Reference Study name Country Study design Follow‐up Funding	Original cohort (N total) Exclusion criteria Study population	Ascertainment of outcome	Exposure groups n/person‐years Exposure assessment method	Incident cases	Model covariates	Results
Melhus et al. (1998) Swedish Mammography Cohort Sweden NCC 2–64 months Funding: Public	N = 1247 Population sampled: Females, 40–76 years old living in two counties in Sweden from 1987 to 1990 Exclusion criteria: Incorrect diagnosis of hip fractures, fractures due to cancer or high‐energy trauma n = 1120 Sex: Females Age: Cases: 67.6 ± 7.3 years Controls: 67.7 ± 7.3 years Other characteristics BMI (kg/m2): Cases: 24.4 ± 3.9 Controls: 25.9 ± 4.3 Current smokers: n = 137 Alcohol consumption: NR Vitamin D intake (μg/day): NR Calcium intake (mg/day): NR β‐carotene intake (mg/day): Cases: 800 ± 700 Controls: 750 ± 600 Current HRT users: 56 Previous osteoporotic fracture= Cases: 82/247 Controls: 165/874	Hip fractures (cervical, trochanteric or subtrochanteric femoral fracture): from hospital records	Preformed vitamin A intake (diet only), μg RE/day: Mean ± SD Cases: 960 ± 480 Controls: 880 ± 430 Categories 1: ≤ 500 (ref.) 2: 510–1000 3: 1010‐1500 4: > 1500 No participant had intake less than 270 μg RE/day n/person‐years per category: NR Total person‐years: NR Exposure assessment: A 60‐item FFQ at baseline (validation not published)	Hip fractures: Cases = 247 Controls = 873 Matching criteria = age, county of residence	Model 1: Univariate Model 2: Adjusted for BMI, energy intake, age at menopause, lifetime physical activity during leisure time, cigarette smoking, hormone replacement therapy, diabetes mellitus, use of oral contraceptives or cortisone, previous osteoporotic fracture of the distal forearm or proximal humerus, menopause at time of the 2. questionnaire, former athletic activity	OR (95% CI) Per 1000 μg RE preformed vitamin A: Univariate model: 1.56 (1.15–2.11) Adjusted model: 1.68 (1.18–2.40) Per category: Univariate model: C1. 1.0 (ref) C2. 0.93 (0.61–1.41) C3. 1.27 (0.80–2.02) C4. 1.95 (1.11–3.43) Adjusted model: C1. 1.0 (ref) C2. 0.92 (0.57–1.46) C3. 1.34 (0.77–2.31) C4. 2.05 (1.05–3.98) Additionally adjusted for iron, magnesium, vitamin C, and calcium intake: OR high vs. low: 1.54 (1.06–2.24), p = 0.02
Feskanich et al. (2002) Nurses' health study USA PC Up to 18 years Funding: Mixed	N = 121,700 Population sampled: Postmenopausal registered female nurses Exclusion criteria: Premenopausal, previous hip fracture, diagnosis of cancer, heart disease, stroke or osteoporosis % lost to follow up: NR n = 72,377 Sex: Females Age (mean, no SD given) per quintile of total vitamin A intake Q1: 58.3 Q2: 59.3 Q3: 59.7 Q4: 60.0 Q5: 60.5 Other characteristics (mean, no SD given) BMI, kg/m2 Q1: 26.0 Q2: 26.0 Q3: 25.9 Q4: 25.8 Q5: 25.7 Physical activity, h/wk Q1: 2.4 Q2: 2.7 Q3: 2.9 Q4: 3.0 Q5: 3.2	Hip fractures: Self‐reported (questionnaire). Only fractures due to low or moderate trauma were considered cases	Cumulative average intake across quintiles (μg RE/day): Preformed vitamin A intake (diet and supplements): Q1 (ref) (< 500): 487 Q2 (500–849): 763 Q3 (850–1299): 1085 Q4 (1300‐1999): 1607 Q5 (≥ 2000): 3206 n/Person‐years: 72,377/313,308 Preformed vitamin A intake (diet only): Q1 (ref) (< 400): 425 Q2 (400–549): 553 Q3 (550–699): 666 Q4 (700–999): 802 Q5 (≥ 1000):1014 n/Person‐years: 34,386/313,138 Total vitamin A intake (diet and supplements) Q1 (ref) (< 1250): 965 Q2 (1250‐1699): 1442 Q3 (1700‐2249): 1890 Q4 (2250‐2999): 2491 Q5 (≥ 3000): 4274 n/Person‐years: 72,377/313,308 Total vitamin A intake (diet only): Q1 (ref) (< 1000): 811 Q2 (1000‐1299): 1146	Hip fracture cases Preformed vitamin A intake (diet and supplements) Q1 (ref): 102 Q2: 122 Q3: 111 Q4: 122 Q5: 146 Preformed vitamin A intake (diet only) Q1 (ref): 31 Q2: 36 Q3: 29 Q4: 42 Q5: 52 Total vitamin A intake (diet and supplements) Q1 (ref): 118 Q2: 123 Q3: 121 Q4: 124 Q5: 137 Total vitamin A intake (diet only) Q1 (ref): 22 Q2: 30 Q3: 25 Q4: 32 Q5: 30	Model 1: Adjusted for age Model 2: Adjusted for age, follow‐up cycle, body mass index, use of postmenopausal hormones, smoking, hours of leisure‐time activity per week, use of thiazide diuretics, and intakes of calcium, protein, vitamin D, vitamin K, alcohol, and caffeine + total energy	Preformed vitamin A intake (diet and supplements) Model 1 Q1 (ref): 1.00 Q2: 1.12 (0.86–1.46) Q3: 0.99 (0.76–1.30) Q4: 1.08 (0.83–1.40) Q5: 1.25 (0.97–1.60) p for trend = 0.03 Model 2 Q1 (ref): 1.00 Q2: 1.25 (0.95–1.65) Q3: 1.18 (0.88–1.59) Q4: 1.43 (1.04–1.96) Q5: 1.89 (1.33–2.68) p for trend <0.001 Preformed vitamin A intake (diet only) Model 1 Q1 (ref): 1.00 Q2: 1.20 (0.74–1.94) Q3: 0.92 (0.55–1.53) Q4: 1.34 (0.84–2.15) Q5: 1.67 (1.07–2.61) p for trend = 0.05 Model 2 Q1 (ref): 1.00 Q2: 1.27 (0.77–2.07) Q3: 0.96 (0.57–1.63) Q4: 1.41 (0.86–2.32) Q5: 1.69 (1.05–2.74) p for trend = 0.05t Total vitamin A intake (diet and supplements) HR (95% CI)
	Total calcium intake, mg/day Q1: 719 Q2: 827 Q3: 887 Q4: 947 Q5: 1058 Total vitamin D intake, μg/day Q1: 4.2 Q2: 5.6 Q3: 7.1 Q4: 9.5 Q5: 13.8 Current use of postmenopausal hormones, % Q1: 29 Q2: 30 Q3: 33 Q4: 34 Q5: 34		Q3 (1300‐1599): 1427 Q4 (1600‐1999): 1763 Q5 (≥ 2000): 2507 n/Person‐years: 28,676/217,635 Exposure assessment Up to 5 times repeated (61 to > 130‐items, depending on the iteration) semi‐quantitative validated FFQs Conversion factor for pro‐vitamin A carotenoids: NR			Model 1 Q1 (ref): 1.00 Q2: 0.79 (0.60–1.02) Q3: 0.89 (0.69–1.14) Q4: 0.88 (0.69–1.14) Q5: 0.94 (0.74–1.21) p for trend = 0.55 Model 2 Q1 (ref): 1.00 Q2: 0.92 (0.70–1.22) Q3: 1.13 (0.85–1.49) Q4: 1.24 (0.92–1.68) Q5: 1.48 (1.05–2.07) p for trend = 0.003 Total vitamin A intake (diet only) Model 1 Q1 (ref): 1.00 Q2: 1.33 (0.77–2.31) Q3: 1.16 (0.66–2.05) Q4: 1.34 (0.77–2.34) Q5: 1.40 (0.81–2.42) p for trend = 0.53 Model 2 Q1 (ref): 1.00 Q2: 1.51 (0.86–2.66) Q3: 1.37 (0.74–2.51) Q4: 1.74 (0.96–3.14) Q5: 1.82 (0.97–3.40) p for trend = 0.24
