Modifiable Lifestyle Factors for Primary Prevention of CKD: A Systematic Review and Meta-Analysis

Significance Statement

Although CKD incidence is increasing, no evidence-based lifestyle recommendations for CKD primary prevention apparently exist. To evaluate evidence associating modifiable lifestyle factors and incidence of CKD, the authors undertook a systematic review and meta-analysis. Their analysis, which included 104 observational studies of 2,755,719 participants, demonstrated consistency of evidence for a number of measures associated with preventing CKD onset, including increasing dietary intake of vegetables and potassium (21% reduced odds and 22% reduced odds, respectively), increasing physical activity levels (18% reduced odds), moderating alcohol consumption (15% reduced risk), lowering sodium intake (21% increased odds), and stopping tobacco smoking (18% increased risk). In the absence of clinical trial evidence, these findings can help inform public health recommendations and patient-centered discussions in clinical practice about lifestyle measures to prevent CKD.

Abstract

Background

Despite increasing incidence of CKD, no evidence-based lifestyle recommendations for CKD primary prevention apparently exist.

Methods

To evaluate the consistency of evidence associating modifiable lifestyle factors and CKD incidence, we searched MEDLINE, Embase, CINAHL, and references from eligible studies from database inception through June 2019. We included cohort studies of adults without CKD at baseline that reported lifestyle exposures (diet, physical activity, alcohol consumption, and tobacco smoking). The primary outcome was incident CKD (eGFR<60 ml/min per 1.73 m²). Secondary outcomes included other CKD surrogate measures (RRT, GFR decline, and albuminuria).

Results

We identified 104 studies of 2,755,719 participants with generally a low risk of bias. Higher dietary potassium intake associated with significantly decreased odds of CKD (odds ratio [OR], 0.78; 95% confidence interval [95% CI], 0.65 to 0.94), as did higher vegetable intake (OR, 0.79; 95% CI, 0.70 to 0.90); higher salt intake associated with significantly increased odds of CKD (OR, 1.21; 95% CI, 1.06 to 1.38). Being physically active versus sedentary associated with lower odds of CKD (OR, 0.82; 95% CI, 0.69 to 0.98). Current and former smokers had significantly increased odds of CKD compared with never smokers (OR, 1.18; 95% CI, 1.10 to 1.27). Compared with no consumption, moderate consumption of alcohol associated with reduced risk of CKD (relative risk, 0.86; 95% CI, 0.79 to 0.93). These associations were consistent, but evidence was predominantly of low to very low certainty. Results for secondary outcomes were consistent with the primary finding.

Conclusions

These findings identify modifiable lifestyle factors that consistently predict the incidence of CKD in the community and may inform both public health recommendations and clinical practice.

CKD affects 10% of the world’s population¹ and ranks in the top ten noncommunicable diseases contributing to disease and disability.² Its incidence is increasing worldwide,¹ and mortality owing to CKD rose between 2005 and 2017 from 0.9 million to 1.2 million deaths annually.² CKD imposes a significant economic burden on patients and society; many developed countries spend between 2% and 3% of their annual health care expenditure on RRTs (that is, chronic dialysis or kidney transplantation) for 0.2% of the total population.³

Because prevention is more effective than cure, avoiding exposures to hazards that may cause CKD in the community is a key priority. This is particularly important because the management of patients with established CKD may require dietary adaptations that diametrically differ from those that are needed for primary prevention.⁴ However, no evidence-based lifestyle recommendations for the primary prevention of CKD are apparent. Instead, current advice extrapolates from recommendations and inferences drawn from literature concerning cardiovascular disease, hypertension, and diabetes.^5–7 Furthermore, the World Health Organization discusses CKD as a complication of cardiovascular disease and diabetes, which further complicates effective public health messages.⁸ To address this critical knowledge gap, we hereby evaluated the consistency of evidence associating modifiable lifestyle factors (specifically diet, physical activity, alcohol consumption, and tobacco smoking) and the incidence of CKD.

Methods

This systematic review and meta-analysis was conducted according to the Meta-analysis of Observational Studies in Epidemiology checklist for observational studies,⁹ and it was prospectively registered in PROSPERO (CRD42019137328).

Data Sources and Searches

We searched MEDLINE, Embase, and CINAHL (database inception through June 2019) without date or language restriction (Supplemental Table 1). We used EndNote to deduplicate and manage citations. For publications in languages other than English, we used translations by native speakers and where not possible, Google Translate. Author pairs completed the title and abstract screening and full-text screening, and they manually searched the reference lists of eligible studies and reviews to identify other potentially relevant studies by snowballing. We included a third investigator (J.T.K. or J.-J.C.) and used consensus to resolve disagreements.

