Skip to main content

Table 3 Overview of longitudinal studies on the association between PA and the outcome of diabetes

From: The association between physical activity with incident obesity, coronary heart disease, diabetes and hypertension in adults: a systematic review of longitudinal studies published after 2012

Author Country Characteristics Follow-up time Predictor variable: Physical activity Outcome of interest Main results
Carlsson et al. (2013) [36] Sweden (Swedish Twin Registry) N = 23,539 Baseline: 1967–1972 Self-reported LTPA Incident type 2 diabetes Risk of type 2 diabetes decreased with PA: HR [95% CI]
- Low - Low: 1.0 (reference)
Follow-up: 1998–2002 - Moderate - Moderate: 0.77 [0.61, 0.96]
- high - High: 0.53 [0.37, 0.75]
Elwood et al. (2013) [40] UK (Caerphilly Prospective Study CaPS) N = 2235 men, 45–59 years 30 years Self-reported PA Incident diabetes (self-reported) OR [95% CI] for regular activity and incident diabetes
Baseline: 1979–1983 - 0.63 [0.46, 0.85]
Follow-up: 1984–1988, 1989–1993, 1993–1997, 2009
Grøntved et al. (2014) [41] USA (Nurses’Health Study NHS I and II) N = 99,316 women 8 years Self-reported PA (time spent on resistance exercise per week, lower intensity muscular conditioning exercises (yoga, stretching, toning), aerobic MVPA) Incident diabetes (self-report confirmed using standardized criteria; validated in sub-sample through medical chart review) RR [95% CI] for incident diabetes for aerobic MVPA
Baseline (NHS I): 2000 - None: 1.0 (reference)
- 1–29 min: 0.83 [0.74, 0.92]
Baseline (NHS II): 2001 - 30–59 min: 0.73 [0.65, 0.82]
- 60–150 min: 0.66 [0.60, 0.73]
Follow-up (NHS I): 2008 - None - ≥ 150 min: 0.46 [0.41, 0.50]
- 1–29 min/wk - Trend: p < 0.001
Follow-up (NHS II): 2009 - 30–59 min/wk Engaging in at least 150 min/wk of aerobic MVPA and at least 60 min/wk of muscle-strengthening activities was significantly associated with lower risk of incident diabetes compared with being inactive (pooled RR = 0.33 [0.29, 0.38]).
- 60–50 min/wk
- >  150 min/wk
Hjerkind et al. (2017) [37] Norway (Nord- Trøndelag Health Study) N = 38,413 with information on PA, 47% males 11 years Self-reported LTPA Incident diabetes (self-reported; validated through medical record) Risk of diabetes decreased with PAa: RR [95% CI] Women | Men:
Baseline: 1984–1986 - Low
- Medium - Low: 1.0 (reference)
Follow-up: 1995–1997 - High - Medium: 0.81 [0.65, 1.00] | 0.80 [0.66, 0.98]
- High: 0.76 [0.61, 0.95] | 0.65 [0.51, 0.84] p = 0.01 | p < 0.01
Gradual inverse association between frequency, duration, intensity and risk of incident diabetes for males
Gradual inverse association between frequency, intensity and risk of incident diabetes for females
Ekelund et al. (2012) [42] 8 European countries (EPIC–InterAct Study) N = 11,669 men, 15,695 women Median 12.3 years Self-reported PA (OPA, LTPA) Incident diabetes A one level difference in PA (e.g. between inactive and moderately inactive) was associated with a 13% relative reduction in risk of incident diabetes in males (HR [95% CI] 0.87 [0.80, 0.94]) and 7% risk reduction in females (0.93 [0.89, 0.98])b
- Inactive
N = 15,934 subcohort (6009 men, 9925 women) Baseline: 1991 - Moderately inactive
Follow-up: 2007 - Moderately active
- Active
Increased risk of incident diabetes associated with lower levels of PA evident across BMI strata in both sexes, with the exception of obese women
Jefferis et al. (2012) [38] UK N = 3012 men, 68.3 years Median 7.1 years Self-reported PA Incident type 2 diabetes (self-report included after validation through medical record) Risk of diabetes decreased with PA: Dose-response associationc: HR [95% CI]
- None
Baseline: 1996, 1998–2000
Follow-up:
- Occasional - None: 1.0 (reference)
- Light - Occasional: 0.54 [0.31, 0.96]
2006 - Moderate - Light: 0.34 [0.18, 0.65]
- Moderately vigorous - Moderate: 0.33 [0.17, 0.65]
- Vigorous - moderately vigorous: 0.32 [0.16, 0.61]
- vigorous: 0.26 [0.13, 0.53] p < 0.01
Taking up at least moderate intensity PA also associated with lower risk of diabetes.
