A systematic review and meta-analysis of acute kidney injury in the intensive care units of developed and developing countries

Objectives Although the majority of the global population lives in developing countries, most of the epidemiological data related to intensive care unit (ICU) acute kidney injury (AKI) comes from developed countries. This systematic review aims to ascertain the methodology of studies on ICU AKI patients in developing and developed countries, to determine whether epidemiological comparisons between these two settings are possible, and to present a summary estimate of AKI incidence. Methods A systematic review of published studies reporting AKI in intensive care units (2005–2015) identified in PubMed, LILACS, and IBECs databases was conducted. We compared developed and developing countries by evaluating study methodology, AKI reference serum creatinine definitions, population characteristics, AKI incidence and mortality. AKI incidence was calculated with a random-effects model. Results Ninety-two studies were included, one of which reported data from both country categories: 60 from developed countries (1,057,332 patients) and 33 from developing countries (34,539 patients). In 78% of the studies, AKI was defined by the RIFLE, AKIN or KDIGO criteria. Oliguria had 11 different definitions and reference creatinine 23 different values. For the meta-analysis, 38 studies from developed and 18 from developing countries were selected, with similar AKI incidence: 39.3% and 35.1%, respectively. The need for dialysis, length of ICU stay and mortality were higher in developing countries. Conclusion Although patient characteristics and AKI incidence were similar in developed and developing countries, main outcomes were worse in developing country studies. There are significant caveats when comparing AKI epidemiology in developed and developing countries, including lack of standardization of reference serum creatinine, oliguria and the timeframe for AKI assessment. Larger, prospective, multicenter studies from developing countries are urgently needed to capture AKI data from the overall population without ICU access.

It is widely accepted that AKI characteristics are different in developed and developing countries due to contrasting socioeconomic patterns, government health expenditures, heath service infrastructure and AKI etiology [7].
Although the vast majority of the global population lives in developing countries, most of the epidemiological data from ICU patients with AKI comes from developed countries. Comparisons of these two different settings are scarce. A multicenter prospective study found higher mortality for ICU AKI patients in developing countries [8], which might be related to an inadequate number of ICU beds in relation to population size and the difficulty of health care access, among other reasons. To determine whether the outcomes in these populations are comparable, it is necessary to evaluate whether there are differences in patient characteristics as well as the methodological aspects among the analyzed studies.
This systematic review covers methodological aspects, including AKI and reference serum creatinine definitions, as well as the main characteristics and outcomes of ICU AKI patients from studies in developed and developing countries. We aim to determine whether epidemiological comparisons between these two country categories are appropriate with the available data and to estimate their AKI incidence using meta-analysis.

Database search
This systematic review was conducted using the recommendations of the "Cochrane Handbook for Systematic Reviews of Interventions" [9] (Fig 1). A systematic electronic search was performed to identify all original studies which might include acute kidney injury patients in intensive care units published from 2005 until 2015, including the keyword terms: "acute kidney injury", "acute kidney failure", "acute kidney insufficiency", "acute renal injury", "acute renal failure", "acute renal insufficiency", "intensive care units", "critical ill patient and critical ill". We accessed the PubMed, CENTRAL (Cochrane Controlled Register of Trials), LILACS (Latin American and Caribbean Health Sciences Library), and IBECs (Spanish Bibliographical Index of Health Sciences) databases. The search strategies for each database can be found in S1 File. This search was last updated on July 31, 2015, and the language was limited to English, Spanish, French, Italian and Portuguese. The manuscripts were manually analyzed in order to Reviews of Interventions" [9]. When necessary, a final decision was achieved by consulting a third coauthor.

Data collection process
Two authors performed independent data extraction using standardized data extraction forms.
The following data were retrieved: 1. Research place and study description: country, study type, length of data collection, number of ICUs, type of ICU, number of patients.
2. AKI characteristics: incidence, definition, criteria used for definition, timeframe for AKI assessment, percentage of oliguric patients, oliguria definition, reference serum creatinine (SCr) defined as the value used for comparing increased SCr to establish the diagnosis of AKI.
3. AKI patient characteristics: mean age, gender ratio, previous comorbidities, patient source, severity of illness score. 4. Outcomes: length of ICU stay, length of hospital stay, frequency of RRT use, type of RRT, mortality.