Study Selection

Inclusion criteria were adult participants without established CKD (i.e., GFR>60 ml/min²) at baseline,¹⁰ including those that reported people without CKD as a subgroup of a larger study; cohort studies, prospective or retrospective; and outcome data (primary or secondary) reported on the basis of levels of exposure to a lifestyle factor, including diet (foods or nutrients), physical activity, alcohol consumption, and tobacco smoking. We excluded studies of participants with albuminuria at baseline, whenever this information was available. We did not consider diet pattern exposures in our review, as these have recently been systematically evaluated.^11,12

The primary outcome was incident CKD, defined as the development of GFR<60 ml/min per 1.73 m² during follow-up.¹⁰ Secondary outcomes were other surrogate measures of CKD: RRT, GFR decline, or albuminuria.

Data Extraction and Quality Assessment

One author from each team extracted the data from the included studies using standardized data extraction forms created in Microsoft Excel, which were checked by a second author before analysis. Supplemental Table 2 shows the data extracted for analysis. Where we identified more than one publication for a single cohort study reporting on the same lifestyle factor, we used the dataset from the published study contributing the highest number of participants. Within each meta-analysis, we conducted a sensitivity analysis examining the effects of substituting these alternative sources of data on the same cohort. Where questions arose concerning a study’s eligibility, missing data, or further information on results, we contacted corresponding authors to request this information. We used author responses to make decisions on a study’s eligibility and analysis approach.

We assessed risk of bias for each included study using the Newcastle–Ottawa Quality Assessment Scale,¹³ which assesses the methodological quality of the studies across three domains: selection (or representativeness) of cohorts, comparability (of cohorts due to design or analysis), and outcomes (assessment and follow-up) (Supplemental Table 2). We applied the Grading of Recommendations Assessment, Development and Evaluation methodology (GRADE) criteria to rate the quality of evidence (low, moderate, or high) for each outcome. In analyses that included five or more studies, when the primary outcome showed low or moderate heterogeneity, we constructed funnel plots to examine for effects that may represent publication, selection, or reporting bias.

Data Synthesis and Analyses

We used RevMan 5.3 and Stata version 16.0 (StataCorp, College Station, TX) to meta-analyze data when a particular lifestyle factor was reported for the same outcome from three separate cohorts; otherwise, associations were tabulated and reported using frequencies. Given the observational design and highly varied sample sizes of the primary studies, we used the random effects model (DerSimonian and Laird) for all meta-analyses. Robustness of the association estimate was evaluated by comparing if any statistically significant association identified through random effect models became nonsignificant under a fixed effects model. The overall association estimates for all binary outcomes are expressed as odds ratios (ORs) or relative risks (RRs), depending on which ratio was predominantly used in the original data, together with 95% confidence intervals (95% CIs). For studies reporting changes in mean values at the end of their observation follow-up periods, we extracted and tabulated the end of the study values with their associated variance. If the association compared a lifestyle factor with a reference exposure that was not homogenous with the other included lifestyle exposures (e.g., high adherence versus low adherence), then the ratio methods were inverted and combined into the meta-analysis. We used chi-squared testing at an α of 0.05 and with the I² statistic, rated low (<25%), moderate (25%–75%), or high (>75%) to assess for heterogeneity. Supplemental Table 2 gives details of planned subgroup and sensitivity analyses to explore reasons for heterogeneity. In addition to these linear analyses for alcohol consumption, we also conducted a dose-response random effects metaregression analysis to account for potential changes in the risk estimate at different dosages of alcohol consumption (methods are reported in Supplemental Table 2).

Results

The electronic search retrieved 45,111 citations, of which a total of 104 studies encompassing 2,755,719 participants met the inclusion criteria (Supplemental Figure 1). The characteristics of included studies are reported in Supplemental Table 3. Lifestyle exposures and associations to kidney outcomes varied across the studies, with n=56 reporting diet exposures, n=18 reporting physical activity, n=27 reporting alcohol consumption, and n=30 reporting tobacco smoking. The outcomes evaluated included incident CKD in n=51 studies, RRT in n=17 studies, GFR decline in n=32 studies, and albuminuria in n=20 studies.

Lifestyle Hazards and Incident CKD

The association between lifestyle factors and incident CKD was reported in 51 studies of 1,221,018.^14–72 The main results of the meta-analysis are summarized in Supplemental Figure 2.

