In 1985, a nationally representative sample of Australian schoolchildren participated in the Australian Schools Health and Fitness Survey (ASHFS) and had their health and fitness assessed . A subset of children aged 9, 12 and 15 years had their grip strength measured. Participants were followed up and attended clinics as part of the Childhood Determinants of Adult Health (CDAH) Study in 2004–06 when aged 28–36 years (young adulthood) and in 2014–19 when aged 38–49 years (mid-adulthood). During these adult follow-ups, participants had their grip strength reassessed and provided a fasting blood sample that was tested for glucose and glycated haemoglobin (HbA1c). Included in analyses were 263 participants who had their grip strength measured in childhood, young- and mid-adulthood and who provided a fasting blood sample in mid-adulthood, did not have type 1 diabetes and were not pregnant. A flowchart of participation is presented in Fig. 1. The ASHFS was approved by the State Directors General of Education. Follow-up studies were approved by the Southern Tasmania Health and Medical Human Research Ethics Committee and the Tasmania Health and Medical Human Research Ethics Committee. All participants provided written informed consent.
Grip Strength Across the Life Course
In childhood, young- and mid-adulthood, right and left grip strength was measured by maximum voluntary contraction using an isometric dynamometer (Smedley’s Dynamometer, TTM, Tokyo, Japan) that was adjusted to fit the size of the participant’s hand. Grip strength was measured by participants gripping the dynamometer with maximum force with one hand, whilst the dynamometer rested on the opposite shoulder. In childhood, participants had one attempt at right and left grip strength. In adulthood, the maximum of two attempts was used in analyses. At each time point, participants reported whether their dominant hand was right or left. To remove the influence of body mass on grip strength performance, dominant grip strength not attributable to body mass at all three life stages was created by regressing dominant grip strength on body mass and using the residuals added to the grand mean , and standardized for age and sex.
Regularly calibrated scales measured body mass to the nearest 0.5 kg in childhood and Heine scales (Heine, Dover, NH) measured body mass to the nearest 0.1 kg in adulthood. Height was measured to the closest 0.1 cm using a KaWe height tape (KaWe Kirchner & Wilhelm, Aspeg, Germany) in childhood and a Leicester height measure (Invicta, Leicester, UK) in adulthood. BMI was calculated as body mass (kg) divided by height (m) squared. Using a constant tension tape, child waist circumference was measured to the nearest 0.1 cm at the level of the umbilicus and adult waist circumference was measured at the narrowest point between the lower costal border and the iliac crest. Triceps, biceps, subscapular, and suprailiac skinfolds were measured using Holtain calipers (Holtain, Crymych, UK) to the nearest 0.2 mm in childhood and using Slim Guide Calipers to the nearest 0.5 mm in adulthood. Using the log of sum of four skinfolds, body density and fat percentage were calculated according to age-specific regression estimates . Using the Siri formula, body fat was calculated from body density . The difference between total body mass and fat mass was used to estimate fat-free mass.
Cardiorespiratory fitness (CRF) was estimated as physical work capacity at 170 beats per minute (PWC170) using a Monark 818E bicycle ergometer (Monark Exercise AB, Vansbro, Sweden) in childhood, a Monark 828E bicycle ergometer (Monark Exercise AB, Vansbro, Sweden) in young adulthood and a Monark 928G3r bicycle ergometer (Monark Exercise AB, Vansbro, Sweden) in mid-adulthood. Participants pedalled at a cadence of 60 RPM and the test included three 3-min workloads (childhood) or three 4-min workloads (adulthood) that increased resistance stepwise. In the final minute of each workload, watts and heart rate were measured, and the regression lines were extrapolated to estimate PWC170. To remove the influence of muscle mass, measures of PWC170 not attributable to fat-free mass were created by regressing PWC170 on fat-free mass and using the residuals added to the grand mean.
Prediabetes and Type 2 Diabetes
In mid-adulthood, participants provided a blood sample that was tested for glucose using a Siemens Advia 2400 Chemistry analyzer (Siemens Healthcare Diagnostics Inc., Deerfield, IL, USA) and HbA1c using a Bio-Rad D100 HbA1c analyzer (Bio-Rad Laboratories Inc., Hercules, CA, USA). A fasting status of ≥ 8 h was confirmed with the participant upon clinic arrival. Participants were categorized as having prediabetes or type 2 diabetes if they self-reported having type 2 diabetes or being on medication for type 2 diabetes, or if their fasting glucose levels were ≥ 5.6 mmol/L and/or HbA1c levels were ≥ 5.7% (≥ 39 mmol/mol), as defined by the American Diabetes Association .
Participant characteristics were examined using Stata (Version 15.0, StataCorp, College Station, Texas). For continuous variables, mean and standard deviation (SD) are presented. For categorical variables, percentage and number of participants are reported.
Bayesian Model for Life Course Investigation
In R (Version 3.5.3, R Foundation for Statistical Computing, Vienna, Austria)  using the Stan package to fit Bayesian models , the Bayesian relevant life course exposure model (BRLM) was used to identify the relative importance of grip strength measured in childhood, young adulthood and mid-adulthood on prediabetes or type 2 diabetes in mid-adulthood [15, 16]. Full methodological detail outlining the BRLM has been published previously [15, 16] and is summarized in the supplementary material. Briefly, the relative importance of grip strength at each period to the development of prediabetes or type 2 diabetes is assumed by weights (childhood = W1, young adulthood = W2, mid-adulthood = W3), allowing grip strength to associate with prediabetes or type 2 diabetes at different levels depending on the life stage at which it was measured. The relative weights and the joint posterior distribution of the weight parameters at each of the three life stages, visualized using a ternary plot, help determine the life course model best supported by the data. When the posterior distribution of weights cluster along vertices of the ternary plot, the model indicates critical periods for the corresponding life stage, and when the posterior distribution clusters in the central area of the plot, the model suggests an accumulation model [15, 16]. The BRLM also estimates an overall effect for the lifetime exposure of grip strength, representing the maximum accumulated effect of grip strength across the life course on prediabetes or type 2 diabetes, and derives life stage-specific effects (a combination of the overall effect and relative weights), representing the time-dependent association between grip strength and prediabetes or type 2 diabetes [15, 16]. Posterior distributions were used to compute mean and 95% credible intervals (95% CrI) for weights (interpreted as relative importance) and odds ratios (OR) for the overall effect.
In a sensitivity analysis, lifetime average standardized values of CRF and waist circumference were included as covariates. These covariates were derived by age- and sex-standardizing CRF and waist circumference in childhood, young- and mid-adulthood and creating a numerical average from across the life course.