Michaëlsson et al. (2003) Uppsala Longitudinal Study of Adult Men Sweden PC up to 11 years Funding: Mixed	N = 2322 Population sampled: General population born between 1920 and 1924, living in Uppsala Exclusion criteria: Missing serum retinol samples % lost to follow up: NR n = 2032/1221 a Sex: Males Baseline variables reported according to quintiles of serum retinol Age, years Q1: 49.7 ± 0.6 Q2: 49.6 ± 0.6 Q3: 49.7 ± 0.6 Q4: 49.7 ± 0.6 Q5: 49.7 ± 0.6 Other characteristics BMI, kg/m2 Q1: 24.1 ± 3.0 Q2: 25.1 ± 3.3 Q3: 24.9 ± 3.2 Q4: 25.5 ± 3.4 Q5: 25.6 ± 3.1 Leisure physical activity ≥3 h per week, % Q1: 42 Q2: 48 Q3: 49 Q4: 47 Q5: 41 Vitamin D intake: NR Calcium intake: NR	Any fracture site Ascertained through medical records and linkage to the Hospital Discharge Register. Cases of fracture caused by cancer were excluded	Preformed vitamin A intake (diet and supplements), RE μg/day Mean ± SD (Range) Q1 (ref): 410 ± 90 (< 530) Q2: 640 ± 60 (540–740) Q3: 880 ± 80 (750–1040) Q4: 1290 ± 150 (1050–1560) Q5: 2370 ± 770 (> 1560) Preformed vitamin A intake (diet only) Q1 (ref): 410 ± 90 (< 530) Q2: 640 ± 60 (530–730) Q3: 860 ± 70 (730–990) Q4: 1230 ± 150 (1000‐1500) Q5: 2250 ± 720 (> 1500) n/Person‐years: NR Exposure assessment method: 7‐day dietary records in subsample.	111, among those with dietary information (n = 1138) Cases per quintile of intake NR	Model 1: Adjusted for total energy intake. No other model covariates reported for models with dietary vitamin A.	RR (95% CI) of any fracture for Preformed vitamin A (diet + supplements) Q1 (ref): 1 Q5: 1.99 (0.98, 4.01) p for trend: NR Preformed vitamin A (diet only) Q1 (ref): 1 Q5: 2.00 (1.00, 3.99) p for trend: NR Intermediate quartile RRs not reported
Lim et al. (2004) Iowa Women's Health Study USA PC 9.5 years (mean) Funding: Public	N = 41,836 Population sampled: Females, aged 55–69 from the general population Exclusion criteria: Premenopausal at baseline, implausible energy intakes, incomplete FFQ, history of cancer (except skin cancer) % lost to follow up: 21 n = 34,703 Sex: Females Median age 61 years Age‐adjusted characteristics stratified by vitamin A supplementation and total vitamin A intake (measure of central tendency or dispersion not given for all variables): Supplement users: BMI, kg/m2: 26.6 Physically active (%): 56.4 Prior fracture (%): 14.1 Mean calcium intake, mg/day: 1295 Mean vitamin D intake, IU/day: 671 Non‐supplement users: BMI, kg/m2: 27.2 Physically active (%): 49.2 Prior fracture (%): 13.4 Mean calcium intake, mg/day: 984 Mean vitamin D intake, IU/day: 269	Hip and total fractures: Self‐reported (via questionnaire)	Quintiles of Preformed vitamin A (food and supplements), mean (range), μg RE/day Q1 (ref): 274 (8–422) Q2: 609 (422–886) Q3: 1157 (886–1397) Q4: 1730 (1397‐2100) Q5: 3783 (2101‐63,315) n/Person‐years hip fractures: Q1: 6940/65,807 Q2: 6941/67,194 Q3: 6941/65,468 Q4: 6941/66,052 Q5: 6940/65,290 n/Person‐years all fractures: Q1: 6940/58,648 Q2: 6941/60,455 Q3: 6941/58,304 Q4: 6941/59,129 Q5: 6940/58,527 Quintiles of total vitamin A (food and supplements) (Mean (range), IU/day [μg RE/day]) Q1 (ref): 1534 (66.3–2117) Q2: 2631 (2117–3145) Q3: 3679 (3146–4263) Q4: 5029 (4263–5968) Q5: 8771 (5968–71,097) n/Person‐years hip fractures: Q1: 6940/64,989 Q2: 6941/65,688 Q3: 6941/58672 Q4: 6942/66,724 Q5: 6940/66,068	Hip fractures Preformed vitamin A (food and supplements) Q1: 109 Q2: 84 Q3: 116 Q4: 101 Q5: 125 Total vitamin A (food and supplements) Q1: 93 Q2: 122 Q3: 102 Q4: 99 Q5: 119 All fractures Total Vitamin A (food and supplement) Q1: 1298 Q2: 1319 Q3: 1256 Q4: 1311 Q5: 1319 Preformed vitamin A (food and supplement) Q1: 1324 Q2: 1238 Q3: 1346 Q4: 1270 Q5: 1324	Hip fractures Total and preformed retinol dietary intake Model 1: Age Model 2: Age, BMI, waist‐to‐hip‐ratio, diabetes mellitus, past irregular menstrual duration, physical activity, steroid medication, oestrogen replacement and energy intake Total and preformed vitamin A supplements Model 1: Age Model 2: Age, BMI, waist‐to‐hip ratio, diabetes mellitus, physical activity, steroid medication and oestrogen replacement therapy All fractures Total and preformed vitamin A dietary intake Model 1: Age Model 2: Age, BMI, diabetes mellitus, cirrhosis, past irregular menstrual duration, thyrotropic medication, sedative medication, steroid medication, antiepileptic medication, diuretic medication, education, alcohol use, energy intake Total and preformed vitamin A supplements Model 1: Age Model 2: Age, BMI, diabetes mellitus, cirrhosis, thyrotropic medication, antiepileptic medication, sedative medication, steroid medication, education	Hip fractures RR (95% CI) Preformed vitamin A intake quintiles (food and supplements) Model 1 Q1: 1.00 (ref) Q2: 0.72 (0.54–0.96) Q3: 1.03 (0.79–1.33) Q4: 0.88 (0.67–1.15) Q5: 1.10 (0.85–1.42) p for trend = 0.21 Model 2 Q1: 1.00 (ref) Q2: 0.69 (0.52–0.93) Q3: 1.03 (0.79–1.34) Q4: 0.86 (0.65–1.14) Q5: 1.10 (0.84–1.43) p for trend = 0.19 Total vitamin A intake quintiles (food and supplements) Model 1 Q1: 1.00 (ref) Q2: 1.26 (0.96–1.65) Q3: 1.03 (0.78–1.37) Q4: 0.97 (0.73–1.29) Q5: 1.17 (0.89–1.54) p for trend = 0.85 Model 2 Q1: 1.00 (ref) Q2: 1.27 (0.97–1.67) Q3: 1.08 (0.81–1.44) Q4: 1.02 (0.76–1.37) Q5: 1.25 (0.94–1.68) p for trend = 0.49 All fractures Preformed vitamin A intake quintiles (food and supplements)
	1st and 5th quintile of vitamin A intake BMI, kg/m2: Q1: 27.1 Q5: 26.9 Physically active, %: Q1: 40.0 Q5: 62.0 Prior fracture, % Q1: 13.4 Q5: 14.0 Mean calcium intake, mg/day: Q1: 874 Q5: 1318 Mean vitamin D intake, IU/day: Q1: 219 Q5: 632		n/Person‐years all fractures: Q1: 6940/58,073 Q2: 6940/66,434 Q3: 6940/59,622 Q4: 6942/59,717 Q5: 6940/58,979 Exposure assessment 127‐item validated semi‐quantitative FFQ Conversion factors for carotenoids not reported.		Non‐users only Preformed vitamin A intake Model 1: Age Model 2: Age, body mass index, diabetes mellitus, cirrhosis, past irregular menstrual duration, thyrotropic medication, sedative medication, steroid medication, antiepileptic medication, diuretic medication, education, alcohol use, and energy intake Total vitamin A intake Model 1: Age Model 2: Age, body mass index, waist‐to‐hip ratio, diabetes mellitus, past irregular menstrual duration, physical activity, steroid medication, oestrogen replacement, and energy intake	Model 1 Q1: 1.00 (ref) Q2: 0.90 (0.83–0.97) Q3: 1.01 (0.94–1.09) Q4: 0.94 (0.87–1.01) Q5: 0.99 (0.91–1.06) p for trend = 0.86 Model 2 Q1: 1.00 (ref) Q2: 0.89 (0.82–0.96) Q3: 1.00 (0.93–1.08) Q4: 0.92 (0.85–1.00) Q5: 0.96 (0.89–1.04) p for trend = 0.61 Total Vitamin A intake quintiles (food and supplements) Model 1 Q1: 1.00 (ref) Q2: 1.00 (0.92–1.06) Q3: 0.93 (0.86–1.01) Q4: 0.97 (0.89–1.04) Q5: 0.98 (0.91–1.06) p for trend = 0.43 Model 2 Q1: 1.00 (ref) Q2: 0.98 (0.91–1.06) Q3: 0.92 (0.85–0.99) Q4: 0.94 (0.87–1.02) Q5: 0.95 (0.87–1.03) p for trend = 0.098