Diet

The association between 57 different dietary factors with incident CKD was described in 31 studies of 176,625 participants.^{14,20,24,26,27,30,38,39,41,46,47,49,51,57,60,61,66,69,70} Meta-analysis was possible for nine dietary factors (Table 1), of which higher vegetable intake (OR, 0.79; 95% CI, 0.70 to 0.90; I²=57%; evidence quality: low) and higher potassium intake (OR, 0.78; 95% CI, 0.65 to 0.94; I²=48%; evidence quality: low) were consistently associated with lower odds of CKD. Higher sodium intake was associated with higher odds of CKD (OR, 1.21; 95% CI, 1.06 to 1.38; I²=59%; evidence quality: moderate) (Figure 1, Table 1). Meta-analysis of studies evaluating carbohydrate intake, fish, fruit, phosphorus, protein, and sugar-sweetened beverages consumption showed no clear association (Table 1).

Table 1.

Meta-analyses performed in the review showing the association of alcohol consumption, diet, physical activity, and smoking exposure with the risk of incident CKD

Exposure	Studies	Participants	Association Ratio [95% CI]	Heterogeneity I2, %	Evidence Qualitya
Diet factors
Fish	3	21,226	OR, 0.94 [0.86 to 1.02]	5	Low
Fruit	4	28,779	OR, 0.91 [0.79 to 1.06]	60	Very low
Vegetables	5	32,054	OR, 0.79 [0.70 to 0.90]	57	Low
Sugar-sweetened beverages	4	22,760	OR, 1.45 [0.97 to 2.15]	68	Very low
Carbohydrates	3	20,238	OR, 1.08 [0.85 to 1.36]	63	Very low
Protein	3	19,835	OR, 1.08 [0.91 to 1.28]	47	Very low
Phosphorus	3	23,466	OR, 1.00 [0.75 to 1.32]	72	Very low
Potassium	7	32,647	OR, 0.78 [0.65 to 0.94]	48	Low
Sodium	6	43,772	RR, 1.21 [1.06 to 1.38]	59	Moderate
Physical activity
High versus low levels	9	70,828	RR, 0.82 [0.69 to 0.98]	83	Very low
Alcohol consumption
High versus low intakes	13	216,291	RR, 0.87 [0.79 to 0.95]	43	Low
Moderate versus low intakes	7	165,415	RR, 0.86 [0.79 to 0.93]	50	Moderate
Tobacco smoking
Never versus former	6	944,931	OR, 1.09 [1.01 to 1.17]	90	Very low
Current or former versus never	12	985,086	OR, 1.18 [1.10 to 1.27]	81	Very low

^aSummary of findings, which details the reasons for downgrading evidence quality, is reported in Supplemental Table 7.

Figure 1.

There was a reduced odds of CKD in people who were exposed to a higher vegetable intake and potassium intake and an increased odds of CKD in people who were exposed to a higher sodium intake. Association of (A) vegetable intake, (B) potassium intake, (C) sodium intake, and incident CKD. Note that the association estimate for each lifestyle factor is presented on the ratio (OR or RR) that was predominantly used in the included studies. IV, inverse variance.

We were unable to pool data on 32 additional dietary factors reported across 21 studies (Supplemental Table 4).^{15,16,18,19,23,25,27,30–32,37,45–47,57,59–62,71,73} We identified 19 dietary factors for which one or both available studies showed an association in the direction indicating decreased risk: cereal fiber, coffee, dairy, fiber, folate, legumes, magnesium, nitrate, nuts, nuts and legumes, plant protein, omega-3, DHA, EPA, vitamin B12, vitamin C, vitamin D, vitamin E, and zinc. We identified two dietary factors for which one or both available studies indicated a harmful relationship: a higher sodium-potassium ratio and red and processed meat consumption (Supplemental Table 4).

Physical Activity

The association of different levels of physical activity with incident CKD was described in ten studies of 78,301 participants (Figure 2, Table 1).^{28,34,42,44,53,54,58,59,63,72} Meta-analysis of nine studies showed lower odds of CKD in people who were more physically active compared with those who were less physically active (OR, 0.82; 95% CI, 0.69 to 0.98; I²=83%; evidence quality: very low).

Figure 2.

There is a significant reduced odds of incident CKD from increased levels of physical activity.

