Menopause age and type and dementia risk: a pooled analysis of 233 802 women

Abstract

Objectives

It is not clear whether the association between younger age at menopause and increased risk of dementia is modified by type of menopause. We examined the association of age at menopause or hysterectomy with dementia risk in three groups of women: those with natural menopause, premenopausal bilateral oophorectomy (surgical menopause) or premenopausal hysterectomy (without bilateral oophorectomy).

Study design

Individual-level data from 233 802 women in five prospective cohort studies (from four countries) were harmonized and pooled. Cox proportional hazards models were used to assess the associations of age at natural menopause, surgical menopause or premenopausal hysterectomy, with age at dementia, death (where available) or end of follow-up, whichever came first.

Results

The study followed women to the median age of 72 years (quartiles 67, 76 years). The median follow-up time was 13 years, with 3262 dementia cases during this period. Compared with women with menopause at 50–52 years, women with menopause <40 years had a higher risk of dementia (adjusted hazard ratio (aHR): 1.47, 95% confidence interval (CI): 1.39, 1.56). This level of risk was comparable to that of current smoking and stroke, which are well-established risk factors for dementia. Increased risk of dementia associated with surgical menopause or premenopausal hysterectomy (compared to natural menopause) was not apparent after adjustment for age at menopause (aHR 0.99, 95% CI: 0.93, 1.04 and aHR 0.97, 95% CI: 0.95, 1.00, respectively).

Conclusion

Women who experience menopause before the age of 40 years have a higher risk of dementia irrespective of type of menopause.

Key Points

• In a pooled analysis of individual data from 223 802 women, risk of dementia increased with earlier age at menopause.

• The highest risk was in women with menopause <40 years, equal to the risk seen in current smokers and those who had a stroke.

• Type of menopause (natural, surgical, premenopausal hysterectomy) was not associated with dementia risk in adjusted models.

• Our results support the hypothesis that greater exposure to endogenous oestrogen confers neuroprotective effects.

Introduction

Dementia is a leading cause of death globally, and poses a heavy burden on patients, their caregivers and healthcare systems [1]. It was estimated that over 50 million people lived with dementia globally in 2018, and this figure is projected to triple by 2050 [2]. There is currently no effective cure for dementia, and it is important to identify modifiable risk factors for this disease in order to develop tailored preventive strategies. A Lancet Commission Report identified 14 modifiable risk factors for dementia (e.g. smoking) and suggested that avoiding these risk factors might delay or prevent 45% of dementia cases [3].

There is a higher prevalence of dementia in older women compared with men of the same age [1], and female sex hormones appear to be neuroprotective [4], indicating that age at menopause may be associated with risk of dementia. In addition, higher dementia risk may be associated with surgical menopause (i.e. premenopausal bilateral oophorectomy) which results in an abrupt decline in hormone levels, or premenopausal hysterectomy, where removal of the uterus, even with ovarian conservation, may reduce menopause age by two to four years [5]. The Lancet Commission Report considered the role of menopause and menopausal hormone therapy but could not reach a conclusion as to whether these factors were causally associated with dementia risk [3]. A number of studies from countries including the United Kingdom, France, Sweden, the Netherlands, the United States and Korea have looked at age at menopause and the risk of dementia [6–14], with most [6–8, 10, 14], but not all [9, 12, 13], suggesting an increased dementia risk with younger age at menopause. Across these prior studies, menopause is defined in a number of different ways: natural menopause only [7, 12–14]; natural or surgical menopause (excluding hysterectomy) [6, 10]; natural menopause, surgical menopause, or premenopausal hysterectomy [8, 9]. However, when women with surgical menopause or premenopausal hysterectomy were included in the definition of menopause, the analytic models did not account for type of menopause. Therefore, it remains unclear whether the association between age at menopause and dementia is modified by type of menopause.

The present study used individual participant data from five cohort studies which contributed to the International collaboration for a Life course Approach to reproductive health and Chronic disease Events (InterLACE) consortium. The aim was to examine the association of age at menopause or hysterectomy with dementia risk in three groups of women: those with natural menopause, premenopausal bilateral oophorectomy or premenopausal hysterectomy (without bilateral oophorectomy).