White et al. (2006) Leisure World Cohort Study USA PC Median follow‐up times for males: Hip fractures: 9.8 years Wrist fractures: 4.7 years Spine fractures: 5.3 years Median follow‐up times for females: Hip fractures: 11.8 years Wrist fractures: 7.4 years Spine fractures: 9.1 years Funding: Mixed	N = 13,978 Population sampled: Residents of a retirement community, mostly white, upper‐middle socioeconomic class Exclusion criteria: NR % lost to follow up: NR n = 13,978 Sex: 63.5% females Age, years: Males: 74.9 ± 7.2 Females: 73.7 ± 7.4 Other characteristics Males : BMI 24.1 ± 2.9 kg/m2 Smoking pack‐years 25.9 ± 29.7 Alcohol consumption drinks/day: 1.64 ± 1.6 Active activities, h/day: 1.1 ± 1.3 Previous fracture after age 40, but before study entry: 14% Females: BMI: 23.1 ± 3.5 kg/m2 Smoking pack‐years: 12.7 ± 22.2 Alcohol consumption drinks/day: 1.17 ± 1.22 Postmenopausal oestrogen: 54% Active activities h/day: 0.9 ± 1.1 Previous fracture after age 40, but before study entry: 30% Vitamin D intake: NR Calcium intake: NR	Hip, wrist and spine fractures: Self‐reported (follow‐up surveys), hospital discharge records and death certificates	Preformed vitamin A supplement, b μg RE/day Mean ± SD Males 14,610 ± 26,010 Females 15,930 ± 24,120 N/person‐years Males Hip fractures: 4769/49,586 Wrist fractures: 4177/31,654 Spine fractures: 3890/32,570 Females Hip fractures: 6850/98,290 Wrist fractures: 6393/62,734 Spine fractures: 7153/67,239 Exposure assessment method: Non‐validated survey/questionnaire.	Males Hip: 278 Wrist: 56 Spine: 167 Females Hip: 949 Wrist: 389 Spine: 562	Model 1, all fracture sites (hip, wrist, spine): Age Model 2 (females only) Hip: age, previous fracture, BMI, current smoker, pack‐years of smoking, diabetes, glaucoma, attitude, ever pregnant Wrist: age, previous fracture, BMI, hysterectomy, heart attack, alcohol,, cola consumption	HR (95% CI) of fracture per 3000 μg RE/day increase in supplemental preformed vitamin A: Males Model 1 Hip: 1.00 (1.00–1.00) Wrist: 1.00 (0.99–1.00) Spine: 1.00 (1.00–1.00) Model 2 NR for any fracture site Females Model 1 Hip: 1.00 (1.00–1.00) Wrist: 1.00 (1.00–1.00) Spine: 1.00 (1.00–1.00) Model 2 Hip: 1.07 (1.00–1.15) Wrist: 1.15 (1.07–1.23) Spine: NR Note: The multiple regression models used a backwards elimination procedure and retained variables significant at 0.05 in the models. Vitamin A supplement use was not a significant variable in model 2 for males (any fracture site) or females (spine fractures)
Hayhoe et al. (2017) EPIC‐Norfolk UK PC 12.5 years (mean) Funding: Public	N = 25,639 Population sampled: Aged 39–79, between 1993 and 1997 Exclusion criteria: NR % lost to follow up: NR n = 25,439 Sex: 54.8% females Age, years: Females: 58.9 ± 9.3 Males: 59.7 ± 9.3 BMI, kg/m2 Females: 26.2 ± 4.3 Males: 26.5 ± 3.3 Dietary calcium intake (mg/day): Females: 766 ± 249 Males: 919 ± 298 Current smoker (%) Females: 12.1 Males: 12.8 Physically active (%) Females: 15.3 Males: 21.5 Corticosteroid use, > 3 months (%) Females: 3.4 Males: 3.0	Osteoporotic hip, spine or wrist fractures: Self‐reported, corroborated via linkage to hospital attendance database	Quintiles of preformed vitamin A (diet and supplements), μg RE/day Mean ± SD, Range Females: Q1: 146 ± 49 (0–216) Q2: 271 ± 33 (217–331) Q3: 431 ± 71 (331–593) Q4: 889 ± 140 (594–1109) Q5: 2505 ± 2179 (1109–43,483) Males: Q1: 184 ± 59 (0–265) Q2: 330 ± 38 (265–398) Q3: 493 ± 60 (398–617) Q4: 880 ± 157 (617–1158) Q5: 2911 ± 2832 (1158–57,714) N per quintile: Females: Q1: 2786 Q2: 2786 Q3: 2786 Q4: 2786 Q5: 2785 Males: Q1: 2302 Q2: 2302 Q3: 2302 Q4: 2302 Q5: 2302 Exposure assessment method: 7‐day food diaries	Diet + supplements analyses Hip Fractures Females: Q1: 132 Q2: 128 Q3: 114 Q4: 136 Q5: 155 Males: Q1: 41 Q2: 45 Q3: 46 Q4: 55 Q5: 44 Total fractures (hip, wrist and spine) Females: Q1: 238 Q2: 224 Q3: 211 Q4: 242 Q5: 250 Males: Q1: 113 Q2: 79 Q3: 88 Q4: 92 Q5: 95 Wrist Fractures Females: Q1: 91 Q2: 75 Q3: 73 Q4: 83 Q5: 76 Males: Q1: 31 Q2: 27 Q3: 17 Q4: 24 Q5: 16	Age, BMI, family history of osteoporosis, menopausal and hormone replacement therapy status in women, corticosteroid use, smoking status, physical activity, calcium intake, total energy intake, calcium and vitamin D‐containing supplement use, days of food diary completed and the ratio of energy intake: estimated energy requirement	Preformed vitamin A (diet + supplements) HR (95% CI) Hip Fractures Females: Q1: 1 Q2: 0.89 (0.69, 1.14) Q3: 0.78 (0.60, 1.02) Q4: 0.98 (0.72, 1.33) Q5: 0.97 (0.72, 1.30) Males: Q1: 1 Q2: 1.16 (0.75, 1.79) Q3: 1.20 (0.77, 1.88) Q4: 1.17 (0.71, 1.92) Q5: 1.32 (0.81, 2.16) Total Fractures Females: Q1: 1 Q2: 0.88 (0.73, 1.06) Q3: 0.83 (0.68, 1.01) Q4: 0.93 (0.74, 1.18) Q5: 0.87 (0.70, 1.10) Males: Q1: 1 Q2: 0.67 (0.50, 0.90) Q3: 0.72 (0.53, 0.96) Q4: 0.77 (0.55, 1.07) Q5: 0.75 (0.54, 1.05) Wrist Fractures Females: Q1: 1 Q2: 0.80 (0.58, 1.09) Q3: 0.78 (0.56, 1.10) Q4: 0.73 (0.49, 1.10) Q5: 0.64 (0.43, 0.96) Males: Q1: 1 Q2: 0.72 (0.42, 1.21) Q3: 0.37 (0.20, 0.69) Q4: 0.52 (0.27, 0.99) Q5: 0.35 (0.17, 0.75)
				Spine Fractures Females: Q1: 57 Q2: 48 Q3: 41 Q4: 56 Q5: 47 Males: Q1: 46 Q2: 15 Q3: 30 Q4: 28 Q5: 30		Spine Fractures Females: Q1: 1 Q2: 0.82 (0.55, 1.23) Q3: 0.72 (0.47, 1.12) Q4: 1.07 (0.65, 1.75) Q5: 0.82 (0.50, 1.34) Males: Q1: 1 Q2: 0.31 (0.17, 0.56) Q3: 0.59 (0.36, 0.96) Q4: 0.54 (0.31, 0.97) Q5: 0.54 (0.30, 0.97)
Key et al. (2007) EPIC‐Oxford UK PC 5.2 years (mean) Funding: Public	N: 57,450 Population sampled: General population and vegetarians living in the UK, aged 20 and above between 1993 and 2000 Exclusion criteria: did not answer follow‐up question about fractures; reported fractures of the digits or ribs, had any type of fracture before recruitment; unreliable nutrient intake (≥ 20% FFQ missing, or daily energy intakes <500 kcal or > 3500 kcal for females or < 800 kcal or > 4000 kcal for males)	All fractures: Self‐reported fractures in bones other than the digits or ribs, accruing after study recruitment	Categories of preformed vitamin A (food only) (μg RE/day) C1: < 200 C2: 200–299 C3: 300–449 C4: 450–999 C5: ≥ 1000 Mean (SD) intakes per category of preformed vitamin A (food only) (μg RE/day)* Females: C1: 127 (52) C2: 251 (29) C3: 369 (43) C4: 627 (160) C5: 1463 (879)	Incident fractures: Female: n = 1555 Male: n = 343 Females: C1: 289 C2: 299 C3: 398 C4: 367 C5: 202 Males: C1: 75 C2: 58 C3: 81 C4: 85 C5: 44	Age at recruitment, smoking, intakes of energy and calcium, protein, vitamins D and C, carotene, potassium and magnesium, alcohol consumption, BMI, walking, cycling, vigorous exercise, other exercise, physical activity at work, marital status and, for females, parity and use of hormone replacement therapy	Incidence rate ratio (95% CI) Females: C1: (ref) 1.00 C2: 0.96 (0.80–1.14) C3: 0.99 (0.82–1.19) C4: 1.03 (0.84–1.27) C5: 0.93 (0.73–1.18) p for trend = 0.97 Males: C1: (ref) 1.00 C2: 0.92 (0.62–1.36) C3: 0.97 (0.65–1.45) C4: 0.91 (0.58–1.42) C5: 0.80 (0.47–1.34) p for trend = 0.54