Statistical analysis
The frequencies of studies were calculated by considering those that assessed the respective data. The number of studies used for each calculation is shown after the presented frequency. Weighted means and percentages of reported data were estimated using the study population as weight.
The pooled point AKI incidence of developed and developing countries´studies was estimated for studies that used the RIFLE, AKIN or KDIGO criteria for the AKI definition. All estimates and their 95% confidence intervals (95% CI) were calculated using a random-effects model for descriptive data analysis. Subgroup analyses were conducted with studies grouped by each different criterion for the utilized AKI definition. Heterogeneity and consistency were evaluated using Cochran's Q and the I 2 statistics, respectively. Funnel plots were used to evaluate publication bias. The analysis was performed in Microsoft Excel, using the step-by-step approach constructed by Neyeloff et al. [11] to analyze descriptive data. The pooled AKI incidence of both country groups were compared using the estimated 95% confidence interval.
The 92 studies report data from 1,091,871 patients. The number of patients from developed countries was 30 times higher; 1,057,332 vs. 34,539 patients in developed versus developing countries, respectively. Larger cohorts with more than 5,000 patients were more frequent in developed countries [33.3% (20/60)], whereas in developing countries, 87.8% (29/33) of the studies included less than 1,000 patients (Tables 2 and 3). The number of ICUs included in developed countries was significantly higher than in developing countries (990 vs. 86 ICUs). In fact, in developed countries, 41.6% of the studies (25/60) assessed more than five ICUs, while in developing countries, 81.8% of studies (27/33) assessed only one ICU. The majority of studies (82.7%, i.e., 77/93) evaluated patients from ICUs classified as "mixed" (Tables 2 and 3).
Definition of AKI. Both developed and developing country studies frequently used RIFLE, AKIN, KDIGO, as defined or modified (Fig 2). In developed countries, AKIN and RIFLE were the most frequently used criteria [37.9% (22/58) and 34.4%, (20/58), respectively], followed by increased SCr (24.1%, 14/58). In developing countries, RIFLE and AKIN were also  The oliguria definition was stated in 81.6% (49/60) of developed country studies and in 97% (32/33) of developing country studies. In five studies from developed countries and two from developing countries, the oliguria definition was not available in the manuscript and was obtained through contact with the researcher by electronic mail. In total, 11 different definitions for oliguria were used. The most frequent was "urinary volume < 0.5 ml/kg/h for 6 h", which was found in 81.6% (40/49) of developed and in 68.7% (22/32) of developing country studies (Tables 4 and 5 and Fig 2).
Reference serum creatinine definition. The reference SCr definition was available in 68.3% (41/60) of studies in developed countries and 57.5% (19/33) in developing countries. In 12 studies from developed countries and 8 from developing countries, this information was not available in the manuscript, and it was obtained through contact with the researcher. We found 29 different definitions for reference SCr (Tables 4 and 5 and Fig 2), and there was no particular dominant definition.

Timeframe for AKI assessment
The timeframe for AKI diagnosis was available in 58.3% (35/60) of studies in developed countries and 69.6% (23/33) in developing countries. Among those reporting this information, the most used definition for timeframe was "until ICU discharge or death" found in 42.8% (15/35) of developed and in 39.1% (9/23) of developing country studies (Tables 4 and 5).

Incidence and AKI etiology
The incidence of AKI was reported in 91.3% (85/93) of the analyzed studies (Tables 4 and 5).
Most studies of both developed (54.7%, 29/53) and developing countries (82.1%, 23/28) reported an AKI incidence up to 40%. According to different AKI definitions, AKI incidence varied from 2.1% [12] to 78.7% [13] in developed countries and from 0.5% [14] to 65% [15] in developing countries. AKI etiology was described in 46.6% (28/60) of studies from developed countries and in 66.6% (22/33) of studies from developing countries. Sepsis and shock were the most common causes of AKI in both developed and developing countries (see Tables 6 and 7). The frequency of sepsis as the cause of AKI in developed countries ranged from 4.4% [13] to 100% [16], and half of the studies had frequencies greater than 40% (9/18). In developing countries, the frequency of sepsis as a cause of AKI ranged from 2.9% [17] to 100% [18], and 66.7% of the studies reported a frequency greater than 40% (12/18). In developed countries, shock was less frequently reported as a cause of AKI than in developing countries. Only one study reported tropical diseases (leptospirosis) as contributing to AKI etiology (Tables 6 and 7).