Materials and methods

Study design

InterLACE is an international consortium that aims to understand the associations between women’s reproductive health and chronic diseases and includes 27 observational studies. InterLACE has individual-level data on reproductive, sociodemographic and lifestyle factors, as well as disease outcomes and mortality. Data harmonization was conducted after studies joined InterLACE, and detailed information on InterLACE has been published previously [15, 16]. The present study analysed data from five cohort studies included in InterLACE which had complete information on natural menopause, surgical menopause, premenopausal hysterectomy, age at natural or surgical menopause, age at hysterectomy and dementia. The included studies were the Australian Longitudinal Study on Women’s Health for women born in 1946–51 (ALSWH) [17], Swedish Women’s Lifestyle and Health Study (WLHS) [18], English Longitudinal Study of Ageing (ELSA) [19], UK Biobank [20] and Prospect-EPIC Utrecht [21].

Sample

A total of 250 997 postmenopausal women (either natural, bilateral oophorectomy or premenopausal hysterectomy) were eligible for inclusion. The analytical sample consisted of 233 802 women after the exclusion of 17 195 women for the following reasons: having a record or report of dementia before menopause or hysterectomy (n = 1773); reported age at natural or surgical menopause or hysterectomy was less than 30 years or more than 60 years (n = 2903); missing information on hysterectomy or oophorectomy (n = 11 190) or dementia (n = 1329).

Exposures

The exposure variables were natural menopause, premenopausal bilateral oophorectomy, premenopausal hysterectomy without bilateral oophorectomy and age at natural menopause or the two types of surgery. Natural menopause was defined as spontaneous cessation of menstruation for 12 months without prior hysterectomy or oophorectomy. Premenopausal bilateral oophorectomy referred to the surgical removal of both ovaries before natural menopause (with the term ‘surgical menopause’ used hereinafter). Premenopausal hysterectomy was defined as removal of the uterus before natural menopause without bilateral oophorectomy (hereinafter called ‘premenopausal hysterectomy’ for simplicity). Women with hysterectomy or bilateral oophorectomy after natural menopause were included in the ‘natural menopause’ group in this study.

In women with natural menopause, the age when they met the definition of natural menopause was used as their age at menopause. In women with surgical menopause, age at bilateral oophorectomy was used as their age at menopause. In women in the premenopausal hysterectomy category, we used their age at hysterectomy as their age at menopause, while recognising that hormonal decline for these women may be delayed [5]. Age at natural menopause or surgery was self-reported and categorized into five groups: <40, 40–44, 45–49, 50–52 and ≥ 53 years.

Outcomes

The outcome variable was age at first record of dementia or death (where data were available) or the end of the study period. Dementia was ascertained from self-reports of doctor diagnosis in ELSA, from hospital data in WLHS and from self-reports and linked administrative records (hospital data and death registry) in ALSWH, Prospect-EPIC Utrecht and UK Biobank. For instance, in ELSA, the dementia question was the following: ‘Has a doctor ever told you that you have (or have had) dementia?’. For linked data, the International Classification of Disease versions 9 and 10 (ICD 9 and 10) were used to identify dementia cases, and the specific ICD codes used were: 290.0, 290.1, 290.2, 290.4, 290.8, 290.9, 331.0 and 331.1 (ICD 9), and F00, F01, F03, G30 and G31.0 (ICD 10).

Confounders

The following variables were considered as possible confounders as they have been reported to be associated with both age at menopause and dementia: year of birth [22, 23], race/ethnicity [22, 24], education [3, 22], parity [25, 26], age at menarche (the first menstrual period in a female) [25, 26], body mass index (BMI) [3, 27], smoking status [3, 28] and hypertension, diabetes or stroke before dementia [3, 29]. Year of birth was categorized into <1940, 1940–1949, 1950–1959 and ≥ 1960. Race/ethnicity was categorized into four groups: Caucasian, Asian, Black and others. Education was categorized into three groups: ≤10, 11–12 and ≥ 12 years of schooling. Age at menarche was categorized into ≤11, 12, 13, 14 and ≥ 15 years. Parity was categorized into 0, 1, 2 and ≥ 3 children. BMI was categorized into four groups according to the World Health Organization criteria: <18.5, 18.5–24.9, 25.0–29.9 and ≥ 30 kg/m². Smoking status was categorized into three groups: never smokers, former smokers and current smokers. Information on diabetes and stroke was collected from self-reports in ELSA and from self-reports and linked data in ALSWH, WLHS, Prospect-EPIC Utrecht and UK Biobank. Data on hypertension was self-reported in ELSA and Prospect-EPIC and was collected from self-reports and linked data in ALSWH, WLHS and UK Biobank. The hypertension, diabetes and stroke variables were binary (yes or no). As each cohort had information on the age when hypertension, diabetes and stroke were first recorded, we were able to identify women with each of these conditions before the first record of dementia. Aside from hypertension, diabetes and stroke, information on all other covariates was from baseline surveys. Missing data on included confounders were less than 1% for all variables except age at menarche with 2.3% missing (see the footnotes to Appendix 1 in the Supplementary Data section for details of missing data for each confounder).