Alcohol Intake

The association between alcohol intake and incident CKD was described in 14 studies of 211,072 participants.^{22,23,28,40,43,50,52,55,63–65,67,68,72} Compared with lower intakes, meta-analysis showed that higher consumption of alcohol (RR, 0.87; 95% CI, 0.79 to 0.95; I²=43%; evidence quality: low) and moderate alcohol consumption (RR, 0.86; 95% CI, 0.79 to 0.93; I²=50%; evidence quality: moderate) were associated with lower risk of CKD (Figure 3, Table 1). In a dose-response metaregression analysis, we modeled the relationship between incident CKD and alcohol intake using restricted cubic splines (Supplemental Figure 3, Supplemental Table 5) and found that higher alcohol intake was associated with lower risk of incident CKD. This association was generally observed throughout the whole range of alcohol consumption considered with a significant effect size (P values for nonlinearity =0.03). There was no apparent association when people who had never consumed alcohol were compared with those who had consumed it in the past (former intake) (Table 1).

Figure 3.

There was a reduced risk of CKD in people who has a higher alcohol consumption and an increased odds of CKD in people who were exposed to higher tobacco smoking. Association of (A) alcohol consumption, (B) tobacco smoking, and incident CKD. Note that the association estimate for each lifestyle factor is presented on the ratio (OR or RR) that was predominantly used in the included studies. IV, inverse variance.

Smoking

The association between smoking and incident CKD was described in 12 studies of 985,086 participants.^{28,29,35,43,50,52,53,59,63,65,68,74} Compared with people who never smoked, current and former smokers (ever smokers) showed higher odds of CKD (OR, 1.18; 95% CI, 1.10 to 1.27; I²=81%; evidence quality: very low) (Figure 3, Table 1). Compared with people who had never smoked, former smokers had increased odds of CKD (OR, 1.09; 95% CI, 1.01 to 1.17; I²=90%; evidence quality: very low) in six studies of 944,931 participants (Table 1). There was no consistent association with CKD in two studies comparing current with former smokers (Supplemental Table 4).

Lifestyle Risk Factors and Secondary Outcomes (RRT, GFR Decline, and Albuminuria)

The consistency of associations between the lifestyle factors identified in our primary analysis and other markers of kidney damage is shown in Figure 4.

Figure 4.

The association of vegetable intake, potassium intake, physical activity, alcohol consumption, sodium intake, and tobacco smoking was consistent across incident CKD and KRT, however had varying association to GFR decline and albuminuria. NA, not applicable. ^aThe four studies of sodium intake and GFR decline were not meta-analyzable because two of the studies reported means and variance but not ratio data.

The association between lifestyle factors and RRT was described in 17 studies of 990,723 participants.^{33,35,37,75–88} The association between lifestyle factors and GFR decline was described in 32 studies of 108,936 participants,^{14,20,23,34,46,49,73,86,89–105} and the association between lifestyle factors and albuminuria was described in 20 studies of 512,403 participants.^{23,39,41,88,95,100,101,105–116} Table 2 and Supplemental Material have the details. The risk of RRT was higher among current and former smokers (RR, 1.59; 95% CI, 1.30 to 1.94; I²=68%; evidence quality: moderate) compared with never smokers (Supplemental Figure 4). The risk of GFR decline was lower in persons with higher potassium intake (RR, 0.49; 95% CI, 0.31 to 0.79; I²=90%; evidence quality: very low) and those who were physically active (OR, 0.76; 95% CI, 0.59 to 0.97; I²=75%; evidence quality: very low) (Supplemental Figure 5). The risk of albuminuria was lower among physically active persons (OR, 0.88; 95% CI, 0.81 to 0.96; I²=0%; evidence quality: low) and higher among tobacco smokers (OR, 1.67; 95% CI, 1.23 to 2.26; I²=88%; evidence quality: very low) (Supplemental Figure 6).

Table 2.

Meta-analyses performed in the review showing the association of alcohol consumption, diet, physical activity, and smoking exposure with secondary outcomes (RRT, GFR decline, and albuminuria)

Exposure	Studies	Participants	Association Ratio [95% CI]	Heterogeneity I2, %	Evidence Qualitya
RRTb
Tobacco smoking
Former versus current	7	1,126,913	RR, 1.25 [1.13 to 1.39]	39	Moderate
Current or former versus never	8	1,230,390	RR, 1.59 [1.30 to 1.94]	68	Moderate
GFR declinec
Diet factors
Potassium	4	10,729	RR, 0.49 [0.31 to 0.79]	81	Very low
Protein	5	18,507	OR, 1.07 [0.96 to 1.19]	42	Low
Physical activity
High versus low levels	5	15,161	OR, 0.77 [0.63 to 0.93]	75	Very low
Alcohol consumption
High versus low intakes	5	16,580	OR, 0.88 [0.72 to 1.07]	15	Very low
Albuminuria
Diet factors
Sodium	3	11,751	OR, 1.01 [0.89 to 1.14]	0	Very low
Physical activity
High versus low levels	4	110,154	RR, 0.88 [0.81 to 0.96]	0	Low
Alcohol consumption
High versus low intakes	7	220,479	RR, 1.03 [0.88 to 1.20]	59	Low
Tobacco smoking
Current or former versus never	7	184,302	OR, 1.67 [1.23 to 2.26]	88	Very low