Patient characteristics
Age. Almost 40% (22/56) of the studies in developed countries described a mean age above 65 years in AKI patients (ranging from 37 [12] to 72 years [19]), while only 12.1% (4/33) of the studies in developing countries reported an age higher than 65 years in AKI patients (ranging from 26 [20] to 70 years [21]) (Tables 6 and 7). The weighted mean ages were 62.0 and 56.8 years for developed and developing country patients, respectively.
Gender and ethnicity. Male sex was predominant in AKI patients in both groups of countries: 59.6% (34/57) and 67.8% (19/28) of the studies in developed and developing countries, respectively, reported a male frequency above 60%. The weighted male frequencies were 67.1% and 64.5% for developed and developing country patients, respectively. Only 9.5% (9) of the studies reported the patients' ethnic background (Tables 6 and 7). Comorbidities. Comorbidities were assessed in 51.6% (31/60) and 78.7% (26/33) of studies from developed and developing countries, respectively. The most prevalent comorbidities were cardiovascular diseases (CVD), diabetes and chronic respiratory disease.
The frequency of diabetes was similar in both groups of countries. In approximately 60% (16/27 in studies of developed and 10/18 of developing countries) of studies, the prevalence of diabetes was over 20% in the studied population (Tables 6 and 7). The weighted diabetes frequencies were 27.3% and 24.5% for developed and developing country patients, respectively.
Severity scores. The most reported severity scores were APACHE II and SOFA. The Apache II score in AKI patients was reported in 32 and 18 studies from developed and developing countries, respectively. The APACHE II score had a similar distribution in the two groups of countries, ranging from 9 [22] to 56 [23] in developed country studies and from 10    [17] to 50 [24] in studies from developing countries. Approximately half of the studies had an APACHE II score up to 20 (16/32) in developed and in developing countries (9/18) (Tables 6  and 7). The weighted APACHE II scores were 18.7 and 21.0 for developed and developing country patients, respectively. The SOFA score was reported in 21 and 10 studies from developed and developing countries, respectively. The distribution of the SOFA score was similar between groups, ranging from 3 [25] to 13.4 [26] in studies from developed countries and from 3 [27] to 9.8 [28] in developing country studies. In studies where this information was available, a SOFA score over 5 was reported by 71% (15/21) of studies from developed countries and 80% (8/10) of studies in developing countries (Tables 6 and 7). The weighted SOFA scores were 7.6 and 8.2 for developed and developing country patients, respectively.
Patient location before ICU. Most of the AKI patients who were admitted to the ICU came from surgical and clinical wards units. The majority of manuscripts from both developed countries (13/16) and developing countries (9/11) reported that up to 50% of patients with AKI had hospital admission in emergency situations (Tables 6 and 7).