Ethical approval

Ethical approval was granted by the Institutional Review Board or Human Research Ethics Committee at each participating institution. All participants provided written informed consent.

Statistical analysis

Cox proportional hazards models were used to assess the associations of age at natural menopause, surgical menopause or premenopausal hysterectomy with time (in years) from birth to dementia, death (where available) or end of the follow up, whichever came first. To avoid immortal time bias [30], age at natural menopause, surgical menopause and premenopausal hysterectomy was treated as a time-dependent variable, with the time from birth to natural menopause, surgical menopause or premenopausal hysterectomy treated as an unexposed period. The proportional hazards assumption was examined using Kaplan–Meier curves, and correlations between the Schoenfeld residuals and time for each covariate and globally. If the study identification was included as a fixed effect there was strong evidence that the proportional hazards assumption was violated, so study was treated as a stratifying variable. Robust variance–covariance estimates were used to account for within-study clustering. Ethnicity was not included in the analyses because more than 95% of participants in all studies were Caucasian. Three categories of BMI were used because very few women in any study had BMI <18.5 kg/m². Year of birth was not included in the stratified analyses because the large differences between studies were accounted for when study was used as a stratifying variable. The possibility that the association between age at menopause and dementia might vary by menopause type was assessed by fitting an interaction term between age at menopause and type of menopause.

Separate models were also fitted for each study and combined using random effects meta-analyses for menopause type and age categories.

All analyses were conducted using Stata/SE 18.0 for Windows [31].

Results

Population characteristics

There were 233 802 women included in the analysis, of whom 3262 (1.4%) developed dementia. Details of each study are shown in Table 1. Overall, 74.2% experienced natural menopause, 11.3% experienced surgical menopause and 14.5% had premenopausal hysterectomy; 7.8% of all women experienced menopause before 40 years of age. The characteristics of the women by type of menopause, are shown in Table 2. Compared to those with natural menopause, women with surgical menopause or premenopausal hysterectomy had younger age at menopause. Women with surgical menopause or premenopausal hysterectomy were also more likely to have ≤10 years of schooling, higher body mass index, hypertension or stroke (Table 2).

Table 1.

Baseline characteristics of participants in each study.