	Lost to follow up: NR n = 34,696 Sex: 77% female Age: Female: 45.8 ± 13.1 Male: 49.5 ± 13.5 Other characteristics BMI (kg/m2): Female: 23.6 ± 3.9 Male: 24.2 ± 3.3% Current smokers: Female: 8.9 Male: 11.5 Alcohol consumption (g/day): Female: 7.7 ± 9.6 Male: 15.1 ± 17.8 Vitamin D intake (μg/day): Female: 2.7 ± 1.9 Male: 2.73 ± 1.99 Calcium intake (mg/day): Female: 996 ± 329 Male. 1046 ± 363% current HRT users: 13.5% ≥ 3h vigorous exercise/week: Female: 27.6 Male: 34.3		Males C1: 111 (58) C2: 251 (28) C3: 370 (43) C4: 636 (153) C5: 1531 (1189) N/person‐years per category*: Females: C1: 5139 C2: 5564 C3: 7149 C4: 6330 C5: 3403 Males: C1: 1449 C2: 1299 C3: 1913 C4: 2157 C5: 1299 *Data received from study authors for total N of 35,702, which is ~ 1000 more participants than what is reported in the publication Exposure assessment: 130‐item, validated FFQ, covering the previous 12 months, at baseline
Caire‐Juvera et al. (2009) Women's' Health Initiative Observational Study USA PC 6.6 years (mean) Funding: Public	N: 93,676 Population sampled: Postmenopausal females aged 50–79 years, unlikely to move or die ≤ 3, not participating in other clinical trials Exclusion criteria: Missing FFQ data; previous fractures or diagnosis of osteoporosis Lost to follow up: 1,9% n = 75,747 Sex: female Age (years) per quintile of vitamin A intake: Q1: 63.7 ± 7.3 Q2: 63.8 ± 7.3 Q3: 63.7 ± 7.3 Q4: 63.6 ± 7.2 Q5: 63.1 ± 7.2 Other characteristics BMI (kg/m2) per quintile of vitamin A intake: Q1: 27.1 ± 5.8 Q2: 26.8 ± 5.5 Q3: 26.7 ± 5.5 Q4: 27.0 ± 5.6 Q5: 27.6 ± 6.1 Vitamin D intake (ug/day) per quintile of vitamin A intake: Q1: 5.6 ± 3.3 Q2: 8.9 ± 4.2 Q3: 11.5 ± 4.7 Q4: 13.4 ± 4.6 Q5: 15.4 ± 5.6	Hip fractures: self‐reported (by participants or proxy) and adjudicated by central review of radiology and other medical reports Other fractures: self‐report (by participants or proxy	Quintiles of preformed vitamin A (food and supplements), μg RE/day: Q1 (ref): < 474 Q2: 474–764 Q3: 765–1092 Q4: 1093–1425 Q5: ≥ 1426 μg N Q1 (ref): 15,149 Q2: 15,149 Q3: 15,150 Q4: 15,149 Q5: 15,150 Quintiles of total vitamin A (food and supplements) (μg RE/day): Q1 (ref): < 5055 Q2: 5055–5824 Q3: 5825–6550 Q4: 6551–7507 Q5: ≥ 7508 N Q1 (ref): 15,149 Q2: 15,149 Q3: 15,150 Q4: 15,149 Q5: 15,150 Nutrient intakes were calculated as the mean of the intakes at baseline and year 3 of follow‐up	Total fractures: 10,405 Hip fractures: 588 Per quintile: Hip fractures Preformed vitamin A Q1: 112 Q2: 129 Q3: 94 Q4: 124 Q5: 129 Total vitamin A Q1: 122 Q2: 121 Q3: 113 Q4: 113 Q5: 119 Total fractures Preformed vitamin A Q1: 1977 Q2: 2044 Q3: 2110 Q4: 2152 Q5: 2122 Total vitamin A Q1: 1993 Q2: 2054 Q3: 2102 Q4: 2137 Q5: 2119 Incidence of total fractures: 221/10,000 person‐years	Model 1: Adjusted for age Model 2: Adjusted for age, energy, vitamin K, protein, alcohol, and caffeine intake; smoking; BMI; hormone therapy use; total METs per week; ethnic group; region Model 3: Model 2 plus vitamin D and calcium	HR (95% CI) Hip fractures: Preformed vitamin A Model 1: Q1: 1 Q2: 1.22 (0.95, 1.57) Q3: 0.87 (0.66, 1.15) Q4: 1.13 (0.87, 1.46) Q5: 1.21 (0.94, 1.55) p for trend: 0.373 Model 2: Q1: 1 Q2: 1.23 (0.94, 1.60) Q3: 0.89 (0.67, 1.19) Q4: 1.10 (0.84, 1.45) Q5: 1.25 (0.95, 1.64) p for trend: 0.415 Model 3: Q1: 1 Q2: 1.19 (0.91, 1.57) Q3: 0.84 (0.61, 1.14) Q4: 1.00 (0.73, 1.39) Q5: 1.13 (0.81, 1.59) p for trend: 0.925 Total vitamin A Model 1: Q1: 1 Q2: 0.99 (0.77, 1.27) Q3: 0.94 (0.73, 1.22) Q4: 0.96 (0.74, 1.23) Q5: 1.06 (0.83, 1.37) p for trend: 0.612 Model 2: Q1: 1 Q2: 1.01 (0.78, 1.32) Q3: 0.99 (0.75, 1.31) Q4: 1.03 (0.76, 1.38) Q5: 1.24 (0.88, 1.73) p for trend: 0.2
	Calcium intake (mg/day) per quintile of vitamin A intake: Q1: 847 ± 429 Q2: 1081 ± 607 Q3: 1240 ± 498 Q4: 1390 ± 462 Q5: 1622 ± 547 Alcohol intake (g/day) per quintile of vitamin A intake: Q1: 4.1 ± 7.9 Q2: 5.2 ± 9.2 Q3: 5.9 ± 10.2 Q4: 6.4 ± 10.9 Q5: 6.8 ± 12.5% current smokers per quintile of vitamin A intake: Q1: 6.8 Q2: 6.1 Q3: 5.0 Q4: 4.8 Q5: 4.9 Physical activity (METs/week) per quintile of vitamin A intake: Q1: 12.9 ± 4.1 Q2: 13.8 ± 14.3 Q3: 14.5 ± 14.3 Q4: 14.7 ± 14.4 Q5: 14.8 ± 14.8% current HRT users, per quintile of vitamin A intake: Q1: 41 Q2: 45 Q3: 47 Q4: 49 Q5: 48		Exposure assessment: 122‐item FFQ at baseline and year 3. Intakes were averaged across the two FFQs. Supplement intake was assessed by a computerised inventory and by participants bringing supplements to the clinic for an in‐person interview. Conversion factors for carotenoids not reported			Model 3: Q1: 1 Q2: 0.99 (0.75, 1.30) Q3: 0.94 (0.69, 1.28) Q4: 0.96 (0.68, 1.35) Q5: 1.14 (0.76, 1.71) p for trend: 0.445 Total fractures: Preformed vitamin A Model 1: Q1: 1 Q2: 1.05 (0.99, 1.12) Q3: 1.09 (1.03, 1.16) Q4: 1.13 (1.06, 1.20) Q5: 1.10 (1.03, 1.17) p for trend: < 0.001 Model 2: Q1: 1 Q2: 1.03 (0.96, 1.11) Q3: 1.08 (1.00, 1.15) Q4: 1.11 (1.04, 1.18) Q5: 1.08 (1.01, 1.16) p for trend: 0.002 Model 3: Q1: 1 Q2: 1.01 (0.94, 1.08) Q3: 1.03 (0.96, 1.11) Q4: 1.04 (0.96, 1.13) Q5: 1.00 (0.92, 1.10) p for trend: 0.8 Total vitamin A Model 1: Q1: 1 (ref) Q2: 1.03 (0.97, 1.10) Q3: 1.07 (1.00, 1.14) Q4: 1.09 (1.02, 1.16) Q5: 1.09 (1.02, 1.16) p for trend: < 0.001
						Model 2: Q1: 1 Q2: 1.02 (0.96, 1.09) Q3: 1.06 (0.99, 1.13) Q4: 1.08 (1.01, 1.16) Q5: 1.09 (1.01, 1.19) p for trend: < 0.001 Model 3: Q1: 1 Q2: 0.99 (0.92, 1.05) Q3: 1.00 (0.93, 1.07) Q4: 1.00 (0.92, 1.08) Q5: 0.98 (0.89, 1.08) p for trend: 0.330 ‘Among the females in the lower vitamin D strata, there was a modest risk of fractures in the highest quintile of both total vitamin A (HR: 1.19, 95% CI: 1.04, 1.37; P for trend <0.05) and retinol (HR: 1.15; 95% CI: 1.03, 1.29; P for trend: 0.056), compared with the lowest quintile. There were no significant risks in the groups with higher vitamin D intakes or lower and higher calcium intakes. The combination of lower vitamin D and calcium intakes resulted in an HR of 1.17 (95% CI: 1.01, 1.36; P for trend <0.05) for total fractures among females in the highest compared with the lowest quintile of retinol.’
