Acute kidney injury predicts the risk of adverse cardio renal events and all cause death in southeast Asian people with type 2 diabetes

Abstract

Patients with diabetes are susceptible to acute kidney injury (AKI) as compared to counterparts without diabetes. However, data on the long-term clinical outcome of AKI specifically in people with diabetes are still scarce. We sought to study risk factors for and adverse cardio-renal outcomes of AKI in multi-ethnic Southeast Asian people with type 2 diabetes. 1684 participants with type 2 diabetes from a regional hospital were followed an average of 4.2 (SD 2.0) years. Risks for end stage kidney disease (ESKD), major adverse cardiovascular events (MACE) and all-cause death after AKI were assessed by survival analyses. 219 participants experienced at least one AKI episode. Age, cardiovascular disease history, minor ethnicity, diuretics usage, HbA1c, baseline eGFR and albuminuria independently predicted risk for AKI with good discrimination. Compared to those without AKI, participants with any AKI episode had a significantly high risk for ESKD, MACE and all-cause death after adjustment for multiple risk factors including baseline eGFR and albuminuria. Even AKI defined by a mild serum creatinine elevation (0.3 mg/dL) was independently associated with a significantly high risk for premature death. Therefore, individuals with diabetes and any episode of AKI deserve intensive surveillance for cardio-renal dysfunction.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-77981-8.

Introduction

Acute kidney injury (AKI) is a clinical syndrome characterized by an abrupt kidney function decline. It may occur in 10–15% hospitalized patients and is associated with prolonged hospital stay and repeated unplanned hospital readmission^1,2. Moreover, AKI is independently associated with long-term adverse clinical outcomes such as incident chronic kidney disease (CKD), cardiovascular disease (CVD) and premature death^3–7. A meta-analysis with more than two million participants found that individuals with a history of AKI had 4.81 (95% CI 1.99–3.58) folds increased risk for progression to end stage kidney disease (ESKD) and 1.80 (98% CI 1.61–2.02) folds increased risk for death⁸. In the real world setting, AKI is often unrecognized and underdiagnosed due to its silent nature⁹. Due to the lack of specific treatments, management of AKI mainly relies on an early detection of abrupt decline of kidney function and timely management of the underlying causes¹⁰. Hence, intensive surveillance and early treatment of modifiable risk factors may mitigate the risk of AKI in high-risk subpopulations.

The incidence rate and the subsequent clinical outcomes after AKI are modified by baseline kidney function and comorbidities including diabetes and CVD⁸. Patients with diabetes are more susceptible to AKI as compared to their counterparts without diabetes^4,11,12. A recent cohort study reported that the risk for incident AKI was 4.7- fold higher in patients with diabetes than those without diabetes¹². Moreover, as compared to individuals without diabetes, patients with diabetes had higher risks for CVD, progression to advanced kidney disease and mortality after development of AKI¹³. Due to these reasons, AKI has been proposed as a bona fide complication of diabetes¹⁴.

Of note, most studies on AKI are conducted in general population or specific subgroups such as those with CVD^8,15. To our knowledge, few prospective studies have examined the long-term adverse outcome of AKI specifically in patients with type 2 diabetes. Thakar et al. followed 3679 patients with diabetes in United States from 1999 to 2008 and found that each AKI episode doubled the risk for progression to stage 4 CKD independent of other major risk factors of progression¹³. In a European cohort with 1371 type 2 diabetes patients followed for 12 years, Monseu et al. observed that incident AKI predicted all-cause mortality and CVD outcomes independent of a low eGFR and albuminuria¹⁶. Recently, Jiang et al. reported that AKI was associated with a high risk for all-cause mortality and an increased risk for progression to ESKD in Chinese type 2 diabetes¹⁷. However, data on major adverse cardiovascular events (MACE) were not reported in this study. Given that the risks for cardiovascular and renal complications differ greatly between Asian and Caucasian descents with diabetes^18,19, further studies in Asian population are needed to address this knowledge gap.