^aSummary of findings, which details the reasons for downgrading evidence quality, is reported in Supplemental Table 7.

^bMeta-analysis was not possible for diet factors, physical activity, or alcohol consumption.

^cMeta-analysis was not possible for tobacco smoking.

Other lifestyle hazards did not have sufficient studies to allow meta-analysis, but their associations suggested that they may be potentially protective, be potentially harmful, or have no association to secondary study outcomes on the basis of the direction of the majority (≥50%) of studies. Results are shown in Supplemental Figure 7 and fully detailed in Supplemental Tables 3 and 4. In brief, coffee consumption was associated with lower risk of incident CKD and RRT in two of two and one of two studies, respectively. Dairy intake was associated with lower risk of incident CKD and albuminuria in two of four and one of two studies, respectively. Red and processed meats were associated with lower risk of incident CKD, RRT, and albuminuria in one of two, one of one, and one of one studies, respectively.

Subgroup and Sensitivity Analyses

Subgroup analyses did not show consistent relationships with duration of exposure or geographic region. Replacing individual datasets (substituting alternative reports from the same cohort) did not meaningfully change results (Supplemental Material). We also conducted a post hoc sensitivity analysis on the way that the frequency of alcohol consumption was reported: higher “daily” consumption of alcohol had no significant association with incident CKD (seven studies; RR, 0.98; 95% CI, 0.82 to 1.18; I²=43%), whereas higher weekly alcohol consumption was associated with a lower risk of incident CKD (six studies; RR, 0.82; 95% CI, 0.75 to 0.90; I²=32%) (Supplemental Table 6). Subgroup analysis was conducted to compare studies with participants with baseline GFR ≥89.9–70 ml/min per 1.73 to participants with GFR≥90 ml/min per 1.73 and showed significant associations between study exposures (except for physical activity) and the risk of incident CKD in studies with participants baseline GFR ≥89.9–70 ml/min per 1.73. The association of vegetable intake and incident CKD was robust in both baseline GFR subgroups. Whereas in studies reporting participant baseline GFR ≥90 ml/min per 1.73 (which were 37 of 104; 12 studies were able to be meta-analyzed), these associations did not attain statistical significance. However, the direction and magnitude of observed risks in both strata were very similar (Supplemental Table 6).

Quality Assessment

The certainty (GRADE) of the evidence for each outcome meta-analyzed is detailed in Supplemental Table 8. The overall risk of bias across the studies was low (Supplemental Figures 8 and 9). Eighty-eight percent of studies had a low risk of bias for sample, whereas 10% were rated high due to being conducted in populations with established diseases (such as type 2 diabetes or hypertension) that are known to increase the risk of incident CKD. With respect to follow-up, 82% of studies had low risk of bias, and bias was judged unclear in 13% of studies that reported <4 years of follow-up; three studies (5%) had a high risk of bias due to very short follow-up periods of <2 years. The risk of bias in the "exposure" domain was rated high in 92% of studies because lifestyle behaviors were self-reported; objective measures were used in eight studies (8%). Outcomes were captured using appropriate tools (and rated low risk) in 88% of studies, and rated high risk in 6% of the studies. Potential analysis bias was low in 73% of studies, where 23% were unclear because data were not analyzed using ratio statistics to allow statistically data pooling.

Discussion

This study evaluated the association between modifiable lifestyle factors and incident CKD in the community. In predominately low- to very low–certainty evidence, we found higher intake of potassium and vegetables, lower intake of sodium, physical activity, moderate alcohol consumption, and avoidance of tobacco smoking to be consistently associated with lower risk of CKD. We identified consistency with these main results when we examined their association with other surrogates of kidney damage: RRT, GFR decline, and albuminuria.