Rejnmark et al. (2004) Danish Osteoporosis Prevention Study Denmark PC/NCC 5 years Funding: Mixed	N = 2016 Population sampled: Perimenopausal females, participating in an open‐label trial, aged 45–58 years, Caucasian, 3–24 months after the last menstrual bleeding or having experienced perimenopausal symptoms and having elevated serum FSH Exclusion criteria: Known metabolic bone disease, osteoporosis (defined as nontraumatic vertebral fractures), current oestrogen use, ever treatment with glucocorticoids for ≥ 6 months, current or past malignancy, newly diagnosed or uncontrolled chronic disease, hospitalisation due to alcohol or drug addiction	NCC Fractures: Self‐reported validated against hospital discharge records Some cases were added upon spinal x‐ray review A fracture was defined as more than 20% reduction in the height of a vertebra compared with the highest vertical distance of that vertebrae PC Femoral neck and lumbar spine BMD measurements were performed by using DXA	NCC – Fractures analyses Preformed vitamin A intake (food and supplements) Median (IQR), μg RE/day For cases (n = 163) 1190 (700–1420) For controls (n = 978) 1210 (740–1430) Categories of Preformed vitamin A intake (food and supplements), μg RE/day C1: < 500 (ref) C2: 500–1500 C3: > 1500 n per category C1: 175 C2: 707 C3: 219 Quintile of preformed vitamin A intakes or n per quintile = NR PC – BMD analyses Median (IQR), μg RE/day Total vitamin A intake (food and supplements): 1740 (1290‐2360)	163 fractures Cases per category of intake: C1 (ref): 24 C2: 107 C3: 28 Cases per quintile of intake: NR	NCC – Fractures Model 1: Crude Model 2: Adjusted for age, years postmenopausal, previous fracture, body weight, physical activity, total energy intake, dietary calcium intake, dietary vitamin D intake, use of vitamin D supplements, alcohol intake, smoking, thiazide diuretics, loop diruetics, thyroid hormones, antipsychotic/anxiolytic/antidepressant therapy, diagnosis of thyretoxicosis, chronic obstructive lung disease, lumbar spine and femoral neck BMD PC ‐ BMD Adjusted for age, years postmenopausal, hormone replacement therapy status, previous fracture, body weight, baseline lumbar spine and femoral neck BMD, physical activity, total energy intake, dietary calcium intake, dietary vitamin D intake, use of vitamin D supplements, alcohol intake, smoking, use of thiazide diuretics, loop diruetics, thyroid hormones or antipsychotic/anxiolytic/antidepressant therapy, diagnosis of thyretoxicosis, non‐insulin dependent diabetes mellitus, insulin‐dependent diabetes mellitus, chronic obstructive lung disease, the other dietary vitamin A intake variables	NCC – Fractures Logistic regression for preformed vitamin A and fractures, OR (95% CI) Categories of preformed vitamin A intake (food and supplements) Model 1 C1: 1 (ref) C2: 1.12 (0.70, 1.81) C3: 0.96 (0.71, 1.28) Model 2 C1: 1 (ref) C2: 1.27 (0.65, 2.51) C3: 1.03 (0.56, 1.89) Quintiles of preformed vitamin A intake (food and supplements) Model 1 Q1: 1 (ref) Q2: 1.37 (0.81, 2.32) Q3: 1.04 (0.79, 1.36) Q4: 1.05 (0.88, 1.25) Q5: 0.99 (0.86, 1.14) Model 2 Q1: 1 (ref) Q2: 1.79 (0.86, 3.75) Q3: 1.05 (0.78, 1.42) Q4: 0.94 (0.43, 2.07) Q5: 1.00 (0.75, 1.35) PC ‐ BMD Regression models for total and preformed vitamin A intake with lumbar spine and femoral neck BMD, (β (95% CI))
	% lost to follow up: NR n = 1690 (longitudinal analysis on lumbar spine BMD) n = 1694 (longitudinal analyses on femoral neck BMD). n = 1141 NCC Sex: Females Age, years (median (IQR)) PC: 50 (48–52) NCC: Cases 50 (48–52) Controls 50 (48–52) Other characteristics at baseline, Median (IQR) PC: Years postmenopausal: 0.5 (0.2–1.5) Body weight, kg: 65.7 (59.8–74) Physical activity, h/week: 19 (9–30) Dietary calcium intake, mg/day: 807 (626–1040) Dietary vitamin D intake, μg/day: 2.2 (1.6–3.2) Dietary β‐carotene intake, mg/day: 2.94 (1.44–5.50)		Preformed vitamin A intake (food and supplements): 1210 (680–1450) n/Person‐years: NR Exposure assessment: 4‐ or 7‐day food records. A dietician used food models and photographs during a 15‐min validation interview to evaluate serving sizes and cooking habits. Conversion factor for β‐carotene: 6:1			Change in BMD at the lumbar spine (g/cm 2 ) per total vitamin A intake (food and supplements) (mg RE/day) increase: 0.043 (−0.193, 0.284) Change in BMD at the lumbar spine (g/cm 2 ) per total preformed vitamin A intake (food and supplements), (mg RE/day) increase: 0.101 (−0.180, 0.390) Change in BMD at the femoral neck (g/cm 2 ) per total vitamin A intake (food and supplements) (mg RE/day) increase: −0.122 (−0.349, 0.105) Change in BMD at the femoral neck (g/cm 2 ) per total preformed vitamin A intake (food and supplements) (mg RE/day) increase: −0.065 (−0.340, 0.209), p = NR
Barker et al. (2005) UK NCC 3.7 years (mean) Funding: Public	N = 2602 Population sampled: Cases and controls sampled from the placebo arm of a study, general population of females > 75 years Exclusion criteria: bilateral hip arthroplasty, hypocalcaemia, leucopoenia, impaired hepatic function, known malignancy, inflammatory bowel disease, impaired renal function, and use of bisphosphonate or calcium supplement of > 500 mg/day % lost to follow up: NR n = 1611 Any osteoporotic fracture Cases: 312 Controls: 934 Hip fracture Cases: 92, Controls: 273 Sex: Females Age, years, mean (95% CI) Any osteoporotic fracture Cases: 80.1 (79.6–80.6) Controls 79.3 (79.1–79.5) Hip fracture Cases: 80.8 (80.0–81.7) Controls 79.2 (78.7–80.0)	Hip and any fractures: self‐reports (house visits), medical records with radiology or surgical report confirmation. Verification in a subsample by comparing ICD codes	Serum retinyl palmitate (upper quartile vs. rest of population). Concentrations not reported. Measured by HPLC‐MS/MS Retinol supplement use. Concentrations or contrast in analysis not reported. Intake estimated from multivitamin/cod liver oil use Exposure assessment method: Serum retinyl palmitate and surveys for multivitamin/cod liver oil use which was in turn used to estimate retinol supplement use	Any osteoporotic fracture: 312 Hip fracture: 92	Models were unadjusted Retinyl palmitate and retinol supplementation was only evaluated in univariate Cox PH models with fracture as outcome. Because they did not satisfy the authors' criterion of p < 0.1 on fracture risk they were not included in multivariable models. The final model only included age, total hip BMD and weight and no dietary variables	Serum retinyl palmitate HR (95% CI) Any fracture: Q1‐3: ref. Q4: 0.97 (0.74–1.26) p = 0.800 Hip fracture: Q1‐3: ref. Q4: 0.91 (0.56–1.46) p = 0.687 Preformed vitamin A supplement use HR (95% CI) Any fracture: Non‐users: ref. Users: 0.76 (0.60–0.95) p = 0.021 Hip fracture: Non‐users: ref. Users: 0.86 (0.56–1.33) p = 0.507 In the multivariate step‐wise analysis, serum retinyl palmitate, and retinol supplement use, was not associated with fracture risk