	ALSWH	WHL	ELSA	UK Biobank	Prospect EPIC	Total
Country	Australia	Sweden	England	UK	Netherlands
Number of participants	9030	17 112	5514	189 250	12 896	233 802
Dementia number	175	38	120	2692	237	3262
Dementia, %	1.94	0.22	2.18	1.42	1.84	1.40
Baseline, year	1996	1991–1992	2002	2006–2010	1993–1997
Follow-up, year	2022	2012	2010–2011	2022	2010
Age at endpointa, years
Median (quartile1, quartile2)	73 (71, 74)	63 (59, 66)	66 (60, 75)	73 (68, 77)	72 (67, 77)	72 (67, 76)
Year of birth
< 1940	0.00	0.00	38.30	4.12	59.24	7.51
1940—< 1950	74.36	67.85	36.56	54.49	40.76	55.05
1950—< 1960	25.64	29.91	24.65	35.65	0.00	32.62
> = 1960	0.00	2.24	0.49	5.74	0.00	4.82
Ethnicity, %
Caucasian	96.52	100.00	97.42	95.93	100.00	96.51
Asian	2.03	0.00	1.03	1.84	0.00	1.59
Black	0.18	0.00	0.63	1.50	0.00	1.23
Other	1.27	0.00	0.91	0.73	0.00	0.66
Education, %
<=10 yrs	46.31	36.65	69.94	52.45	45.00	51.06
11–12 yrs	16.87	24.14	6.88	11.77	38.30	14.21
>12 yrs	36.82	39.21	23.18	35.78	16.70	34.73
Parity, %
0	6.57	10.24	13.78	16.62	12.70	15.49
1	8.34	14.04	15.78	12.66	7.65	12.40
2	40.15	44.65	39.74	45.32	37.40	44.51
> = 3	44.94	31.08	30.70	25.40	42.25	27.61
Menopause type, %
Natural	69.18	63.15	75.86	75.86	67.89	74.18
Surgical	10.83	15.82	10.72	10.72	13.82	11.30
Premenopausal hysterectomy	19.99	21.03	13.42	13.42	18.29	14.52
Age at menopause, % by years
<40	10.51	7.90	8.51	7.46	10.35	7.80
40–44	10.47	12.30	13.15	11.63	12.97	11.75
45–49	20.84	27.27	24.21	23.50	26.14	23.83
50–52	28.28	25.49	28.73	30.47	28.61	29.88
> = 53	29.90	27.05	25.41	26.93	21.92	26.74
Age at Menarche, % by years
<=11	18.71	13.26	20.92	20.71	9.73	19.49
12	21.87	23.47	16.16	19.02	20.80	19.48
13	28.28	29.42	23.24	24.03	23.33	24.52
14	16.89	21.42	20.88	19.65	22.55	19.87
> = 15	14.25	12.43	18.79	16.59	23.59	16.64
BMI (kg/m2), %
<18.5	1.48	1.44	0.80	0.72	0.64	0.80
18.5–24.9	51.25	66.98	32.81	37.27	45.46	40.29
25.0–29.9	28.90	24.66	37.93	38.05	39.74	36.84
> = 30	18.37	6.92	28.46	23.96	14.17	22.08
Smoking, %
Never	55.73	40.92	45.78	58.41	44.47	55.95
Former	28.93	37.25	35.18	33.28	34.49	33.52
Current	15.34	21.82	19.04	8.31	21.04	10.53
Hypertension before endpointa, %	57.33	27.34	46.35	39.22	54.37	40.05
Diabetes before endpointa, %	20.83	8.89	9.83	7.85	10.92	8.64
Stroke before endpointa, %	6.17	2.75	4.95	3.36	7.17	3.67

ABBREVIATIONS: ALSWH = Australian Longitudinal Study on Women’s Health (cohort of participants who were born in 1946–1951); WLHS = Swedish Women’s Lifestyle and Health cohort; ELSA = English longitudinal study of ageing; UK Biobank = ; Prospect-EPIC Utrecht = Prospect—European Prospective Investigation into Cancer and Nutrition (EPIC) Utrecht

^aEndpoint is dementia or censored (i.e. death from another cause, loss to follow-up or end of follow-up).

Table 2.

Baseline characteristics of women by menopause type, pooled over all five studies (N = 233 802).

	Natural menopause	Surgical menopause	Premenopausal hysterectomy
Number of participants	173 435	26 408	33 959
Dementia number	2316	382	564
Dementia %	1.34	1.45	1.66
Age at endpointa, years
Median (quartile1, quartile2)	72 (67, 76)	72 (66, 76)	72 (66, 76)
Study %
ALSWH	3.60	3.70	5.32
WHL	6.23	10.25	10.60
ELSA	2.34	2.50	2.36
UK Biobank	82.78	76.80	74.79
Prospect EPIC	5.05	6.75	6.95
Age at menopause or surgery, years, mean (SD)	50.48 (4.33)	46.73 (6.82)	42.37 (6.36)
Age at menopause or surgery, % by years
<40	1.64	15.29	33.39
40–44	6.96	18.68	30.82
45–49	23.26	29.43	22.40
50–52	36.49	16.54	6.48
> = 53	31.65	20.05	6.91
Education, % by years
<=10	49.25	55.29	57.05
11–12	14.01	14.48	15.00
>12	36.74	30.23	27.95
Parity, %
0	16.34	16.72	10.20
1	12.62	13.08	10.69
2	44.42	43.48	45.76
> = 3	26.62	26.72	33.35
Age at Menarche, % by years
<=11	18.50	22.38	22.32
12	19.31	19.97	19.98
13	24.98	23.09	23.28
14	20.39	18.37	18.35
> = 15	16.82	16.18	16.06
BMI ((kg/m2), %
<25.0	42.91	35.82	36.85
25.0–29.9	36.50	37.22	38.27
> = 30	20.59	26.97	25.88
Smoking, %
Never	56.51	53.92	54.69
Former	33.50	33.91	33.31
Current	9.99	12.17	12.00
Hypertension before endpointa, %	37.81	46.48	46.49
Diabetes before endpointa, %	7.72	11.60	11.07
Stroke before endpointa, %	3.42	4.37	4.42