In this prospective cohort study, we sought to examine the incidence, clinical risk predictor and the subsequent cardio-renal outcomes of AKI in patients with type 2 diabetes in Singapore, a city state in Southeast Asia with 3 major Asian ethnic groups.

Methods

Participants

A total of 1684 participants with type 2 diabetes, age between 21 and 90 years old, and estimated glomerular filtration rate (eGFR) above 15 ml/min/1.73m²were consecutively recruited in a regional hospital in northern Singapore between January 2011 and October 2017²⁰. Type 2 diabetes was diagnosed by attending physicians based on the prevailing American Diabetes Association criteria after excluding type 1 diabetes and diabetes attributable to specific causes. Exclusion criteria were pregnancy, autoimmune diseases or cancer on active treatments, clinically manifest infection, and kidney diseases attributable to specific causes. Participants were followed by reviewing the centralized electronic database which has integrated data on imaging and biochemical examinations, medication dispensary, surgical operation procedure, outpatient clinical visits and hospitalization discharge summary. Follow-up for the current study was censored on March 31st, 2020.

Clinical and biochemical variables

Sex, ethnicity, and smoking status were self- reported. CVD history (nonfatal acute myocardial infarction and stroke) was ascertained by reviewing medical reports after cohort enrolment. Data on medication usage at baseline were retrieved from medication dispensing database. Blood pressure was measured twice by a semi- automatic sphygmomanometer in sitting position with 5- minute rest in-between and the average value was used. Mean artery pressure was calculated as (systolic pressure + 2X diastolic pressure)/3. HbA1c was measured using immunoturbidimetric method (Cobas Integra 800 Chemistry Analyzer, Roche, Basel, Switzerland). Creatinine was quantified by an enzymatic method which was traceable to isotope dilution mass spectrometry reference. All eGFR values in the current study were calculated by 2009 chronic kidney disease- Epidemiology Collaboration (CKD- EPI) formula based on serum creatinine²¹. Spot urine albumin was measured by an immunoturbidimetric assay (Roche Cobas c, Roche Diagnostics, Mannheim Germany). The level of albuminuria was presented as urinary albumin-to-creatinine ratio (ACR). Frequency of hospitalization was categorized into 4 levels: no hospitalization, 1 to 4, 5 to 9 and, ≥ 10 hospitalization admissions during follow-up.

Definition of AKI

Only inpatient serum creatinine measurements were used to define AKI because, (1) it was challenging to define AKI when patients were managed at outpatient clinics and blood tests were infrequent and, (2) it was particularly challenging to distinguish potential outpatient AKI episodes from nonlinear eGFR trajectories or rapid progression of CKD^22,23. In accordance with early studies, we defined AKI as 0.3 mg/dl (26.5 µmol/L) or greater than 50% relative difference between the peak and nadir serum creatinine concentrations during a single hospitalization, to partly account for the irregular intervals in-between creatinine measurements^3,16. AKI was further classified by stages according to the following ratios: stage I 150–199% or 0.3 mg/dl increment, stage II 200–299%, stage III 300% and above or on dialysis. For each patient, we considered only the first episode of AKI or the episode with the highest stage if multiple episodes of AKI were identified during a single hospitalization. Given that identification of AKI by 0.3 mg/dl creatinine change might lead to possible over-diagnosis²⁴, we included AKI defined by this criterion in the primary analysis but excluded it in a sensitivity analysis. We did not use the urine output criteria to define AKI in the current study.

Definition and identification of incident ESKD, MACE and all-cause death

ESKD was defined as progression to eGFR < 15 mL/min/1.7m²with confirmation by at least one additional eGFR measurement 3 months apart, initiation of maintenance dialysis for more than 3 months or death attributable to renal causes. MACE was a composite outcome which included nonfatal stroke, nonfatal myocardial infarction, and death attributable to cardiovascular cause, whichever occurred first. All-cause death was identified by data linkage with national death registry. Death attributable to CVD or renal causes was identified by the primary cause of death on death-certificates issued by qualified clinical practitioners²⁰.