	Other characteristics, mean (95% CI) Any osteoporotic fracture cases Height, cm: 156 (155–157) Weight, kg: 61.7 (60.6–62.9) Total hip BMD, g/cm2: 0.68 (0.66–0.69) Serum 25(OH)D, nM [geometric mean (95% CI)]: 40.1 (38.3–42) Any osteoporotic fracture controls Height, cm: 156 (156–157) Weight, kg: 65.4 (64.6–66.1) Total hip BMD, g/cm2: 0.76 (0.76–0.77) Serum 25(OH)D, nM [geometric mean (95% CI)]: 41.9 (40.8–43.0) Hip fracture cases Height, cm: 155 (154–156) Weight, kg: 58.8 (56.8–60.7) Total hip BMD, g/cm2: 0.65 (0.62–0.68) Serum 25(OH)D, nM [geometric mean (95% CI)]: 37.5 (34.5–40.7) Controls Height, cm: 157 (156–157) Weight, kg: 65.7 (64.1–67.1) Total hip BMD, g/cm2: 0.76 (0.74–0.78) Serum 25(OH)D, nM [geometric mean (95% CI)]: 39.6 (37.7–41.7)
de Jonge et al. (2015) The Rotterdam study The Netherlands PC 13.9 years (mean) Funding: Mixed	N: 7983 Population sampled: Aged 55 years and older from the general population Exclusion criteria: No dietary intake data/unreliable data, missing BMD data Lost to follow up: NR n = 5288 Sex: 59% female Age: ≥ 55 years Other characteristics (median (IQR) per quintile of vitamin A intake) Vitamin D intake, μg/day: Q1: 3.62 (2.68–4.61) Q2: 3.29 (2.36–4.45) Q3: 3.01 (2.18–4.12) Q4: 3.02 (2.20–4.15) Q5: 3.16 (2.25–4.32) Calcium intake, mg/day: Q1: 1009 (804–1279) Q2: 1046 (857–1282) Q3: 1097 (878–1341) Q4: 1097 (894–1340) Q5: 1130 (872–1374) Height, cm: Q1: 168 (161–174) Q2: 167 (161–175) Q3: 167 (161–174) Q4: 166 (161–173) Q5: 166 (160–174)	Radius/ulna, tibia/fibula, hand/foot, femur and other fractures: Reported by general practitioners; verified by two research physicians Femoral neck BMD: By DXA, assessed at four visits	Quintiles of energy‐adjusted total vitamin A intake (food only), μg RE/day Median Q1: 709 Q2: 939 Q3: 1124 Q4: 1384 Q5: 2012 Energy‐adjusted preformed vitamin A intake, median μg RE/day: Q1: 186 Q2: 200 Q3: 294 Q4: 518 Q5: 1099 N/person‐years: NR (author requested) Exposure assessment: 170‐item, validated, semi‐quantitative FFQ through interviews at baseline ‘Energy‐adjusted’ nutrient intakes = unstandardised residuals from linear regression Conversion factor for β‐carotene 6:1	Fractures: 1301 Cases per quintile: Total vitamin A, energy‐adjusted: Q1: 258 Q2: 234 Q3: 291 Q4: 279 Q5: 239 Retinol, energy‐adjusted Q1: 228 Q2: 263 Q3: 307 Q4: 269 Q5: 234	Model 1: Age, sex Model 2: Model 1 + smoking, dietary calcium, alcohol intake, education, net income, disability index, physical activity Model 3: Model 2 + BMI Model 4: Model 3 + Baseline BMD	Total fractures: HR (95% CI) Total vitamin A: Model 3: Q5 vs. Q3: 0.82 (0.69, 0.97) p for trend: NR Retinol: Model 3: Q5 vs. Q3: 0.81 (0.68, 0.96) p for trend: NR Model 4: Not significant (data not shown) Significant lower fracture risk in subjects in the highest quintile of retinol intake only in those with a BMI ≥ 25 kg/m2 (HR (95% CI) = 0.78 (0.68–0.89) versus 1.04 (0.87–1.24) with BMI ⩽25 kg/m2) No significant interaction between total vitamin A, retinol or β‐carotene and vitamin D intake (p for all interactions > 0.45) on fracture risk. No significant interactions between dietary intake of vitamin A and vitamin D plasma concentrations BMD (g/cm 2 ) Baseline (median (IQR)): 0.86 (0.77–0.96) At follow‐up: Per 100 μg/day RE total vitamin A (β (95% CI): Model 1: 0.53 (0.06–0.99)
	BMI, kg/m2: Q1: 25.6 (23.7–27.8) Q2: 25.9 (23.8–28.1) Q3: 25.9 (23.8–28.4) Q4: 26.0 (24.2–28.5) Q5: 26.4 (24.2–29.1) Current smokers, %: Q1: 25 Q2: 23 Q3: 22 Q4: 20 Q5: 25 Alcohol intake, g/day: Q1: 3.5 (0.1–16.9) Q2: 3.2 (0.2–14.4) Q3: 3.1 (0.2–14.4) Q4: 3.5 (0.2–14.8) Q5: 3.9 (0.2–15.0) Physical activity, h/day: Q1: 5.8 (4.2–7.5) Q2: 5.8 (4.1–7.8) Q3: 5.9 (4.2–7.8) Q4: 5.8 (4.3–7.9) Q5: 5.9 (4.2–7.9) Prevalent osteoporosis, % Q1: 11 Q2: 10 Q3: 10 Q4: 10 Q5: 11 Current or past HRT use, %: Q1: 13 Q2: 13 Q3: 13 Q4: 17 Q5: 12					Model 2: 0.46 (0.00–0.91) Model 3: 0.14 (−0.28–0.56) (model 3 n.s.) Per 100 μg/day retinol: Model 1: 0.31 (−0.23, 0.87) Model 2: 0.45 (−0.09, 1.01) Model 3: 0.13 (−0.40, 0.75) (model 3 n.s.) Significant interaction between dietary intake of vitamin D and total vitamin A (p for interaction = 0.016) in relation to BMD. However, stratified analysis for dietary vitamin D intake above or below the median did not show significant associations between total vitamin A and BMD. No significant interactions between dietary intake of vitamin A and vitamin D plasma concentrations
Toraishi et al. (2021) Japan PC 1 year Funding: Public	N = 41 Population sampled: Collegiate distance runners Exclusion criteria: NR % lost to follow up: 0 n = 41 Sex: Males Age, years: 19.4 ± 1.0 Other characteristics BMI, kg/m2: 19.3 ± 1.2 Baseline vitamin D intake, μg/day: Overall: 9.6 ± 2.2 Baseline calcium intake, mg/day: Overall: 515.1 ± 175.8 Stress‐fracture group: 684 ± 320 Non‐stress‐fracture group: 497 ± 150	Stress fractures at any site, i.e. due to repeated mechanical load: Self‐reported	Mean intake of total vitamin A (food only), μg RAE/day All subjects: 1441.3 ± 802.4 Stress‐fracture group: 2792 ± 1136 No stress‐fracture group: 1295 ± 619 Mean intake at follow‐up: Stress‐fracture group: 3747 ± 309 No stress‐fracture group: 2943 ± 1204 Exposure assessment: Semi‐quantitative FFQ Conversion factors for carotenoids not reported	N = 4 (3 at tibia and one at metatarsus)	NR In logistic regression: Calcium and iron intake	OR (95% CI) for stress fracture (at any site) Per 100 μg RAE increase: 1.22 (0.99–1.50) Threshold intake for stress fracture (from ROC): 3206 μg
Kaptoge et al. (2003) EPIC‐Norfolk UK PC 3 years (mean; range 2–5 years) Funding: Public	N = 30,411 Population sampled: Elderly males and females Exclusion criteria: < 2 BMD scans, treatment with bone active medication, different side of the hip scanned at follow‐up, incomplete food diaries % lost to follow up: NR n = 944 (470 males, 474 females) Sex: 50.2% female Mean (5th, 95th percentile) age, years Males: 72.0 (68.0, 77.4) Females: 71.9 (67.9, 77.0)	Total hip BMD: DXA	Preformed vitamin A intake (food only), μg RE/day: Mean (5th, 95th percentile) Males: 358 (109, 3836) Females: 289 (98, 3517) Analyses are also reported by tertiles but intakes are not reported T1: NR T2: NR T3: NR	NA	Continuous analyses Model 1: Weight change, FEV, Stair climbing, Activities in daily living (ADL) score change, Past activity score Categorical analyses Estimated mean BMD loss (% per annum) Males Model 1: Rate of weight change and FEV Females Model 1: Rate of weight change, stair climbing, change in ADL score and past activity	Continuous analyses on rate of total hip BMD percentage loss per annum (β coefficient (SE)) Model 1 Vitamin A (/1000 μg): Males: 0.029 (0.043) p = 0.508 Females: −0.024 (0.039) p = 0.539 Categorical analyses on total hip BMD percentage loss per annum (mean rates) Males Model 1 T1: −0.11 T2: −0.14 T3: −0.21 p for trend = 0.773