ABBREVIATIONS: ALSWH = Australian Longitudinal Study on Women’s Health (cohort of participants who were born in 1946–1951); WLHS = Swedish Women’s Lifestyle and Health cohort; ELSA = English longitudinal study of ageing; UK Biobank; Prospect-EPIC Utrecht = Prospect—European Prospective Investigation into Cancer and Nutrition (EPIC) Utrecht

^aEndpoint is dementia or censored (i.e. death from another cause, loss to follow-up or end of follow-up).

Age at menopause and type of menopause and dementia

Early age at menopause was associated with higher risk of dementia. Compared to women who experienced menopause at age 50–52 years, women who experienced menopause before the age of 40 had the highest risk of dementia (adjusted hazard ratio [adjHR] 1.47; 95% confidence interval (CI) 1.39, 1.56; Fig. 1 and Supplementary Data Appendix 1). There was little evidence of any effect of type of menopause after adjustment for age at menopause and other covariates (adjHR 0.99; 95% CI 0.93, 1.04 for surgical menopause and adjHR 0.97; 95% CI 0.95, 1.00 for premenopausal hysterectomy compared to natural menopause; Fig. 1 and Supplementary Data Appendix 1). This was confirmed by an investigation of any interaction which showed little improvement in model fit (Supplementary Data Appendix 2). There was also little evidence of association with age at menarche and when this variable was omitted, the model fit improved. Consistent with other studies the risk of dementia decreased with increasing levels of education and increased with smoking, hypertension, diabetes and stroke (Fig. 1 and Supplementary Data Appendix 1). When BMI was included in the model, higher BMI appeared to be associated with lower risk of dementia; but high BMI was inversely associated with diabetes and, to a less extent with hypertension, and when BMI was omitted, the risk associated with diabetes decreased slightly (Supplementary Data Appendix 2). Analyses of data from each study separately showed little evidence of heterogeneity when the sample sizes and small numbers of dementia cases were considered (Supplementary Data Appendix 3; see also Supplementary Data Appendix 4 for the meta-analysis results).

Figure 1.

Forest plot showing associations between type of menopause, age at menopause and confounders and dementia in a pooled analysis of five cohort studies (N = 228.111). Model is stratified by study and adjusted for all variables included in the forest plot.

Discussion

This pooled analysis of individual participant data from five longitudinal studies showed that risk of dementia increased with earlier age at menopause. The highest risk was seen in women with premature menopause (i.e. before the age of 40), with the level of risk equivalent to that of smoking and stroke, and higher than the risk from hypertension. Increased risk associated with premenopausal bilateral oophorectomy or premenopausal hysterectomy (without bilateral oophorectomy) was not apparent once age at menopause was accounted for.

Our findings are consistent with two studies that also used data from the UK Biobank [7, 11]. Hao and colleagues considered women with natural menopause and women with surgical menopause (defined as bilateral oophorectomy) in separate analyses and reported higher dementia risk for women with natural menopause <40 years (HR 1.36, 95% CI 1.01, 1.83 vs natural menopause between 46–50 years), and surgical menopause <40 years (HR 1.94, 95% CI 1.38, 2.73 vs surgery 46–50 years) [7]. Liao and colleagues included all postmenopausal women (without distinguishing type of menopause) in their analysis and reported a 71% higher risk of dementia in menopause <40 years than menopause ≥50 years [11]. We have extended these UK Biobank studies by pooling cohorts from Australia, Sweden, England and the Netherlands with the UK Biobank data to increase the generalizability of the results. Additionally, by considering women with natural menopause, surgical menopause and premenopausal hysterectomy in the same analytic model we have shown that the mechanism of early menopause seems relatively unimportant.

