Systematic review

Cardiovascular diseases Metabolic diseases Environmental health
Volume 208 - Issue 2

Salt consumption by Australian adults: a systematic review and meta-analysis

Authors:  Mary-Anne Land, Bruce C Neal, Claire Johnson, Caryl A Nowson, Claire Margerison and Kristina S Petersen

Med J Aust 2018; 208 (2): 75-81. || doi: 10.5694/mja17.00394
Published online: 29 January 2018

Further action to reduce salt consumption in Australia and robust monitoring of population salt intake are needed

Download PDF (823 KB)
Share: Share on Blue Sky Share on Facebook Share on LinkedIn Share by email

Abstract

Objective: Salt reduction is a public health priority because it is a leading contributor to the global burden of disease. As in Australia there is uncertainty about the current level of salt intake, we sought to estimate current levels.

Study design: Random effects meta-analysis of data from 31 published studies and one unpublished dataset that reported salt or sodium consumption by Australian adults on the basis of 24-hour urine collections or dietary questionnaires.

Data sources: MEDLINE (via Ovid) and EMBASE (to August 2016).

Data synthesis: Thirty-one published studies and one unpublished dataset (1989–2015; 16 836 individuals) were identified. The mean weighted salt consumption estimated from 24-hour urine collections was 8.70 g/day (95% CI, 8.39–9.02 g/day); after adjusting for non-urinary salt excretion, the best estimate of salt intake in Australia is 9.6 g/day. The mean weighted intake was 10.1 g/day (95% CI, 9.68–10.5 g/day) for men and 7.34 g/day (95% CI, 6.98–7.70 g/day) for women. Mean weighted consumption was 6.49 g/day (95% CI, 5.94–7.03 g/day) when measured with diet diaries, 6.76 g/day (95% CI, 5.48–8.05 g/day) when assessed with food frequency questionnaires, and 6.73 g/day (95% CI, 6.34–7.11) when assessed by dietary recall. Salt intake had not decreased between 1989 and 2015 (R2 = –0.02; P = 0.36).

Conclusion: Salt intake in Australian adults exceeds the WHO-recommended maximum of 5 g/day and does not appear to be declining. Measuring salt intake with methods based on self-reporting can substantially underestimate consumption. The data highlight the need for ongoing action to reduce salt consumption in Australia and robust monitoring of population salt intake.

Despite a steady improvement in mortality rates over the past three decades, cardiovascular disease remains the leading cause of death in Australia1 and elsewhere.2 Strong and consistent evidence indicates that a high dietary salt intake increases blood pressure, which in turn increases the risk of cardiovascular disease.3 Some observational studies have found that lower salt intake may increase all-cause mortality, but methodological flaws, including critical limitations in the measurement of salt intake, mean that their results should be interpreted with caution.4 The World Health Organization recommends a maximum salt intake of 5 g/day. However, salt intake in most countries far exceeds this level, and the global average is estimated to be 10 g/day.5 As reducing population salt intake is projected to be one of the most cost-effective strategies for reducing rates of premature death and disability attributable to high blood pressure and vascular disease,6 all WHO member states have agreed to the target of reducing the mean population salt intake by 30% by 2025.7

Surveillance is a key component of salt reduction strategies. A recent review of salt reduction in Australia found that a number of relevant activities are being implemented, but that further efforts and more robust national monitoring are required.8 To date, the only nationally representative survey of Australian adults found the average salt consumption to be 6.1 g/day in 2011–12.9 This survey was based on 24-hour dietary recall; as acknowledged by the authors of the survey, the Australian Bureau of Statistics, this approach underestimates salt intake because of incomplete reporting, a well described limitation of this method.9,10 Other methods for estimating salt intake include food frequency questionnaires (FFQs), multiple dietary recalls, and urinalysis, including spot, timed and 24-hour assessments. Measuring 24-hour urinary sodium excretion is considered the gold standard, as about 90% of ingested salt is excreted in the urine.11 However, 24-hour urine collections are burdensome for participants, and response rates are therefore typically low.12 In addition, incomplete sample collection and cost can affect the usefulness of this method; as a consequence, 24-hour urine collection is not often undertaken in large population studies.13

A number of programs for reducing dietary salt intake have been implemented in Australia, but there has been no robust assessment of national intake levels that would permit the success or failure of these programs to be quantified. To reliably estimate mean population salt consumption by Australian adults, we therefore undertook a systematic review and meta-analysis of the available data.