Statistical analyses

Data were presented as mean ± standard deviation (SD), median (interquartile range, IQR) or percentage. Differences in baseline clinical and biochemical variables between participants with and without AKI were compared by Student t, Mann-Whitney U or χ2 tests where appropriate (Table 1). Urine ACR was natural log-transformed due to skewed distribution. Incident rate of AKI and the subsequent adverse outcomes were presented as number of events per 100 patient-years. Missing values were < 1% in clinical variables and were handled by listwise deletion.

Table 1.

Participant baseline clinical and biochemical characteristics stratified by AKI development.

	All participants N = 1684	No AKI N = 1465	AKI N = 219	P value
Age (years)	56.5 ± 12.4	55.7 ± 12.4	61.6 ± 11.0	< 0.001
Ethnicity (%)				< 0.001
Chinese	48.5	50.1	37.4
Malay	31.4	29.6	44.3
Asian Indian	20.1	20.3	18.3
Male sex (%)	57.9	58.0	57.1	0.79
Current smoker (%)	14.9	15.1	13.8	0.60
CVD history (%)	17.0	15.4	27.9	< 0.001
Body mass index (kg/m2)	27.4 ± 5.6	27.5 ± 5.6	26.7 ± 5.4	0.06
Diabetes duration (years)	11.6 ± 8.8	11.4 ± 8.6	14.4 ± 9.5	< 0.001
HbA1c (%)	8.7 ± 2.2	8.6 ± 2.1	9.3 ± 2.7	< 0.001
HbA1c (mmol/mol)	72 ± 18	70 ± 17	78 ± 23	--
Blood pressure (mmHg)
Systolic pressure	137 ± 20	137 ± 19	140 ± 22	0.01
Diastolic pressure	76 ± 12	76 ± 11	75 ± 15	0.18
Mean arterial pressure	97 ± 12	97 ± 12	97 ± 15	0.66
eGFR (ml/min/1.73m 2 )	81 ± 32	84 ± 29	58 ± 29	< 0.001
KDIGO eGFR category (%)				< 0.001
CKD Stage 1	48.6	52.6	21.5
CKD Stage 2	23.7	24.5	18.3
CKD stage 3	20.3	17.4	39.7
CKD stage 4	7.4	5.5	20.5
Urine ACR (µg/mg, IQR)	44 (12.0-266)	36 (11–181)	300 (67- 1479)	< 0.001
Medications usage (%)
RAS Blocker	55.1	54.3	60.7	0.07
Insulin	43.4	41.4	57.1	< 0.001
Diuretics	20.0	16.7	42.0	< 0.001

Between group differences were compared by Student t tests (age, body mass index, diabetes duration, HbA1c, blood pressure and eGFR), Mann- Whitney U tests (urine ACR) or χ2 tests (sex, ethnicity, current smoker, CVD history, KDIGO category and medication usage). CVD, cardiovascular disease, ACR, albumin-to-creatinine ratio; RAS, renin-angiotensin system. Variables which differed significantly between groups have been highlighted in bold font.

We employed Cox proportional hazard (PH) regression model to identify potential clinical risk predictors for AKI. Index age, sex, ethnicity (Chinese as reference), smoking status (active smoker versus others), CVD history (yes or no), body mass index (BMI), diabetes duration, HbA1c, mean artery pressure, baseline eGFR, urine ACR and medication usage (insulin, renin- angiotensin system antagonist [RAS], and diuretics, yes versus no) were included as covariates. Variables which independently predicted risk for AKI were selected by Cox regression with backward stepwise elimination. P-value > 0.10 was the criterion for removal of variables whilst p-value < 0.05 was the criterion for re-entry of variables. The p-values were iteratively derived from likelihood ratio statistic. We used Harrell’s c statistics (R package ‘survC1’) to assess the discrimination of the selected clinical variables for prediction of 5-year risk of AKI in the parsimonious model.