Similar results linking early menopause with increased risk of dementia have been found in the Kaiser Permanente Northern California study [6], a nationwide Korean study [32] and a multinational study [8]. In contrast a French cohort study reported no association between premature menopause and dementia [9]; while another two studies (Rotterdam Study and Gothenburg study) reported higher risk of dementia with increasing age at menopause [12, 13]. The Gothenburg study reported age at menopause as a continuous variable and excluded women with menopause <37 years [12]; both the Gothenburg study and the Rotterdam study included only women who experienced natural menopause, although the latter is unlikely to explain the discrepancy in results.

In our study, associations between type of menopause and dementia attenuated when age at menopause was added to the analysis. Although not directly comparable, as other studies have used dichotomous variables with different reference categories (i.e. hysterectomy versus no hysterectomy, surgical menopause versus natural menopause, bilateral oophorectomy versus no bilateral oophorectomy) and different inclusion and exclusion criteria, our results are consistent with age-stratified analyses that indicate that younger age at hysterectomy and younger age at surgical menopause are associated with a higher risk of dementia [7, 14, 33].

We did not include menopausal hormone therapy (MHT) use as a potential effect modifier in our analysis for several reasons. First, only baseline data on MHT use was available in most studies so we would have been unable to explore important factors such as age at initiation or duration of use (with some evidence that older age at initiation and longer duration of use may increase dementia risk [3]). Second, our included studies came from different epochs (i.e. pre- and post-publication of the Women’s Health Initiative randomised controlled trials) with differences in both prescribing trends and types of MHT available [34]. Having said this, women who experience menopause before the age of 40 years (either spontaneous or surgical) are recommended to take hormones (either MHT or combined oral contraceptives) until at least the average age of natural menopause to treat hot flushes and protect against osteoporosis [35]. The exceptions to this are women diagnosed with hormone-dependent cancers for whom systemic use of MHT is not advised [35]. In our study 1.5% of women with natural menopause, 4.7% with surgical menopause and 0.6% with premenopausal hysterectomy had diagnoses of breast, endometrial or ovarian cancer before (or at in the case of some surgeries) menopause. On balance, if MHT is protective against dementia when initiated around the time of menopause, then any bias will have led to an underestimation of our results.

With respect to possible mechanisms, our results support the hypothesis of the neuroprotective and anti-ageing effect of greater exposure to endogenous oestrogen [4], and conversely that prolonged loss of ovarian function that comes with premature menopause may cause brain hypersensitivity to pathologic stressors which increases the risk of neurological diseases [36]. Similarly, oestrogen has protective effects on cardiovascular system, with earlier age at menopause seen as a marker for greater risk of cardiovascular disease [37] that is a key risk factor for dementia.

The Lancet Commission has identified 14 modifiable factors that might prevent or delay dementia [3]. While addressing these factors early in life is preferable, making changes in mid-life or later can also reduce dementia risk. The timing of menopause cannot be controlled, so we suggest clinicians could focus particularly on minimising these modifiable risk factors in women with premature menopause. At an individual level this would include reducing and treating vascular risks (such as high cholesterol, high blood pressure and diabetes), screening for vision and hearing loss, providing support to quit smoking and encouraging women to be physically active.

Several limitations of this study should be acknowledged. First, age at menopause and type of menopause were all self-reported. Although good validity and reproducibility of self-reported age at menopause [38], and self-reported history of hysterectomy [39] have been reported, and a moderate validity of self-reported history of bilateral oophorectomy documented [39], we cannot rule out recall bias. Second, the average age at the end of follow-up was only in the 70’s before the age when dementia risk increases rapidly and when other risk factors may become more important. Finally, although the inclusion of five cohort studies from four countries increased the generalisability of our results, participants in all studies were predominantly Caucasian [17–21], and in three cohorts were more educated than the respective population in each country [17, 18, 20]. Additional studies are needed in more racially and socioeconomically diverse populations.

Conclusions

Women with menopause before the age of 40, of any type, have a higher risk of dementia, similar or greater than that seen for established dementia risk factors such as current smoking, stroke and hypertension.

Disclaimer

Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer/World Health Organization.

Declaration of Conflicts of Interest

None.

Declaration of Sources of Funding

InterLACE project is funded by the Australian National Health and Medical Research Council project grant (APP1027196) and Centres of Research Excellence (APP1153420). GDM is supported by the Australian National Health and Medical Research Council Leadership Fellowship (APP2009577). The funding source played no role in the design; in the collection, analysis and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.