Higher vegetable intake is associated with higher potassium intake: both dietary factors were consistently associated with fewer kidney outcomes. Low vegetable consumption is a strong predictor of mortality in the general population.^117,118 The mechanism is likely, at least in part, through the intermediary diseases that are themselves risks for CKD: cardiovascular disease, hypertension, diabetes, obesity, and metabolic syndrome.¹¹⁹ Whether there is additionally a more direct relationship between vegetable or potassium intake and kidney health requires further study.

In our analysis, higher potassium intake was associated with reduced incident CKD and GFR decline, adding to the growing evidence base for the protective association of potassium intake in other chronic conditions. In the Prospective Urban Rural Epidemiology study, higher potassium excretion was associated with a lower risk of death and cardiovascular events across >18 countries worldwide.¹²⁰ Higher intakes of potassium have also been demonstrated to lower the risk of stroke¹²¹ and lower BP in adults with and without hypertension.^122,123 In patients in the ONTARGET, a high-vascular risk cohort (most of whom did not have CKD at baseline), potassium intake was associated with a less rapid GFR decline or need for RRT.¹²⁴ In ONTARGET, the relationship between potassium intake and the outcome did not seem to differ by baseline level of albuminuria, but there was a statistically significant interaction suggesting that the protective association with potassium might not be observed in patients with GFR<45 ml/min (advanced CKD). In these patients, potassium and with it, fruits and vegetables are often restricted. When discussing dietary advice, it is critical to distinguish whether we are concerned with prevention of incident CKD, prevention of progression (and at what level of GFR), or management of hyperkalemia and management of the risk of hyperkalemia.¹²⁵

Similarly, higher sodium intake showed a consistent association with increased risk of incident CKD, RRT, and GFR decline. This supports the growing evidence from clinical trials demonstrating the effect of sodium intake on BP, proteinuria, and extracellular volume in CKD.¹²⁶ Despite existing controversy surrounding reverse causality at extremely low sodium intakes,¹²⁷ our review supports the message that higher sodium intakes are detrimental, and public health efforts should be prioritized to reduce population-wide sodium consumption. More work is needed on the optimal level of sodium restriction; our analysis focused on high versus low intake, but in the original studies, the highest exposures of salt defining high intake varied, ranging from ≥9.88 to 16.27 g/d (sodium ≥172–283 mmol/d).

A healthy diet pattern, characterized by higher intake of fruit, vegetables, low-fat dairy, fiber, and whole grains and lower intake of sodium, red meat, and sugar, has been associated with a 30% reduced odds of kidney damage.^11,12 Many of the individual diet factors we identified are characteristic of a higher diet quality and overall healthier diet pattern. For example, taking all of the data into account, diet factors possibly protective against incident CKD in our study included fruit, dairy, vegetables, fiber, legumes, nuts, potassium, and unsaturated fatty acids. In contrast, diet factors possibly harmful to incident CKD in our study included excessive carbohydrate intake, high energy intake, red and processed meats, saturated fat, sodium, sodium-potassium ratio, and sugar-sweetened beverages. The protective factors, characteristics of a healthy diet, are also strongly associated with reduced risk of overweight and obesity,¹²⁸ which are linked with increased risk of incident CKD.¹²⁹ These associations fundamentally align with guidelines for general healthy eating and cardiovascular disease prevention, which is helpful and convenient in guiding public health policy for CKD prevention that aligns with prevention of other chronic diseases.¹³⁰

We found that higher levels of physical activity were consistently associated with a lower risk of all study outcomes evaluated. This aligns with the majority of existing studies, which suggest that physical activity is associated with reduced risk of kidney damage through attenuating cardiovascular disease risk,^131,132 mortality,¹³³ and rapid declines in kidney function.¹³⁴ Whether a relationship independent of cardiovascular and vascular damage truly exists is not known on the basis of current data. Possible mechanisms include reduced BP, improved glycemic control, angiogenesis, and vascular regeneration by the upregulation of endothelial nitric oxide production and other antioxidant enzymes.¹³⁵ Although the measurement of physical activity was heterogeneous across the included studies, the most common categorization of higher levels of activity was defined as at least 30 minutes a day.^44,59 These findings suggest that public health messages should emulate those of primary prevention of cardiovascular disease: to achieve and maintain a moderate level of physical activity of at least 30 minutes a day and avoid extended periods of being sedentary.^136,137

