total cholesterol levels, HDL-cholesterol, SBP, and even use of RASi agents had no significant effect on the variability of AER. One limitation of this study was the inability to compare changes over time in medications, HbA1c, eGFR, systolic blood pressure and cholesterol/HDL levels with the changes over time in AER. The definitive answer as to whether a relationship does or does not exist between clinical and biochemical parameters and variability in AER therefore requires a study to determine if there is a relationship between temporal changes in both clinical and biochemical parameters over time and variability in AER. Unfortunately, we were not able to account for factors such as blood pressure, glycaemic control, dietary salt consumption, physical activity and inflammation as information pertaining to the above variables was not collected in a longitudinal fashion for this study. We recognise this as a limitation of the current study. Despite this, several studies have shown that there is no association between the variability of AER and sex [18, 25], age [18], BMI [18], total cholesterol [18, 19, 29] and SBP [18].

In this study, we also examined the theoretical effect of increasing the number of AER measurements per participant on 95 confidence limits for intra-individual coefficient of variation. As seen in Fig. 2, seven to ten measurements of AER can mark the beginning of the plateau of the 95% confidence intervals for the normoalbuminuric and microalbuminuric groups, respectively. It can therefore be inferred from the current study that the commonly accepted definition of two out of the first three samples is insufficient to categorize albuminuria at baseline (3). The RIACE study also suggests that although two AER samples can provide a robust classification of albuminuria status (sensitivity of 90.6 and specificity of 94.6), disease progression and the efficacy of reno-protective treatment such as the use of RASi agents cannot be accurately monitored unless albuminuria is measured at frequent intervals over a prolonged period of time [20]. Similarly, 388 T2DM patients using RASi inhibitors, losartan or irbestartan in the RENAAL and IDNT trials, showed a 30% reduction in albuminuria 3 months after commencement of RASi inhibition and a further decrease of 44.8% in 174 patients after 12 months [30]. The variability of albuminuria within patients suggests that incorporating multiple measurements improves risk algorithms and assessment of treatment effects over time [30].

In this study, 120 out of 617 (19%) participants were excluded from the study due to the lack of sufficient AER samples collected (minimum of five samples were required to be included in the study). This significant portion of participants excluded is a potential selection bias as patients with less samples collected were potentially those who are poorly compliant with follow up requirements. A limitation to the current study and to studies investigating albuminuria, is that there is no reference laboratory method of measuring urine or serum albumin levels [31]. Albuminuria is calibrated to a serum albumin reference material and is diluted to the concentrations measured in urine [31]. There is also a lack of development of a standard procedure for dilution and diluent in measuring albuminuria [31]. Furthermore, in the current study, we focused on albumin excretion rate as it has traditionally been accepted as the reference method for determining albuminuria rather than the albumin to creatinine ratio and did not specifically study albumin to creatinine ratio in the current study. However, it is appreciated that the albumin to creatinine ratio is usually now the preferred method for assessing albuminuria.

It is appreciated that it is often impractical to obtain multiple AER measurements in the clinical setting before deciding on treatment. Furthermore, it is important to appreciate that the relationship between AER and renal/vascular outcomes is continuous [5]. Attempts to classify AER into categories are performed to provide a simple framework for researchers and clinicians to interpret the results of interventions that alter albuminuria and to stratify the risk that individual patients have for the development and progression of renal and vascular disease. Clinicians should be aware of the wide variability of urinary albumin excretion and find a balance between theoretical suggestions for the number of AER measurements in the context of other risk factors. For instance, increases in AER in participants with hypertension are an indication for intervention whereas short-term increases in AER in normotensive participants are not necessarily an indication for intervention. During the time that the current study was conducted, many centres were using AER routinely as the way of determining albuminuria. Over the years, ACR has increasingly been used to determine albuminuria. The current study addressed the variability in AER therefore we are not able to comment as to whether the study will change the way we manage our patients as very few centres are now using AER to measure albuminuria.

The current study highlights an important finding, that there is a high degree of variability of AER in people with diabetes and that this high long-term variability of AER suggests that two out of three AER measurements may not always be adequate for the optimal categorisation and prediction of AER.