Methods

Search strategy

A systematic search of peer-reviewed publications identified articles that reported salt or sodium consumption by Australian adults (at least 18 years old) on the basis of 24-hour urine collections or dietary questionnaires. Relevant articles published before August 2016 were identified by searching MEDLINE via Ovid and EMBASE, combining the search words “sodium” or “salt”, “Australia”, and “intake”, “ingest*”, “eat*”, “consum*”, “diet*”, “urin*” or “excret*”. Reference lists from identified articles were scanned to identify further relevant articles. Australian dietary experts were also consulted to identify research not found by our literature search.

Inclusion and exclusion criteria

Included studies reported mean or median salt or sodium consumption by Australian adults. For intervention studies, we included only the baseline intake data. If both urinary analysis and dietary methods were reported for a study, all results were included in the review. Articles were not excluded because of language, age, sex, study design (including representativeness of the sample), or sample size.

Data extraction

Data for the following study characteristics were extracted from each article: population type, salt intake assessment method, sample size, sex, mean age, mean body mass index (BMI), study location, and mean salt intake (with standard deviation [SD]). If sodium intake (mg/day) was reported, it was converted to salt (g/day) by dividing by 1000 and multiplying by 2.54 (the ratio of the molecular weight of sodium chloride [58.44] to the atomic weight of sodium [23.0]). Titles and abstracts of retrieved articles were independently evaluated by two investigators (MAL, CJ), who were not blinded to authors, institutions, or manuscript journals. If an abstract did not provide adequate information about inclusion and exclusion criteria, the full text was independently evaluated to determine study eligibility. Disagreements were resolved by consensus and, if disagreement persisted, by a third reviewer (KSP). To avoid possible double-counting of patients included in more than one report by the same authors or working groups, patient characteristics and study recruitment dates were evaluated; if necessary, authors were contacted for clarification. Corresponding authors were also contacted to obtain data not included in their published report.

Outcomes

The primary analyses were of salt intake estimated on the basis of 24-hour urine collections. Subgroup analyses assessed differences in salt intake by sex and population type. In addition, the associations between salt intake and age, publication date, and BMI were investigated. Secondary analyses determined mean salt consumption on the basis of estimates derived from dietary assessments.

Data analysis

In the primary and secondary analyses, mean salt intake (with 95% confidence interval [CI]) was estimated using inverse variance random effects meta-analysis. If a study reported salt intake estimates based on 24-hour urine collection as well as dietary methods, the results of 24-hour urine collections were included in the primary analyses, and data for both methods were included in the subgroup analysis of measurement method. For the subgroup analysis by sex, studies that reported salt intake data by sex were pooled to provide salt intake estimates for males and females separately. Studies that did not provide salt intake data by sex were pooled and reported as mixed sex estimates. The proportion of variability attributable to heterogeneity between studies rather than sampling error was expressed as the I2 statistic. The t2 value quantified between-study variance and was used to weight studies in the random effects model. To explore sources of heterogeneity, subgroup analyses of method of salt intake measurement, sex, and health status were undertaken. Random effects meta-regression was performed to determine the relationship between salt intake and mean BMI, mean cohort age, and year of publication. Sensitivity analyses determined whether removal of an individual study altered the estimated mean intake. Statistical analysis was completed in Stata 13 (StataCorp); P < 0.05 was deemed statistically significant.

Results

Data from 31 published studies (including the ABS survey) and one unpublished set of data were included in the meta-analysis (Box 1, Box 2). Salt intake estimates were based on 24-hour urine collections in 26 studies (including 12 that also used dietary questionnaires or diaries), on FFQs in six studies, dietary recall in six studies, and on dietary diaries in eight. The 32 studies were conducted between 1989 and 2015, and included random or volunteer general population samples, or populations with specific disease states, such as hypertension, metabolic syndrome, and types 1 and 2 diabetes.