We found an inverse association between alcohol consumption and incident CKD, RRT, and GFR decline. Although the relationship between alcohol consumption and kidney function decline has not been consistent in the current body of literature, many clinical studies have indeed shown that moderate alcohol consumption is associated with lower occurrence of CKD.^40,55,64,68 In contrast, existing literature has shown that chronic alcoholism is associated with CKD,^56,65 leading to the hypothesis that excessive alcohol consumption directly damages the kidney, independent of liver damage.^138,139 Similarly, some reports included in our review found that heavy drinking, in contrast with moderate drinking, was associated with increased risk of albuminuria and CKD.^56,105,140 This distinction (that moderate, but not high, alcohol consumption is associated with lower prevalence of albuminuria and CKD compared with abstinence or with lower intakes) is consistent with a number of other studies.^139,141,142 There is also the possibility that unadjusted confounders may be at play: for example, social integration as a product of moderate alcohol consumption and overall well-being, which is good for health.¹⁴³ It would seem, despite minor international variations¹⁴⁴ and the heterogenous categorization for the higher alcohol consumption (ranging in grams per day [range, 20–48 g/d], grams per week [range, 210 g/wk], total drinks per day [range, >1–4 drinks per day], and drinks per week [range, 5–>15 drinks per week]), that moderate alcohol consumption in line with public health guidelines seems safe and unlikely to cause kidney damage in the general population.

Current and former smokers had higher risk of incident CKD compared with those who never smoked. Tobacco smoking has been previously identified as a risk factor for incident CKD and RRT in the general population.¹⁴⁵ Although the mechanism is uncertain, smoking may lead to insulin resistance,¹⁴⁶ which might explain the large effect size (67% increase in risk) of smoking on risk of albuminuria observed in our study, consistent with previous observations.¹⁴⁷ The harmful effects of smoking may also be mediated by endothelial cell dysfunction, inducing advanced glycation end products due to glycotoxins present in cigarettes, increased vascular permeability and pathologic vascular changes of kidney disease, and insulin resistance in diabetic and nondiabetic subjects.¹⁴⁵ Taken together, these risk factors for kidney function decline clearly support a public health message to avoid tobacco smoking to prevent kidney disease as well as cardiovascular disease.

This study has strengths as the first evidence synthesis of lifestyle risk and protective factors and primary prevention of CKD: it is broad in scope, and we have taken a rigorous approach to meta-analysis. However, this study has limitations. First, there was scarce evidence for some exposures, particularly for diet factors, precluding meta-analysis; for these factors, our output is simply descriptive. Second, there was variation in the definition of the different exposures across the different lifestyle factors, in the degree of adjustment for potential confounders, and in definition and reporting for study outcomes. For example, there were no consistent adjustments for activity, age, education, and sex (as reported in Supplemental Table 3) in all studies included in the analyses. Furthermore, ten different definitions of incident CKD were used in the included studies, with similar heterogeneity in the definitions of the secondary outcomes. This leads to heterogeneity in the analysis and precludes clear recommendations in terms of possible targets, goals, or thresholds. Third, although we were able to perform meta-analysis to account for the known nonlinear associations of alcohol consumption and kidney outcomes, we were unaware of any such assumptions for the other lifestyle factors, and the original studies did not commonly report their associations in this way. The original papers usually made a linear assumption and analyzed data with linear models. It is possible that nonlinear effects or threshold effects may be missed in this way. Fourth, because we evaluated cohort studies and not randomized trials, the findings do not imply causality. However, there were no trials of this nature in our search, and it is unlikely that studies of sufficient rigor, size, and duration will be conducted, given the resource-intensive nature of such trials. We combined RR and OR; for rare events, these are numerically similar, although calculated differently. Fifth, we used the NOS to assess bias rather than the more comprehensive ROBINS-I tool¹⁴⁸ because our resources to complete the study were finite. Finally, in order to conduct meta-analysis, each putative risk and protective factor was considered separately, and we do not know whether there are quantitatively important interactions between the different factors (e.g., does the effect of carbohydrate intake differ at different levels of physical activity), as no study has evaluated such hypotheses.

Allowing for the limitations of inferences from observational studies, our work suggests that the lifestyle factors that we have evaluated are probably those that patients have the most control over. However, these should not be the only ones that public health authorities target. For example, other potentially modifiable hazards known to increase the risk of kidney damage include environmental pollution,^149,150 excessive body weight related to nutrition,¹²⁹ inappropriate use of medications (including analgesics, antibiotics, antiretrovirals, or proton-pump inhibitors),¹⁵¹ and heavy metal exposure and intake (in particular, cadmium).¹⁵² Finally, society has a further role to play in facilitating more personal choices through food labeling, alcohol and tobacco labeling, health policy around exercise, policy around the ease of walking and cycling in the built environment, and walkability.