Cumulative risks for ESKD, MACE and all-cause death after first AKI episode were plotted by the Kaplan- Meier approach (1- survival) and differences between participants with and without AKI were compared by log- rank tests. The association of incident AKI with the subsequent risks for adverse clinical outcomes were further analysed by multivariable Cox PH regression models. Specifically, AKI was modelled as a time-varying exposure, i.e., time from first AKI episode onwards contributed to risk at time for adverse events whilst time from cohort enrolment to first AKI episode contributed to non- event time. Model 1 adjusted age, sex, ethnicity, CVD history, smoking status, BMI, diabetes duration, HbA1c, baseline eGFR and urine ACR. Model 2 further adjusted frequency of hospitalization above model 1. PH assumption was assessed by Schoenfeld residual. No violation of PH assumption was identified in the current study.

Data analyses were performed by SPSS (version 27) and R software (version 3.4.2). Two-sided P value < 0.05 was considered as statistically significant.

Results

Participant characteristics

A total of 1684 participants were included in the current analysis (average age 56.5 [SD 12.4] years old, diabetes duration 11.6 [SD 8.8] years, baseline eGFR 81 [SD 32] ml/min/1.73m², Chinese 49%, Malay 31% and Asian Indian 20%). During a mean follow-up of 4.2 (SD 2.0) years (median 3.9 [IQR 3.0 to 5.4] years, 7012 patient-years in total), 219 participants were identified to have at least one AKI episode (incidence rate 3.1 [95% CI 2.7–3.5] per 100 patient- years). Among those with AKI occurrence during the follow-up, 30 participants (13.7%) had 2 episodes and 14 (6.4%) had more than 2 episodes.

As shown in Table 1, participants with at least one AKI episode during the follow-up were older, had a longer diabetes duration, were more likely to have CVD history, be Malay ethnicity and on insulin and diuretics treatments as compared to those without AKI events. Also, they had a higher HbA1c and a higher level of albuminuria at baseline. Noteworthy, participants with a lower eGFR at baseline had a high risk for development of AKI. The crude incident rate of AKI in participants with baseline eGFR in CKD stage 1 (≥ 90 ml/min/1.73m²), stage 2 (60–89 ml/min/1.73m²), and stage 3/ 4 (15–59 ml/min/1.73m²) was 1.3, 2.4 and 8.1 per 100 patient-years, respectively.

Potential clinical risk predictors for incident AKI

Index age, BMI, diabetes duration, HbA1c, baseline eGFR, urine ACR, Malay ethnicity, CVD history and usage of diuretics were associated with incident AKI in the univariate analysis (Additional file 1, supplementary Table 1). Backward stepwise Cox regression suggested that index age, minor ethnicity, CVD history, HbA1c, eGFR, urine ACR and usage of diuretics were independent predictors for risk of AKI after accounting for other clinical and biochemical variables (Table 2). The model based on these 7 clinical variables predicted 5-year risk for incident AKI with an area under receiver operating characteristic curve (AUC) of 0.80 (95% CI 0.76–0.84).

Table 2.

Association of potential clinical predictors with risk for incident AKI in the parsimonious Cox regression model.

	HR (95% CI)	p value
Age (years)	1.03 (1.01–1.05)	< 0.001
Malay ethnicity a	1.65 (1.20–2.27)	0.01
Asian Indian ethnicity a	1.50 (1.01–2.24)	0.05
CVD history (yes versus no)	1.58 (1.15–2.18)	0.01
HbA1c (%)	1.15 (1.08–1.22)	< 0.001
eGFR (ml/min/1.73m2)	0.99 (0.98–0.99)	< 0.001
Urine ACR (natural log-transformed)	1.28 (1.18–1.39)	< 0.001
Usage of diuretics (yes versus no)	1.55 (1.13–2.13)	0.01

Cox proportional hazard regression: outcome was time to incident AKI. Six clinical variables were selected by backward Cox regression to build a parsimonious clinical model. ^a Chinese ethnicity was taken as reference.