The 24-hour urine collection datasets included salt intake data for 3896 individuals. The mean weighted salt intake was 8.70 g/day (95% CI, 8.39–9.02 g/day; t2 = 0.55; I2 = 91% [P < 0.001]) (Box 3). Inflation of salt intake by 10% to account for non-urinary sodium excretion45 provides a best estimate of daily salt intake of 9.6 g/day. Removing each study individually did not substantially alter the estimate. The mean weighted intake for men (13 studies16,17,19,25,26,31,33,35,36,38,39,42,43, 1381 men) was 10.1 g/day (95% CI, 9.68–10.5 g/day; t2 = 0.32; I2 = 72% [P < 0.001]); for women (16 studies16,17,25,26,29-33,35,36,38,39,41-43, 1708 women) it was 7.34 g/day (95% CI, 6.98–7.70 g/day; t2 = 0.40; I2 = 82% [P < 0.001]) (Box 4).

For random or volunteer populations (15 studies15-17,19,21,25,26,28,30,32,33,35,39,41,42, 2956 individuals), the weighted mean salt intake was 8.36 g/day (95% CI, 7.95–8.77 g/day; t2 = 0.58; I2 = 89% [P < 0.001]). For people with hypertension (six studies14,18,22-24,29, 486 individuals) it was 8.82 g/day (95% CI, 7.88–9.77 g/day; t2 = 1.11; I2 = 92% [P < 0.001]), for those with type 1 or 2 diabetes (four studies31,36,38,43, 355 individuals) it was 9.66 g/day (95% CI, 9.07–10.3 g/day; t2 = 0.20; I2 = 54% [P = 0.09]), and for people with metabolic syndrome (one study27, 99 individuals) it was 10.4 g/day (95% CI, 8.96–11.8 g/day; t2 = 0.64; I2 = 56.5% [P = 0.13]) (Box 4).

The secondary analyses indicated that salt intake was 6.49 g/day (95% CI, 5.94–7.03 g/day; t2 = 0.55; I2 = 87% [P < 0.001]) when intake was measured with a diet diary (eight studies18-20,30,32,36,41,43, 503 individuals), 6.76 g/day (95% CI, 5.48–8.05g/day; t2 = 2.5; I2 = 97% [P < 0.001]) when measured with an FFQ (six studies20,30,34,40,43,44, 504 individuals), and 6.73 g/day (95% CI, 6.34-7.11 g/day; t2 = 0.21; I2 = 86% [P < 0.001]) when 24-hour recall was used (six studies25,33,37-39,42, 13 428 individuals) (Box 4).

Meta-regression found that salt intake estimated from 24-hour urine collections was significantly correlated with BMI (R2 = 0.35; P = 0.008), but not with mean age of the sample cohort (R2 = –0.02; P = 0.36) or with year of publication (R2 = –0.02; P = 0.36), indicating that population salt intake has not decreased across the study period (online Appendix).

Discussion

We found that the average salt consumption by Australian adults is well above the recommended maximum of 5 g/day. While estimates of salt intake varied according to the method for measuring consumption, all exceeded the WHO-recommended maximum. Our best estimate of 9.6 g salt per day, based on 24-hour urine collections and corrected for non-urinary sodium loss, is almost double the recommended limit. The mean weighted intake was 10.1 g/day for men and 7.3 g/day for women; this sex difference was anticipated, given the generally lower energy intake among women. These data reinforce the need for salt reduction strategies in Australia.

Reducing population salt intake has been on the health agenda in Australia for many years, but the impact of the strategies employed has been unclear because intake has not been monitored at the population level.8 The 2011–12 Australian Health Survey has been the only attempt to assess salt intake in a large representative sample of the population, albeit on the basis of 24-hour dietary recall. In our meta-analysis, we found no relationship between publication date and salt intake level, suggesting there has been no change in consumption since 1989. However, our analysis of change over time is limited by the data available, which are not based on repeat representative samples of the population; it is therefore possible that temporal changes in salt intake have been missed.

Although there was significant variation in the estimates of salt intake by studies that based their assessments on 24-hour urine collections, even the lower end of the prediction interval (mean ± 2 × SD)46 — 7.2–12.9 g/day — exceeds the WHO-recommended maximum. Salt intake in Australia is therefore almost certainly too high and needs to be reduced. The heterogeneity of study results was not unexpected, given the differences in sample characteristics, methods used to measure salt intake, and study designs. Salt intake varies significantly with characteristics such as age, BMI, and disease state,12 and meta-analyses of aggregated data cannot effectively control for such factors. Our meta-regression analysis found a relationship between salt intake and BMI, consistent with the recognised relationship between energy intake and salt intake, but not between salt intake and age.