Outcomes
Length of ICU and hospital stay. In developed and in developing countries, 71.6% (43/ 60) and 69.6% (23/33) of the studies reported the length of ICU stay for AKI patients, which ranged from 1 to 22 days and from 5 to 23 days, respectively. The reported ICU stay was longer than seven days in 38.6% (17/44) and 80% (20/25) of developed and developing country studies, respectively (Tables 8 and 9). The weighted mean ICU stay lengths were 7.2 and 12.2 days for developed and developing country patients, respectively. Acute kidney injury in the ICU: Developed versus developing countries In non-AKI patients, ICU stays longer than 7 days were not reported in developed country studies (0/24) but occurred in 44% (4/9) of studies in developing countries. In developed and developing country studies, 38.3% (23/60) and 21.2% (7/33) of the studies reported the length of hospital stay, which ranged from 8 to 31 days and 10 to 29 days, respectively. Hospital stays were longer than 15 days in 58.3% (14/24) and 66.7% (6/9) of developed and developing country studies, respectively (Tables 8 and 9). In non-AKI patients, the hospital stay was longer than 15 days in 25% (4/16) and 33.3% (1/3) in developed and in developing country studies, respectively. The weighted mean lengths of hospital stay were 15.5 and 23.6 days for developed and developing country patients, respectively.
Renal replacement therapy. Sixty-three percent of the analyzed studies reported the use of renal replacement therapy (RRT) in ICU AKI patients. In developed countries, 21% of the 38 studies with available data referred to the use of RRT in ICU AKI patients as being greater than 30%. In developing countries, 48% of the 21 studies with available data showed that the frequency of RRT use was higher than 30% in ICU AKI patients (Tables 8 and 9). The weighted frequencies of RRT were 8.8% and 23.8% for developed and developing country patients, respectively.
Mortality. Reported ICU mortality in AKI patients was greater in developing country studies. AKI mortality greater than 60% was reported in 15.9% (7/44) of studies from developed countries and 56% from developing countries (14/25) (Tables 8 and 9). The weighted frequencies of mortality were 30.8% and 54.8% for developed and developing country patients, respectively.

Synthesis of AKI incidence
Pooled AKI incidence estimates for developed and developing countries in the meta-analysis are presented in Table 10, according to the AKI definition used. The RIFLE, AKIN or KDIGO criteria for AKI definition was used, as defined, by 39 and 21 studies in developed and developing country studies, respectively. One study in developed countries and 3 in developing countries studies did not report AKI incidence, so 38 and 18 studies, respectively, had an AKI incidence estimation included in the meta-analysis. The pooled estimate of AKI incidence in developed and developing countries is shown in Fig 3. There was a tendency towards a greater incidence in developed countries, although this was not significant. When only prospective studies were analyzed, this tendency disappeared (Table 10). Fig 4 shows the funnel plot for both country groups in which individual study frequency of AKI is a function of their sample size with the pooled incidence of studies that used the RIFLE, AKIN or KDIGO criteria for the AKI definition being depicted as a black line. Note that the Fig 4A (developed countries) had a 10-fold greater sample size than Fig 4B (developing countries). The studies with a greater sample size depart from the polled estimated AKI incidence, suggesting the possibility of publication bias or bias resulting from the lack of standardizing reference creatinine, oliguria, and the timeframe for AKI assessment.