Adverse clinical outcomes after AKI events

151 EKSD events were identified during the follow-up and 41 occurred after an AKI episode. The median time from cohort enrolment to ESKD was 2.1 (IQR 1.1–3.6) years in participants without AKI whilst the median time from AKI to ESKD was 0.9 (IQR 0.3–1.8) years. Cox regression showed that the unadjusted risk for ESKD after AKI was 6.12 (95% CI 4.09–9.17) folds higher as compared to those without development of AKI. The association of AKI with ESKD was attenuated but remained statistically significant after adjustment for demographic and cardio-renal risk factors including eGFR, ACR and frequency of hospitalization (adjusted HR 1.60, 95% CI 1.02–2.49, Table 3; Fig. 1).

Table 3.

Adverse clinical outcomes after AKI occurrence in Cox regression models.

	Unadjusted Model		Multivariable Model 1		Multivariable Model 2 a
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
ESKD	6.12 (4.09–9.17)	< 0.001	1.57 (1.02–2.42)	0.04	1.60 (1.02–2.49)	0.04
MACE	4.28 (2.84–6.45)	< 0.001	2.21 (1.42–3.43)	< 0.001	1.81 (1.16–2.82)	0.01
Death	7.40 (5.43–10.08)	< 0.001	3.51 (2.51–4.89)	< 0.001	3.01 (2.14–4.23)	< 0.001

Cox PH regression models: outcome was time to ESKD, MACE and all-cause death, respectively. Model 1 adjusted age, sex, ethnicity, CVD history (yes or no), smoking status (current smoker versus others), body mass index, diabetes duration, HbA1c, mean artery pressure, eGFR and urine ACR. Model 2 additionally adjusted frequency of hospitalization above model (1) ^a hospitalization due to myocardial infarction and stroke were not included in calculation of hospitalization frequency when MACE was taken as outcome in model (2) ESKD, end stage kidney disease; MACE, major cardiovascular event.

Figure 1.

Risks for end stage kidney disease (ESKD), major adverse cardiovascular disease (MACE) and all-cause death after first AKI episode as compared to those with no AKI events in the follow-up.

A total of 223 MACE events were identified during follow-up and 43 occurred after AKI. The median time from cohort enrolment to MACE was 2.0 (IQR 0.8–3.3) years in participants without AKI whilst the median time from AKI to MACE was 0.5 (IQR 0.1-1.0) years. Cox regression model suggested that the unadjusted risk for MACE was 4.28 (95% CI 2.84–6.45) folds higher after AKI as compared to those without AKI. The magnitude of association was attenuated but remained statistically higher after adjustment for multiple cardio-renal risk factors and frequency of hospitalization (adjusted HR 1.81, 95% 1.16–2.82).

263 death events were identified and 94 occurred after AKI. The median time from cohort enrolment to death was 2.25 (IQR 1.13–3.83) years in participants without AKI whilst the median time from AKI to death was 0.49 (IQR 0.11–1.98) years in those with AKI occurrence. Cox regression showed that the risk for all- cause death increased 7.40 (95% CI 5.43–10.08) folds after AKI. The association remained statistically significant after adjustment for multiple clinical risk factors (adjusted HR 3.01, 95% CI 2.14–4.23, Table 3; Fig. 1).

Analyses after stratifying participants by baseline eGFR category showed that the crude incidence rates (absolute risk) for ESKD, MACE and all-cause death were markedly increased in participants with a lower eGFR (Additional file 1, supplementary Table 2). The relative risk for ESKD after AKI occurrence manifested mainly in participants with preserved eGFR but not in those with eGFR below 60 ml/min/1.73m². The risk for MACE after AKI was significantly higher in all eGFR categories in univariable analysis whereas it remained statistically significant only in subgroup with baseline eGFR between 60 and 89 ml/min/1.73m² in multivariable model. Of note, the relative risk for all-cause death after AKI was significantly elevated in all three eGFR categories in both univariable and multivariable models whilst it was more pronounced in the subgroup with baseline eGFR above 90 ml/min/1.73m² (Additional file 1, supplementary Table 3).