Estimates of salt intake on the basis of assessment methods other than 24-hour urine collections were substantially lower: estimates based on diet diaries (2.2 g/day lower), FFQs (1.9 g/day lower), and dietary recall (2.0 g/day lower) were all below the estimate based on urinalysis. This is consistent with research that has documented under-reporting of total dietary intake, and consequently of salt intake, in dietary surveys.10 The accuracy of the assumed average values of salt content of processed and home-cooked foods and discretionary salt use also pose challenges that are difficult to overcome. The results of our meta-analysis suggest that when dietary methods are the primary basis for estimating salt intake in a study, 24-hour urine samples of a subsample of participants should be assessed to enable calibration of results. Further, our results highlight the need for careful interpretation of results gained with different methods for assessing salt intake.

A 2015 review found that most salt reduction campaigns in Australia have been led by non-government organisations, and that there is a need for stronger government intervention.8 One of the key federal government initiatives was the Food and Health Dialogue, introduced in 2009, which included voluntary salt reduction targets for several food categories. These targets resulted in minor reductions in the sodium content of breads (reduced by 39 mg/100 g, or 9%), ready-to-eat breakfast cereals (by 79 mg/100 g, or 25%) and processed meats (by 101 mg/100 g, or 8%).47 The Food and Health Dialogue has now been replaced by the Healthy Food Partnership, and it is unclear what the objectives of this initiative will be. However, the results of our meta-analysis indicate that salt reduction should remain a priority.

Modelling suggests that if sodium intake in Australia and New Zealand were reduced by 10% over 10 years by a national program, about 11 000 disability-adjusted life-years linked to cardiovascular disease could be averted each year.48 One of the most successful salt reduction strategies was implemented in the United Kingdom. Key components included determining and monitoring population salt intake, setting population salt intake targets, reducing the salt content of the food supply by setting progressively lower targets for various food categories, and moving industry to reformulate food products so that they contained less salt.47 In response to this action, salt intake in the UK decreased from 9.5 g/day in 2001 to 8.6 g/day in 2008, and to 8.1 g/day in 2011, an overall 15% reduction.49 It has been estimated that this campaign cost £15 million to implement, but there were 9000 fewer cardiovascular deaths and £1.5 billion saved each year as a result of the strategy.50

In Australia, a community-based intervention in a rural population showed that reducing population salt intake was possible (by 0.8 g/day between 2011 and 2014) with a multi-faceted approach that included education, community engagement, a smartphone application, provision of a salt substitute, and advertising.51 However, it is unlikely that a program of this type could be implemented nationally, and alternative approaches will be required.

Our meta-analysis is the first step to estimating population salt consumption in Australia by integrating all the relevant data that have been collected. While it was a novel investigation, our analyses were limited by the significant heterogeneity of the studies, most of which had small sample sizes. In addition, some studies included individuals with hypertension, diabetes or metabolic syndrome, and it is possible that some individuals changed their salt consumption after being diagnosed with such conditions, although estimates of salt intake were comparable across the subgroups included. In addition, most studies applied criteria for excluding incomplete 24-hour urine collections or implausible dietary intake data, and these criteria differed between studies.

Conclusion

Our systematic review and meta-analysis found that salt consumption by Australian adults probably exceeds the WHO-recommended maximum of 5 g/day, re-affirming the need for action to reduce salt intake in Australia. Our results also highlight the need for systematic, standardised and repeated assessments of a national sample of the population in order to determine whether salt reduction programs are achieving the target of a 30% reduction in intake by 2025.7

Box 1 – Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram for the selection of publications included in our analysis

FFQ = food frequency questionnaires. * More than one method was employed in some studies.