Koloverou et al. (2017) [39] Greece (Attica Study) N = 1485, 49% males 10 years Self-reported PA (MET-min/wk) Incident diabetes (measured in biological sample or self-reported) Moderate intensity PA associated with lower risk of incident diabetesd: OR [95% CI]
Baseline: 2001–2002 - Very low ≤150
- Low = 150–330 - Very low: 1.0 (reference)
Follow-up: 2011–2012 - Moderate = 331–1484 - Low: 0.77 [0.41, 1.49]
- High ≥1484 - Moderate: 0.47 [0.24, 0.93]
- High: 1.04 [0.59, 1.82]
Medina et al. (2018) [49] Mexico (Mexico City Diabetes Study) N = 1883, median 45 years (IQR 39–52); 42.7% males Median 14.4 person years Self-reported PA (occupational, leisure, total PA in MET-min/wk of MVPA) Incident type 2 diabetes (measured, self-reported, taking medication) Participants with leisure PA < 1 MET-min/wk had increased risk of incident diabetes (HR 1.45 [95% CI: 1.10, 1.92]) as compared to reference group (≥ 1200 MET-min/wk of MVPA; p = 0.008)e
Baseline: 1989–1990
- 1 = < 1
No association between occupational and total PA and diabetes risk.
Follow-up: 1993–1994, 1997–1998, 2008–2009 - 2 = 1–599.9
- 3 = 600–1199.9
- 4 = ≥ 1200
Mehlig et al. (2014) [43] Sweden N = 1448 women, 38–60 years 34 years Self-reported LTPA Incident diabetes LTPA is associated with an elevated risk in incident diabetes: HR [95% CI]f
Baseline: 1968–1969 - Almost inactive: low LTPA
- Some PA at least 4 h/wk - Non-obese, active: 1.0 (reference)
Follow-up: 1974–1975, 1980–1981, 1992–1993, 2000–2001, 2000 - Regular exercise - Non-obese, inactive: 1.79 [1.15, 2.79]
- Regular training and competitive sports - Obese, active: 2.43 [1.44, 4.09]
- Obese, inactive: 11.7 [6.28, 21.8]
Shi et al. (2013) [44] China N = 51,464 men, 54.1 ± 9.3 years Median 5.4 years Self-reported PA MET level (in quintiles) Incident diabetes (self-reported) Total PA is associated with a reduced risk in incident diabetes: HR [95% CI] for MET levelg
Baseline: 2002–2006 - Q1 < 4.3 - Q1: 1.0 (reference)
- Q2 4.3–6.5 - Q2: 0.84 [0.72, 0.99]
Follow-up: 2004–2008, 2008–2011 - Q3 6.5–8.9 - Q3: 0.72 [0.61, 0.85]
- Q4 8.9–12.1 - Q4: 0.66 [0.55, 0.78]
- Q5 ≥ 12.1 - Q5: 0.65 [0.54, 0.77]
Williams & Thompson (2012) [50] USA N = 48,116 Median 6.2 years Self-reported PA MET-h/d Incident diabetes Greater MET-h/d is associated with lower risk of incident diabetes: HR [95% CI]h
- Light
Baseline: 1998–1999 - Moderate - Running: 0.879 [0.83, 0.929]
- Vigorous - Walking: 0.877 [0.82, 0.93]
Follow-up: 2006 - Other vigorous: 0.98 [0.95, 1.007]
- Other moderate: 0.969 [0.908, 1.02]
- Other light: 0.99 [0.736, 1.12]
  1. Abbreviation: BP - CI confidence interval, d day, h hour, HR hazard ratio, IQR interquartile range, LTPA leisure time physical activity, MET metabolic equivalent, min minutes, MVPA moderate to vigorous physical activity, N number of participants, OPA occupational physical activity, OR odds ratio, PA physical activity, RR relative risk, wk week
  2. a: Model adjusted for age; education, alcohol frequency in the past 2 weeks, smoking, blood pressure medication use, prevalent cardiovascular disease, BMI, PA summary score; b Model adjusted for study center, education, smoking status, alcohol consumption, energy intake, BMI; c Model adjusted for age & region; d Model adjusted for age, sex, family history of diabetes, hypertension, hypercholesterolemia, smoking status, education, physical activity, waist circumference, adherence to the Mediterranean diet, fasting glucose, triglycerides; e Model adjusted for sex, age, education levels, marital status, current smoking, alcohol intake, total energy intake, parent history of diabetes, sleeping hours, leisure/working MET-min/wk; f Model adjusted for baseline covariates age, education, smoking, consumption of alcohol, triglycerides, hypertension, parental history of diabetes (diabetes only); g Model adjusted for age at interview, energy intake, smoking, alcohol consumption, education level, occupation, income level, hypertension, family history of diabetes; h Model adjusted for baseline age (age, age2), sex, race, education, smoking, intakes of red meat, fruit, alcohol, preexisting CHD at baseline