In this multicentre study of 3642 Finnish individuals with type 1 diabetes, we compared the revised classification of albuminuria, based on the three most recent urine samples at baseline, with the most advanced stage of albuminuria noted pre-baseline. Regression was defined as a reversion from a higher to a lower category of albuminuria. Consequently, 23% of individuals with pre-baseline micro and macroalbuminuria were reclassified to a less advanced stage of albuminuria at baseline, and were therefore considered to have regressed.
It has previously been shown that baseline levels of albuminuria show a positive correlation with future cardiovascular events in individuals with diabetes [7,8,9]. We not only validated these previous results but we extended them to show that regression of albuminuria in individuals with type 1 diabetes is associated with reduced risk for cardiovascular events to the same level as in those who did not progress in the first place. Interestingly, similar results have been shown in large clinical studies such as the Losartan Intervention for Endpoint Reduction in Hypertension (LIFE) Study , which studied hypertensive individuals with and without diabetes. The cardiovascular consequences of a 50% reduction in microalbuminuria has also been studied in individuals with type 2 diabetes , and an adjusted risk of 0.41 after regression was reported for these individuals. Yet, in a recent paper from DCCT/EDIC , remission of albuminuria was not found to reduce the cardiovascular risk in individuals with type 1 diabetes. However, there are profound differences between DCCT/EDIC and the FinnDiane Study that might explain the contradictory findings. The DCCT/EDIC represents an intervention study comparing intensive with conventional insulin therapy (DCCT) with a subsequent open-label follow-up from the close-out of the DCCT up to 30 years from baseline (EDIC). The FinnDiane Study is not an intervention but includes a substantial number of evenly distributed individuals with type 1 diabetes in Finland. Therefore, our novel observations regarding the beneficial associations between regression of albuminuria and the risk for cardiovascular events may be seen to apply to the type 1 diabetes community in general. However, the single time point ascertainment of the outcome in our study is a limitation compared with the regression assessments based on time-updated measures performed in the DCCT/EDIC. Indeed, these methodological differences could partly explain the contradictory observations made in these two studies.
Several studies are in agreement with our findings that albuminuria predicts mortality rate [12,13,14,15] and that the risk of premature death (both cardiovascular and non-cardiovascular death) increases concomitantly with declining kidney function . We further extended our findings to include the impact of regression on mortality rate, and as with cardiovascular events, regression was associated with a significantly lower risk of premature death. The major cause of death in our cohort was cardiovascular disease and, as expected, cardiovascular mortality rates were reduced by regression of albuminuria in the same manner as all-cause mortality rates.
In contrast to results from the 1980s [1,2,3], some recent studies have indicated that microalbuminuria more frequently reverts to the range of normal AER than remains stable or progresses to ESRD. Two studies, one in type 1 diabetes  and one in type 2 diabetes , that defined regression of microalbuminuria as a 50% reduction in AER obtained similar results (58% and 51% 6 year cumulative incidence for regression, respectively). In the EURODIAB Prospective Complication Study , 50.6% of individuals with type 1 diabetes and microalbuminuria reverted to the normal AER range over a 7 year follow-up. Also in that study, regression occurred more frequently than progression, as the observed progression rate was only 13.9%.
Along similar lines, our study also proposes that microalbuminuria does not always predict inevitable progression to macroalbuminuria and ESRD. Notably, regression was more common in the aforementioned studies than our findings suggest. There may be a few explanations for this discrepancy. First, we did not have the possibility to examine changes in AER or ACR as continuous variables and theoretically we could therefore have lost valuable information by looking at only categorical variables. Yet, both Perkins et al  and Araki et al  showed surprisingly similar results in their studies by using categorical traits (regression from microalbuminuria yes/no; 59% and 54%, respectively) compared with a reduction of 50% in continuous AER measurements. Information on the overlap between the two approaches was unfortunately not shown. Our findings represent a snapshot in time not a cumulative incidence, which also may explain the lower regression rate. We cannot exclude the possibility that undetected regressions may have occurred between the initial and the revised classifications. Specifically, the scenario of fluctuating AER levels in the early stages of diabetic nephropathy, marking a dynamic rather than persistent renal injury, may remain unobserved in our study setting. In addition, the individuals in the present study were older, had longer duration of diabetes and higher BP at baseline compared with the other studies on regression of albuminuria in type 1 diabetes, which may further contribute to the differing results.
Another interesting finding of our study is the 23.4% regression from macroalbuminuria, which is strikingly similar to the regression rate in the microalbuminuric group. However, keeping in mind that the reference limits represent artificial cut-off values, this might not be surprising after all and may be a validation of the results obtained for the individuals with a history of microalbuminuria. Although the presence of macroalbuminuria has traditionally been considered a marker of irreversible renal damage, the concept of regression from higher levels of albuminuria is not completely new [22,23,24,25].
In clinical studies, intensified interventions (hypertension and glucose control) have been shown to reduce progression and even induce regression in both type 1 and type 2 diabetes, and consequently, varying regression rates have been obtained [26,27,28,29,30]. Consistent with the findings in the studies by Perkins et al  and the EURODIAB study group , we did not find evidence that renin–angiotensin–aldosterone system (RAAS) blockade is associated with regression. Regressed vs non-regressed individuals with a history of macroalbuminuria did not differ regarding the use of RAAS inhibitors (88% vs 84%), which is no surprise as the majority of individuals at this stage are designated to antihypertensive medication with RAAS inhibitors as the first choice. However, the difference in RAAS inhibitor use between regressed and non-regressed individuals with a history of microalbuminuria could be secondary to a more favourable health profile among those who had returned to normal AER.
A key strength of this study is the large, well-characterised population. The individuals in the current study represent nearly 10% of all individuals with type 1 diabetes in Finland and although the FinnDiane Study is not by definition a population-based study, the geographical distribution of participating individuals largely follows that of the general population. However, as all clinical variables were measured at the first FinnDiane Study visit, which also was the time point of the third and final urine sample included in the revised classification of albuminuria, we could not reliably assess factors associated with regression in this study. It is possible that the AERs of a number of regressed individuals levelled to less advanced stages already before entering the FinnDiane Study, but as a result of the limitations previously discussed, the exact time point for this remain unknown. Another disadvantage of the multicentre study design was the incomplete coverage of centrally measured AERs, and as a result of which we could not analyse change in AER/ACR as a continuous variable, as mentioned before. However, selection bias is less likely in a nationwide multicentre study compared with single-centre studies, and as all samples for each participant were analysed at the same centre, this should have no impact at the individual level. An additional strength of the study was the high quality and completeness of register data in addition to follow-up visits, which resulted in 100% coverage of cardiovascular events and deaths.
In conclusion, we observed regression in 23% of initially micro/macroalbuminuric individuals at baseline. A novel finding was that regression from micro to normoalbuminuria resulted in a cardiovascular risk and mortality rate no different from individuals who remained normoalbuminuric, and a similar phenomenon was seen for the individuals who regressed from macroalbuminuria compared with individuals with no change in microalbuminuria. Our observations underline the importance of keeping the AER as close to normal levels as possible in order to improve outcomes in type 1 diabetes.