Box 2 – The investigations of the salt intake of Australian adults, 1989–2015, included in our meta-analysis

Study

State, year*

Method of assessment

Population

n

Men

Mean age (years)

Mean BMI (kg/m2)

Mean salt intake (SD), g/day

All

Men

Women

NHMRC DSSMC14

NSW, 1989

24 h urine

Diastolic blood pressure 90–100 mmHg

108

83%

58.4

—

8.1 (0.4)

—

—

Beard15

Tas, 1992

24 h urine

Random sample

54

41%

49

—

8.3 (4.6)†

—

—

Beard16

Tas, 1997

24 h urine

Random sample

194

45%

[range, 18–70]

—

8.4 (2.7)

9.9 (3.0)

6.9 (2.5)

Jones17

Tas, 1997

24 h urine

Random and volunteer sample

154

22%

46.9

—

—

10.7 (2.3)†

8.6 (3.2)†

Bao18

WA, 1998

24 h urine

Overweight non-smoking men and postmenopausal women, taking antihypertensive medication (= 3 months)

63

67%

51.1

31.6

10.5 (8.0)†

—

—

3-day diet diary

7.7 (6.3)†

—

—

Mori19

WA, 1999

24 h urine

Mildly hyperlipidaemic volunteers

56

100%

48.8

28.8

10.2 (4.2)†

10.2 (4.2)†

—

3-day diet diary

8.6 (0.6)†

8.6 (0.6)†

—

Hodge20

Vic, 2000

7-day diet diary

Iron deficient volunteers

63

0

33.3

—

5.9 (1.4)

 

5.9 (1.4)

FFQ

5.4 (1.6)

 

5.4 (1.6)

Nowson21

Vic, 2003

24 h urine

Twins (normotensive and hypertensive)

108

41%

47

26.1

8.1 (3.1)

—

—

Nowson22

Vic, 2004

24 h urine

Treated hypertensive and normotensive volunteers

94

60%

55.6

29.0

8.8 (2.9)

—

—

Ward23

WA, 2004

24 h urine

Untreated hypertensive and normotensive volunteers

66

56%

54.1

26.7

11.3 (7.5)†

—

—

Hodgson24

WA, 2006

24 h urine

Hypertensive volunteers

60

63%

58.5

27.7

9.7 (3.9)

—

—

Margerison25

Vic, 2006

24 h urine

Healthy volunteers

144

55%

55.2

29.2

8.6 (3.2)

9.9 (2.8)

7.0 (2.9)

24 h dietary recall

8.0 (3.3)

6.2 (0.9)

Brinkworth26

SA, 2008

24 h urine

Overweight and obese volunteers

208

33%

52.4

33.6

8.6 (4.5)†

10.6 (5.5)†

7.9 (3.6)†

Keogh27

SA, 2008

24 h urine

Overweight and obese volunteers with at least one additional risk factor for metabolic syndrome

99

—

50

33.7

11.3 (5.8)
9.8 (3.7)

—

—

Dickinson28

SA, 2009

24 h urine

Overweight and obese volunteers

29

24%

52.7

31.6

9.0 (4.0)

—

—

Nowson29

Vic, 2009

24 h urine

Post-menopausal women with normal to high blood pressure

95

0

59.2

29.6

6.3 (3.4)†

—

6.3 (3.4)†

Lassale30

SA, 2009

24 h urine

Healthy volunteers

62

0

49.2

27.6

7.4 (2.5)

—

7.4 (2.5)

4-day diet diary

6.3 (1.7)

—

6.0 (1.7)

FFQ

7.6 (3.0)

—

7.6 (3.0)

Ekinci31

Vic, 2010

24 h urine

Patients with type 2 diabetes attending an outpatient clinic

122

52%

68.6

30.4

10.1 (3.6)

9.9 (3.1)

8.3 (3.0)

Charlton32

NSW, 2010

24 h urine

Healthy volunteers

76

0

38.3

24.0

6.5 (2.7)

—

6.5 (2.7)

3-day diet diary

6.9 (4.4)

—

6.9 (4.4)

Ireland33

SA, 2010

24 h urine

Healthy volunteers‡

43

23%

56

—

7.3 (3.0)
7.9 (2.6)

7.2 (3.8)
9.4 (1.8)

6.9 (3.0)
7.0 (2.4)

24-hour dietary recall

—

8.7 (3.9)
7.2 (2.8)

7.0 (4.2)
7.8 (3.8)