	Other characteristics Mean (5th, 95th percentile) Males: BMI, kg/m2: 27.0 (21.9, 32.3) Past physical activity, z‐score: 0.31 (−1.18, 2.37) Grip strength, kg: 39 (27, 52) Weight loss, kg/year: −0.06 (−2.30, 2.10) Ever fractured (% yes): 39 Vitamin D intake, μg/day: 3.4 (1.1,8.0) Calcium intake, mg/day: 886 (505, 1364) Carotene intake μg/day: 1871 (556, 4664) Females: BMI, kg/m2: 26.8 (21.0, 35.0) Past physical activity, z‐score: −0.21(−1.77, 1.19) Grip strength, kg: 24 (15, 32) Weight loss, kg/year: −1.7 (−7.5, 3.5) Ever fractured (% yes): 34 Vitamin D intake, μg/day: 2.5 (0.9,6.0) Calcium intake, mg/day: 755 (399, 1188) Carotene intake μg/day: 1658 (458, 3885)		The nutrient intakes for categorical analyses were adjusted for total energy intake by taking the residuals from regressing each nutrient on total energy Exposure assessment 7‐day food diary with a 24‐h recall the first day. The food diary has been validated against biomarkers and 16‐day weighted food records. Conversion factors for carotenoids not reported			Females Model 1 T1: −0.42 T2: −0.49 T3: −0.33 p for trend = 0.517
Promislow et al. (2002) The Rancho Bernardo Study USA PC Mean ± SD follow‐up time: 4.1 (0.5) years Funding: Public	N = 1526 Population sampled: Ambulatory community‐dwelling elderly Exclusion criteria: NR % lost to follow up: 37.2% n = 958 females = 570 males = 388 Sex: 59.5% female Age, years Females: 71.3 ± 8.7 Males: 70 ± 8.5 Other characteristics Mean ± SD Females BMI, kg/m2: 24.6 ± 3.7 Total calcium intake, mg/day: 984 ± 573.8 Oestrogen users, %: 39.7 Vitamin D: NR Males BMI, kg/m2: 26.4 ± 3.4 Total calcium intake, mg/day: 797.2 ± 458.2 Vitamin D: NR	Total hip, femoral neck and lumbar spine BMD: DXA Total hip BMD was obtained by summing the bone mineral content at the femoral neck, intertrochanter, and greater trochanter and dividing by the composite area of the three sites. Spine BMD was defined as the average BMD of lumbar vertebrae L1–L4. Instruments were calibrated daily and had measurement precisions of ≤ 1% for the spine and ≤ 1.5% for the Hip	Preformed vitamin A (food and supplements), μg RE/day Mean (SD) Females: 1247 (1573) Males: 1242 (1442) Preformed vitamin A (food only), μg RE/day Mean (SD) Females: 497 (460) Males: 624 (585) Supplement users, n Females: 281 Males: 150 Supplement use defined as taking either a multivitamin or a retinol supplement Exposure assessment 61‐item FFQ validated against four one‐week food records. Conversion factors for carotenoids not reported	NA	Model 1: Adjusted for age Model 2: Adjusted for age, BMI, calcium intake (including supplements), diabetes status; years postmenopausal (females only), current exercise, and current use of cigarettes, alcohol, thiazides, thyroid hormones, steroids, supplemental retinol, and oestrogen, percent change in body weight	Change (%) in BMD per unit increase in energy‐adjusted log preformed vitamin A intake (retinol) (β (95% CI)) Total hip BMD Females Model 1 All: −0.05 (−0.016, 0.07), p = 0.43 Supplement users: −0.28 (−0.5, −0.06), p = 0.01 Non‐users: 0.10 (−0.12, 0.33), p = 0.36 Model 2 All: NR Supplement users: −0.27 (−0.48, −0.04), p = 0.02 Non‐users: 0.13 (−0.09, 0.35), p = 0.25 Males Model 1 All: −0.01 (−0.13 0.10), p = 0.84 Supplement users: −0.15 (−0.41, −0.10), p = 0.24 Non‐users: −0.06 (−0.24, 0.13), p = 0.55 Model 2 All: −0.08 (−0.23, 0.08), p = 0.32 Supplement users: −0.19 (−0.46, 0.08), p = 0.17 Non‐users: −0.04 (−0.23, 0.15), p = 0.68 Femoral neck BMD Females Model 1 All: −0.04 (−0.019, 0.10), p = 0.56 Supplement users: −0.21 (−0.44, −0.02), p = 0.07 Non‐users: 0.13 (−0.20, 0.46), p = 0.44 Model 2 All: NR Supplement users: −0.23 (−0.46, 0.00), p = 0.05 Non‐users: 0.22 (−0.11, 0.56), p = 0.19
						Males Model 1 All: 0.01 (−0.13, 0.15), p = 0.87 Supplement users: −0.10 (−0.37, −0.17), p = 0.45 Non‐users: 0.01 (−0.24, 0.26), p = 0.95 Model 2 All: 0.06 (−0.14,0.25), p = 0.57 Supplement users: −0.15 (−0.42, 0.12), p = 0.28 Non‐users: 0.16 (−0.12, 0.44), p = 0.25 Total spine Females Model 1 All: −0.03 (−0.13, 0.07), p = 0.50 Supplement users: −0.10 (−0.29, 0.08), p = 0.27 Non‐users: 0.04 (−0.15, 0.23), p = 0.67 Model 2 All: NR Supplement users: −0.10 (−0.28, 0.09), p = 0.30 Non‐users: 0.01 (−0.17, 0.20), p = 0.89 Males Model 1 All: 0.00 (−0.12, 0.12), p = 0.97 Supplement users: −0.05 (−0.28, 0.17), p = 0.64 Non‐users: −0.06 (−0.27, 0.14), p = 0.95 Model 2 All: 0.01 (−0.15,0.16), p = 0.92 Supplement users: 0.00 (−0.24, 0.24), p = 0.98 Non‐users: 0.02 (−0.19, 0.23), p = 0.87 ‘Regression models showed an analogous inverse U‐shaped association of retinol intake with percent bone change, suggesting that those with low or high retinol intakes suffered greater bone loss, although statistical evidence for this pattern was weak in men’
Macdonald et al. (2004) Aberdeen Prospective Osteoporosis Screening Study UK Up to 7 years PC Funding: Public	N = 5119 Population sampled: Premenopausal women Exclusion criteria: Chronic medication use, conditions likely to affect bone metabolism, bisphosphonate therapy, wheelchair use, outlying dietary calcium intake % lost to follow up: 17.3 n = 891 Sex: Females Age, years 57.5 ± 1.5 Other characteristics BMI, kg/m2: 24.6 ± 4.0 BMD, g/cm2 Lumbar spine: 1.064 ± 0.16 Femoral neck: 0.886 ± 0.13 Total calcium intake, mg/day 1070 ± 334 Total protein intake, g/day 81.4 ± 22.5 Total vitamin D intake, μg/day 4.5 ± 3.1 Liver enzymes: NR	Femoral neck and lumbar spine BMD: DXA	Preformed vitamin A intake (food only), μg RE/day: Mean ± SD: 820 ± 602 Median (range): 588 (39–4354) Preformed vitamin A intake (food and supplements), μg RE/day: Mean ± SD: 924 ± 666 Median (range): 702 (85–4354) Total vitamin A intake NR Exposure assessment 98‐item semi‐quantitative FFQ validated against weighted food records. Results reported using 6:1 conversion	NA	Model 1: Unadjusted Model 2: Adjusted for age, weight, annual percentage change in weight, height, smoking status, socioeconomic status, physical activity level, baseline BMD at appropriate site, menopausal status and hormonal replacement therapy use Multivariable regression model for dietary retinol and vitamin A and femoral neck BMD: Baseline femoral neck BMD, age, annual percentage weight change, height, hormone replacement therapy use, menopausal status Multivariable regression model for lumbar spine BMD: Vitamin A was not part of this model. It included weight, age, hormone replacement therapy use, menopausal status and alcohol intake.	