The risk for adverse clinical outcomes after AKI remained markedly higher than those without AKI occurrence after excluding 64 AKI events identified by 0.3 mg/dL serum creatinine change (Additional file 1, supplementary Table 4).

Of note, even AKI defined by a moderate increment (0.3 mg/dL) in serum creatinine was associated with adverse clinical outcomes (unadjusted HR [95% CI], 6.22 [3.60-10.74] for ESKD; 3.30 [1.67–6.50] for MACE; 4.41 [2.58–7.54] for death). In the multivariable model, stage 1 AKI defined by 0.3 mg/dL creatinine increment was significantly associated with 1.86 (1.07–3.23) folds increased risk for all-cause death as compared to those with non-AKI after adjustment for multiple clinical risk factors (Additional file 1, supplementary Table 5).

Discussions

In this prospective study in Southeast Asian people with type 2 diabetes, we found that, (1) AKI is a common occurrence among individuals with diabetes. Patients with old age, CVD history, minority ethnicity, usage of diuretics, poor control of hyperglycaemia, a low eGFR and a high level of albuminuria were at increased risk for development of AKI; (2) AKI survivors had a markedly increased risk for adverse cardio-renal outcomes including MACE, progression to ESKD and premature death. Importantly, the increased risk for these adverse clinical outcomes after AKI was independent of a low eGFR, albuminuria and other traditional risk factors; (3) the median time from first AKI episode to the adverse cardio-renal outcomes was less than one year after AKI and, (4) even AKI defined by a mild elevation of serum creatinine (0.3 mg/dL) was independently associated with an increased risk for premature death in individuals with diabetes.

Early studies in general population or specific subpopulations have shown that patients with diabetes are more susceptible to AKI as compared to their counterparts without diabetes^4,11,12. However, data on AKI in people with diabetes are scarce. In the current work from South East Asia, we found that AKI was common in multi-ethnic Asian people with type 2 diabetes (31.2, 95% CI 27.1–35.4, per 1000 patient-years), close to that reported in a recent large cohort study in Chinese population from Hong Kong (29.2, 95% CI 28.2–30.2, per 1000 patient-years)¹⁷. The mechanisms underlying the susceptibility to AKI in people with diabetes may be multifaceted. Diabetic kidney is characterized by hypoxia as shown by blood oxygen- dependent MRI^25,26. Also, hyperglycaemia and glomerular hyperfiltration may promote inflammation, oxidative stress, cellular senescence, tubulointerstitial fibrosis and intrarenal atherosclerosis within kidney^12,27. These structure and functional pathophysiologic abnormalities may enhance the susceptibility of diabetic kidney to acute injury.