Keogh34

SA, 2010

FFQ 1

Healthy volunteers

159

100%

55

27.4

9.4 (3.4)†

9.4 (3.4)†

—

FFQ 2

7.8 (3.2)†

7.8 (3.2)†

—

Huggins35

Vic, 2011

24 h urine

Random sample

783

48%

64

28.4

9.1 (3.7)†

10.4 (3.8)†

7.8 (2.9)†

Villani36

SA, 2012

24 h urine

Overweight and obese volunteers with type 2 diabetes

88

59%

Men, 61.9
Women, 59.4

Men, 34.5
Women, 35.9

9.8 (3.3)

11.4 (4.4)

8.4 (2.4)

4-day diet diary

—

7.2 (1.6)

5.8 (1.3)

ABS [Australian Health Survey]37

All, 2011–2012

24-hour dietary recall

Random sample

12 153

47%

40

25.7

6.0 (3.4)

6.8 (3.8)

5.2 (2.9)

Petersen38

SA, 2013

24 h urine

Volunteers with type 2 diabetes

78

63%

63

34.1

9.9 (4.6)

11.0 (4.4)

7.4 (3.6)

24-hour dietary recall

7.0 (3.2)

7.3 (2.7)

6.3 (2.7)

Land39

NSW, 2014

24 h urine

Random and volunteer sampling

419

45%

55

29.3

8.8 (3.6)

10.2 (3.7)

7.6 (3.0)

Land (unpublished data)

 

24-hour dietary recall

 

412

44%

 

 

6.5 (2.9)

7.5 (3.1)

5.6 (2.4)

Blanch40

SA, 2014

FFQ

Healthy volunteers

35

26%

31

21.7

5.6 (2.1)

—

—

Turner41

SA, 2015

24 h urine

Volunteers with coeliac disease

28

0

51

25

6.3 (3.0)

—

6.3 (3.0)

3-day diet diary

5.0 (1.7)

—

5.0 (1.7)

Nowson42

Vic, 2015

24 h urine

Random sample

598

46%

57

26.6

8.1 (3.1)

9.2 (2.6)

6.7 (2.7)

24-hour dietary recall

6.8 (3.1)†

6.1 (3.2)†

7.7 (3.3)†

Petersen43

SA, 2015

24 h urine

Volunteers with type 1 and type 2 diabetes

67

57%

56

31

8.6 (4.1)

9.0 (4.0)

8.1 (4.2)

3-day diet diary

6.3 (2.9)

7.0 (3.2)

5.3 (2.2)

FFQ

6.2 (2.5)

6.4 (2.3)

6.0 (2.8)

Petersen44

SA, 2015

FFQ

Volunteers with type 1 and type 2 diabetes

118

63%

57.5

32.6

6.4 (2.9)

7.2 (3.1)

5.1 (1.7)

ABS = Australian Bureau of Statistics. FFQ = food frequency questionnaire; NHMRC DSSMC = Australian National Health and Medical Research Council Dietary Salt Study Management Committee; — = not reported. * Year of publication. † Standard deviations calculated for meta-analysis. ‡ Two parallel groups of participants were followed in this study; the salt intake of each was reported separately.

Box 3 – Random effects meta-analysis of overall salt intake, measured by 24-hour urine collection, Australia, 1989–2015

Box 4 – Summary of random effects meta-analysis of salt intake, Australia, 1989–2015, by measurement method, population, and sex

Authors

  • Mary-Anne Land1
  • Bruce C Neal1
  • Claire Johnson1
  • Caryl A Nowson2
  • Claire Margerison2
  • Kristina S Petersen1
  • 1. The George Institute for Global Health, Sydney, NSW
  • 2. Institute for Physical Activity and Nutrition, Deakin University, Geelong, VIC

Correspondence: maland@georgeinstitute.org.au

Competing interests

No relevant disclosures.

Acknowledgements

Bruce Neal is supported by a National Health and Medical Research Council (NHMRC) Principal Research Fellowship. He holds an NHMRC Centre for Research Excellence grant (APP1117300) and an NHMRC program grant (APP1052555).

References

Linked content

  • MJA InSight: Salt intake reduction: life-saving and cost-effective

  • MJA Podcast: Professor Bruce Neal

Provenance: Not commissioned; externally peer reviewed.

View this article on Wiley Online Library

More like this

View more