We recognize that these findings are on the basis of nonrandomized data with risk of residual and unknown confounding, particularly in the setting of studies with low to very low certainty of the evidence. However, the generalizability and overall importance of these findings should not be undermined, as they are well in line with public health recommendations for preventing related chronic conditions and general healthy lifestyle principles. Therefore, the results can represent important indications for public policy and advocacy and be hypothesis generating for future randomized trials. It is not trivial to ask patients to change their lifestyle. Diet, in particular, is a highly social and cultural issue; furthermore, many patients are constrained by external factors, including economic factors, in their decisions. For these reasons, we believe that lifestyle intervention studies to address unanswered questions examining the whole range of clinically important outcomes, including the incidence and progression of CKD, continue to be justified.

Increased vegetable and potassium intake, physical activity, and moderate alcohol consumption were consistently associated with reduced risk of CKD. Increased dietary salt intake and tobacco smoking were consistently associated with increased risk of CKD. In the absence of trial evidence, these findings can inform both public health recommendations and patient-centered discussions in clinical practice.

Disclosures

K. Campbell has received consultation, advisory board membership, or research funding from the Dietitians Association of Australia, Nestle Health Sciences, and Queensland Health, all outside the submitted work. J.J. Carrero reports grant funding from Astellas, AstraZeneca, Vetenskapsrådet (2019-01059), and ViforPharma; consulting for AstraZeneca and Baxter; speaker fees for Abbott, AstraZeneca, Fresenius, Nutricia, and ViforPharma; and support from the Swedish Heart and Lung Foundation, all outside the submitted work. C.M. Clase has received consultation, advisory board membership, or research funding from Amgen, Astellas, AstraZeneca, Baxter, Boehringer-Ingelheim, Fresenius Medical Care, Janssen, Johnson & Johnson, Leo Pharma, the Ontario Ministry of Health, Pfizer, Relyspa, Sanofi, and Vifor, all outside the submitted work. J.T. Kelly has received consultancy fees from Amgen for travel and lecture presentations and consultancy fees from HealthCert for topics unrelated to the submitted work. J.T. Kelly is supported through a Griffith University Postdoctoral Research Fellowship, outside the submitted work. A. González-Ortiz is supported by School of Medicine, Programa de Doctorado en Ciencias Médicas, Odontológicas y de la Salud, Consejo Nacional de Ciencia y Tecnología grant CVU 373297, outside the submitted work. G. Su is supported by the European Renal Association-European Dialysis and Transplantation Association Young Fellowship Programme and by Guangdong Provincial Hospital of Traditional Chinese Medicine grant YN2018QL08, outside the submitted work. H. Xu is supported by Loo and Hans Osterman’s Foundation, outside the submitted work. All remaining authors have nothing to disclose.

Funding

None.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020030384/-/DCSupplemental.

Supplemental Figure 1. Study flow diagram.

Supplemental Figure 2. Summary of the associations between modifiable lifestyle risk and protective factors and risk of incident CKD.

Supplemental Figure 3. Dose-response relationship between alcohol intake (grams per day) and incident CKD estimated with a random effect metaregression-restricted cubic spline model.

Supplemental Figure 4. Association of tobacco smoking and ESKD.

Supplemental Figure 5. Summary of the associations between modifiable lifestyle risk and protective factors and risk of GFR decline.

Supplemental Figure 6. Summary of the associations between modifiable lifestyle risk and protective factors and risk of albuminuria.

Supplemental Figure 7. Consistency of associations in lifestyle factors that could not be statistically pooled across the markers of kidney function decline.

Supplemental Figure 8. Risk of bias across the included studies.

Supplemental Figure 9. Individual assessment of risk of bias across the included studies.

Supplemental Figure 10. Funnel plots for lifestyle hazards and incident CKD.

Supplemental Figure 11. Funnel plots for lifestyle hazards and secondary outcomes.

Supplemental Material. Data synthesis for secondary outcomes.

Supplemental Table 1. Search terms used across the electronic databases.

Supplemental Table 2. Summary of the methods relating to data extraction, risk of bias, metaregression, and planned subgroup and sensitivity analyses.

Supplemental Table 3. Characteristics of the included studies.

Supplemental Table 4. Lifestyle hazards and kidney disease outcomes from individual studies that could not be statistically pooled into meta-analysis.

Supplemental Table 5. Summary of the studies between alcohol intake and incident CKD.

Supplemental Table 6. Subgroup analysis for incident CKD.

Supplemental Table 7. Results from sensitivity analysis.

Supplemental Table 8. GRADE table summarizing the quality of the evidence for each outcome.