Study of clinical predictors which are prospectively associated with AKI may identify high risk patients for intensive surveillance and inform strategies to target modifiable risk factors to mitigate AKI risk. Our finding of an independent association of minority ethnicity with a high risk for AKI is novel although the mechanisms underlying the linkage are only partly understood. Malay and Indian Asian participants were more likely to have a CVD history (19.6%, 19.2% and 14.3%, respectively, P = 0.02) and had a higher HbA1c (8.9 ± 2.4% [74 ± 20 mmol/mol], 9.0 ± 2.1% [75 ± 20 mmol/mol] and 8.4 ± 2.0% [68 ± 16 mmol/mol], P < 0.001) as compared to Chinese counterparts. Also, Malay participants had a lower eGFR (74 ± 31, 86 ± 27, 82 ± 31 ml/min/1.73m², P < 0.001) and high ACR (median [IQR], 81 [18–452], 31 [10–132] and 37 [10–215], P< 0.001) as compared to the other two ethnic groups. Inadequate control of these cardio-renal risk factors may partly account for the high risk for development of AKI. Beside older age and CVD history, both low eGFR and high albuminuria are independent predictor for AKI in our study population, which are in line with most early studies in people with and without diabetes^11,16,28. Together with early studies, findings from our work reinforces the notion that CKD and AKI are inter-related syndromes²⁹. The other modifiable risk factor identified in the current study, which was also reported in two early studies in people with diabetes^17,30, was the strong association of HbA1c with AKI. Data from our study suggested that 1% increment in HbA1c was associated with 15% increased risk for AKI even after adjustment for other clinical risk factors including eGFR and albuminuria, highlighting the importance of better glycaemic control to reduce the risk of AKI in patients with diabetes. On the other hand, usage of diuretics independently predicted AKI in our study. Diuretics has been associated with AKI in controlled trials^31,32. However, it remains unknown whether diuretics is harmful or helpful for AKI prevention³². We cannot exclude the possibility that the association of diuretics usage and a high risk for AKI may be attributable to confounding by indication, i.e., patients with heart failure and fluid retention are more likely to be treated with diuretics whilst these patients are at high risk for development of AKI. Further studies are warranted to address this question.

Our finding that AKI was independently associated with a significantly high risk for progression to ESKD was in line with early studies which showed that any AKI episode was a risk factor for progressive kidney disease in patients with diabetes^13,16,17. The magnitude of association in our population (adjusted HR 1.60, 95% CI 1.02–2.49, Table 3) was closer to that reported by Monseu et al. (adjusted HR 2.47, 95% CI 1.50–4.05), but much lower than that observed by Jiang et al. (adjusted HR 12.1, 95% CI 10.7–13.6). The discrepancy may be attributable to the difference in baseline characteristics among the three cohorts. For example, participants in our cohort and those from Monseu et al. had a longer diabetes duration (12, 15 versus 5 years) and a much higher percentage of participants were on insulin treatment (43%, 60% versus less than 20%) than the Hong Kong cohort^16,17. Hence, the relative risk of ESKD after AKI in our cohort may be lower due to the high risk of ESKD in the non-AKI group. Given that the risk of progressive CKD may persist for up to 10 years in AKI survivors even after apparent full renal recovery³³, our study added new data to support that patients with AKI should have periodic assessment of renal function and urinary ACR measurements to assess prognosis and outcome after hospitalization discharge²⁹. Observational data showed that few patients with AKI received follow-up after hospital discharge. Hence, more efforts are warranted to fill this unmet clinical need.

An increased risk of atherosclerotic cardiovascular diseases after AKI has been observed in our current study, in consistence with the early study in French people with diabetes and several studies in cohorts without diabetes^16,34,35. The pathophysiology of cardiovascular damage after AKI may be multifactorial. A transient episode of ischemic AKI might act as a cardio-depressant by increasing plasma levels of circulating inflammatory mediators such as tumour necrosis factor α and Interleukin-6³⁶. Experimentally induced ischemic AKI in rats also affected the metabolomic profile in the heart, manifested as amino acid depletion and a shift toward anaerobic energy production³⁷. The other pathways might involve cardiac fibrosis, neurohormonal activation, increased uremic toxins and electrolyte disturbances in patients with AKI⁷.

The clinical relevance of mild AKI is still debatable. Some early studies showed that mild AKI defined by a 0.3 mg/dL increase in serum creatinine was associated with a poor long-term outcome^38,39. However, others argue that mild AKI episodes may only reflect changes in hemodynamic rather than intrinsic damage to the kidney⁴⁰. For example, the development of mild AKI has been associated with improved survival in an early study⁴¹. Our findings suggest that stage 1 AKI defined by a mild serum creatinine elevation may be clinically relevant in patients with type 2 diabetes because it was associated with an 86% increase in all-cause mortality even after adjustment for multiple clinical risk factors (Additional file 1, supplementary Table 5). Hence, it is reasonable to postulate that the occurrence of mild AKI may be a marker for reduced renal reserve (renal frailty) in patients with diabetes⁶. This group of patients also deserve intensive surveillance after AKI.