Discussion
We found a high incidence of AKI in both country categories, and a tendency towards a greater incidence in developed countries. Due to the differences in AKI definitions, timeframe and the types of studied population, the reported incidence varied from 0.5% to 78%.
Our review covered a 10-year period. Thus, different AKI definitions were used for AKI assessment, including the RIFLE, AKIN and KDIGO criteria [29,30,31]. Only two-thirds of the studies reported the definition for reference serum creatinine, with 29 different definitions used, which results in high heterogeneity of AKI incidence estimate [32]. In the most recent AKI definitions (RIFLE, AKIN and KDIGO), the reference SCr is the value observed up to seven days or 48 hours before the SCr increase defining the AKI diagnosis. However, we Acute kidney injury in the ICU: Developed versus developing countries Acute kidney injury in the ICU: Developed versus developing countries observed that several studies used as reference SCr values obtained months, or even one year before the AKI episode, which is not consistent with the current AKI definitions. When the reference serum creatinine was not available, the MDRD formula has been used for estimation of the missing SCr value, which is a flawed methodology as it can misdiagnose AKI in CKD patients [12,33,34]. The oliguria definition was more uniform, with recent studies correctly using the RIFLE, AKIN and KDIGO oliguria definitions. The addition of urine output criteria was associated with higher and earlier AKI detection and incidences in critically ill patients [32]. Another important caveat to create a valid comparison between developed and developing countries is the striking differences in the number of studies and the sample sizes. Eighty per cent of the world population lives in developing countries, but only one-third of the studies sample reported data from them, with the majority assessing a single center with a relatively small number of patients [35]. Moreover, approximately half of the studies from developing countries were from Brazil, and only two were from Africa. On the other hand, approximately 40% of developed country studies assessed more than five centers. The sample size from developed countries studies was more than 30-fold greater compared to those from developing countries. There is a clear underrepresentation of developing countries that is probably caused by a lack of health resources and electronic medical records, as well as difficulty in gathering epidemiological data and, consequently, conducting adequate large observational studies. These studies can be more capable to determine the true burden of a disease than trials and more valuable in assessing the incidence and prevalence of the disease [36]. A snapshot of worldwide AKI incidence found more severe AKI presentation in patients from developing countries, which was considered to be due to delay in AKI recognition and treatment, adversely affecting the outcomes [7,37].
The incidence of AKI development in the ICUs was similar in both types of countries, with a numeric tendency to be greater in developed country studies. When only prospective studies were analysed, this tendency disappeared. Developing and developed countries have very distinct healthcare patterns. In developing countries, deficiencies in health structure, long distance from the community to the hospital and poor transportation systems limit patient access to healthcare. Lack of universal health coverage and insufficient funding for the health system Acute kidney injury in the ICU: Developed versus developing countries imposes significant cost of treatment for the patients and family, including high cost procedures such as ICU and renal replacement therapy [38]. Tropical infectious diseases, animal venoms, natural medicine, abortion and eclampsia are known to be important AKI etiological factors in developing countries [39,40]; however, their incidence was extremely low in the ICU population. This is likely due to the limited number of ICUs, which are located mostly in larger urban cities, as well as inadequate recognition of high-risk AKI patients in the primary health system. Furthermore, difficulty transporting patients due to geographical and economic issues may contribute to this situation [39]. As a consequence, developing country's ICUs reflect tertiary hospitals and university hospitals mostly from an urban population. Thus, patient characteristics were similar in both types of country. Sepsis and shock were the main causes of AKI in both developed and developing countries, but the frequency of sepsis Acute kidney injury in the ICU: Developed versus developing countries was approximately 50% greater in developing country studies. In developed country studies, AKI patients were older, which likely reflects higher population longevity, better socioeconomic conditions and more structured health services. Overall, cardiovascular diseases were the most frequently reported comorbidity, although they were more frequent in developed country studies. AKI was associated with poor outcomes, higher length of stay (LOS) and mortality, which is consistent with other studies [41,42]. In developing country studies, AKI had higher LOS and mortality compared to developed countries, although patients were younger, had less CVD and had similar APACHE II scores. Difficulty accessing health services [39,43] and lack of infrastructure, including ICU beds and human resources for care of the Acute kidney injury in the ICU: Developed versus developing countries critically ill in these countries [44,45], are probably the cause of worse outcomes in such a low resource setting. It is possible that the patients treated in developing countries are transferred to an ICU at a late stage of disease progression and have a reduced change of recovering [45]. The finding of higher frequency RRT use in developing countries supports this hypothesis. This systematic review highlighted important caveats for the comparison between ICU AKI epidemiology in developed and developing countries. The vast majority of studies assessed university tertiary hospitals, limiting the generalizability of the results. Different AKI definitions were used over time, and even when the new AKI criteria were used, there is an important lack of standardization for reference serum creatinine. Most of the studies from developing countries were single center. The number of patients and ICUs assessed in developed country studies was greater than 30-fold and 10-fold higher than in developing countries, respectively, highlighting the underrepresentation of developing countries.

Conclusion
AKI incidence was high in both types of countries. Patient characteristics were mostly similar, but outcomes were worse for patients in developing country studies. Despite patient´s similarities, the non-inclusion of secondary hospitals and the differences in the number of studies and sample sizes exemplify the challenge of comparing developing and developed country AKI epidemiology. The widespread application of AKI definitions has made it possible to compare AKI epidemiology across different settings. However, an effort to standardize reference serum creatinine, oliguria and the timeframe for AKI assessment is crucial. There is an urgent need for larger, prospective, multicenter studies that assess broader populations from developing countries.