Most early studies on AKI were retrospective and focused on specific subpopulations^6,8,15. The current work is one of the few prospective studies in patients with type 2 diabetes. The sample size of the study is relatively large. We examined several cardio-renal outcomes concurrently which are of critical importance for patients with diabetes. Unlike studies based on administrative database, multiple clinical risk factors have been considered in our analyses.

Several important limitations should be highlighted. First, we focus on only inpatient AKI in the current study. It is possible that AKI occurred in the community may have different risk factors and clinical outcomes. Second, measurement of serum creatinine was decided by clinical discretion instead of a standardized protocol. Therefore, incident rate of AKI might have been underestimated due to infrequent measurements or irregular interval between measurements of serum creatinine or lack of data on urine output. Third, aetiology of AKI may be an important determinant of clinical outcomes. We did not have information on the causes of AKI. Neither did we consider recovery of AKI in our analyses on clinical outcomes. Also, we could not study the frequency of AKI episodes on clinical outcomes due to the small number of patients with repeated AKI. Fourth, the event number of AKI was relatively small in this study. It was underpowered for subgroup analysis as shown by the few numbers of events and broad 95% CI (supplementary Tables 2 and 3). We have only 64 events of AKI defined by 0.3 mg/dl serum creatinine elevation (supplementary Table 4). We would like to emphasize that the outcomes from these subgroup analyses can only be taken as exploratory in nature awaiting further validation in large and more diverse cohorts to strengthen the robustness. Also, this is a single centre study in Southeast Asia. Therefore, the AKI prediction model and the other findings in this study may not be readily extrapolated to other populations with diabetes or other ethnic groups. Finally, as that for all observational studies, residual confounding is inevitable. For example, we do not have information on social economic status or comorbidities such as liver diseases which may be important determinant of clinical outcomes.

In conclusion, AKI occurrence is common in patients with type 2 diabetes and associated with markedly increased risks for progression to ESRD, development of atherosclerotic cardiovascular disease, and premature death independent of traditional cardio-renal risk factors. More efforts are needed to identify high risk patients for stratified management to prevent AKI. Patients with any episode of AKI deserve intensive surveillance on cardio-renal dysfunction after hospitalization discharge.

Abbreviations

AKI

Acute kidney injury

ACR

Albumin-to-creatinine ratio

BMI

Body mass index

CKD

Chronic kidney disease

CKD-EPI

Chronic kidney disease epidemiology collaboration

CVD

Cardiovascular disease

ESKD

End stage kidney disease

MACE

Major adverse cardiovascular event

RAS

Renin-angiotensin system

Author contributions

JJL and JL designed the study. SL, AL, KA, CC, YMS, RLG, KA and SCL collected data and contributed important intellectual knowledge. JJL, JL and HZ performed data analysis. JJL drafted the manuscript. All co-authors revised the manuscript critically for important intellectual contents and approved publication of the manuscript. JJL and SCL are guarantors of this work and, as such, had full access to all data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

The work was supported by Khoo Teck Puat Hospital STAR Grant (23201 and 20201) and NMRC Clinical Scientist Award (MOH-000714-01). The funder has no role in study design, data analysis, manuscript writing and decision for publication.

Data availability

The datasets generated and analysed during the current study are not publicly available due to ethical restrictions. However, anonymized, or aggregated data are available from the corresponding author on reasonable request and upon approval from Singapore National Healthcare Group Review Board.

Declarations

Ethics approval and consent to participates

This study was approved by Singapore National Healthcare Group Domain Specific Review Board (DSRB 2000/00541). It has been conducted in compliance with principles laid by Helsinki Declaration. Written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.