Pearson's correlation analyses Energy‐adjusted vitamin A and change in femoral neck BMD Model 1: Preformed vitamin A (food only): r = −0.072, p < 0.05 Preformed vitamin A (food and supplements): r = −0.071, p < 0.05 Total vitamin A (food only): r = −0.090, p < 0.01 Total vitamin A (food and supplements): r = −0.004, p = NS (value NR) Model 2 Preformed vitamin A (food only): r = −0.067, p < 0.05 Preformed vitamin A (food and supplements): r = −0.032, p = NS (value NR) Total vitamin A (food only): r = −0.090, p < 0.01 Total vitamin A (food and supplements): r = −0.012, p = NS (value NR) Energy‐adjusted vitamin A and change in lumbar spine BMD Model 1: Preformed vitamin A (food only): r = −0.021, p = NS (value NR) Preformed vitamin A (food and supplements): r = −0.019, p = NS (value NR) Total vitamin A (food only): r = −0.041, p = NS (value NR) Total vitamin A (food and supplements): r = −0.029, p = NS (value NR)
						Model 2 Preformed vitamin A (food only): r = −0.036, p = NS (value NR) Preformed vitamin A (food and supplements): r = −0.004, p = NS (value NR) Total vitamin A (food only): r = −0.061, p < 0.08 Total vitamin A (food and supplements): r = −0.032, p = NS (value NR) Multivariable regression model Femoral neck BMD, β coefficient (95% CI) ‐ Standardised to 8 MJ energy intake Preformed vitamin A (food only) (mg × 10−4): −1.73 (−3.20, −0.30), p = 0.018 Total vitamin A (food only) (mg × 10−4): −1.24 (−2.47, 0.17), p = 0.047 Including the nutrient intake from dietary supplements, the relation between retinol and vitamin A and FN BMD change was no longer significant Lumbar spine BMD Of dietary factors, only alcohol intake was significantly associated with lumbar spine BMD: 0.0893 (0.034, 0.145) p = 0.002
Chan et al. (2011) China PC 4 years Funding: Public	N = 4000 Population sampled: Elderly (≥ 65 years) living in the community Exclusion criteria: Detectable disease or medication known to affect bone mass; incomplete dietary data; extreme energy intake	Total hip and femoral neck BMD: measured by DEXA	Total Vitamin A intake (food only) at baseline, μg RE/day Median (IQR) Males: 940 (667–1315) Females: 939 (676–1277)	NA	Nutrient intakes were adjusted for dietary energy intake by the residual method Model 1: Adjusted for age, baseline weight, baseline height, % change in body weight, education, current drinker, current smoker, use of calcium supplements, physical activity, total energy intake	Change (%) in BMD over the 4‐year follow‐up per IU/day vitamin A increase (β) Males Total Hip: Univariate: −0.653, p = 0.116 Model 1: −0.433, p = 0.259 Model 2: 0.035, p = 0.932 Femoral neck: Univariate: −0.316, p = 0.618 Model 1: 0.068%, p = 0.914
	Lost to follow up: 25% n = 2217 Sex: 45% female Age, years: Male: 71.6 ± 4.6 Female: 72.0 ± 5.1 Other characteristics BMI, kg/m2: Male: 23.5 ± 3.1 Female: 24.0 ± 3.5 Height, cm: Male: 163.2 ± 5.6 Female: 151.2 ± 5.3 Physical activity score, PASE: Male: 101.7 ± 51.3 Female: 87.6 ± 33.9 Baseline BMD, g/cm2: Hip: Males: 0.875 ± 0.122 Females: 0.725 ± 0.114 Femoral neck: Males: 0.696 ± 0.106 Females: 0.594 ± 0.098 Current smoker, % Males: 11.9 Females: 1.8 Current drinker, % Males: 25.2 Females: 2.9 Vitamin D intake (IU/day), median (IQR): Male: 8.2 (2.9–16.6) [0.2 (0.07–0.42 μg] Female: 7.2 (3.1–15.0) [0.2 (0.08–0.38) μg] Calcium intake (mg/day) Male: 638.8 ± 294.4 Female: 571.8 ± 260.0		Exposure assessment: Semi‐quantitative FFQ including 13 food groups at baseline, validated with basal metabolic rate calculation and 24‐h sodium/creatinine and potassium/creatinine Conversion factors for carotenoids not reported		Model 2: As model 1 + energy‐adjusted calcium and vitamin D intake	Model 2: 0.742%, p = 0.274 Females No associations (data not shown)
Houtkooper et al. (1995) USA PC 12–18 months Funding: Mixed	N = 66 Population sampled: Pre‐menopausal women who participated in a RCT on resistance exercise training (and were taking 500 mg of calcium supplements per day). 27 participated in resistance training, rest were sedentary Exclusion criteria: Pregnancy, lactation, < 10 normal menses in previous year, contraceptive use, medication affecting bone metabolism for two years prior to study, smoking > 10 cigarettes per day, regular exercise > 6 months during the 5 years preceding enrolment, weight fluctuations > 3 kg the previous year excluding pregnancy, had intentions of changing weight status by this magnitude within the next year, BMI < 5th or >95th percentile (NHANES 1976–80), history of anorexia, bulimia, cancer, diabetes, thyroid disease or myocardial infarction % lost to follow up: 15.2% for 18 month visit n = 66 at 12 months, 56 at 18 months	Total body, lumbar spine, femoral neck, trochanter and Ward's triangle BMD: DXA at four timepoints: Baseline and at 5, 12 and 18 months	Preformed vitamin A intake (food only), μg RE/day Mean ± SD 1220 ± 472 Exposure assessment: Repeated 4‐day records before 5–12‐ and 18‐months testing periods	NA	Total body BMD: Adjusted for exercise group in original study, baseline fat mass, fat mass annual rate of change and baseline total body BMD	Total body BMD annual rate of change (β, no CI reported) 0.007, p = 0.002 Vitamin A was only included in the model assessing total BMD. Vitamin A was not a significant variables in regression models predicting bone mineral density slopes (rates of change) at any femur site or at the lumbar spine
	Sex: Females Age, years: 34.4 ± 2.7 Other characteristics: Height, cm: 165.4 ± 6.4 Weight, kg: 59.9 ± 9.8 BMD, g/cm 3 Total body: 1.15 ± 0.08 Lumbar spine: 1.22 ± 0.13 Femoral neck: 0.96 ± 0.13 Trochanter: 0.76 ± 0.10 Ward's triangle: 0.89 ± 0.15 Total calcium intake (diet + supplement), mg/day: 1326 ± 232 Vitamin D: NR
Sugiura et al. (2016) Mikkabi Cohort Study Japan PC 4 years Funding: Public	N = 457 Population sampled: post‐menopausal women with T‐scores > 70% at baseline Exclusion criteria: those with T‐scores < 70% % lost to follow up: NR n = 187 Sex: females Age, years: 59–61 (mean range across tertiles) Other characteristics: BMD (g/cm2) Mean (SD) T1: 0.568 (0.068) T2: 0.586 (0.077) T3: 0.562 (0.075) T‐score (%) Mean (SD) T1: 88.0 (10.5) T2: 90.8 (11.9) T3: 87.0 11.6)	Osteroporosis was defined as T‐score less than 70%. Radial BMD at baseline and follow‐up survey using DXA	Tertiles of preformed vitamin A intake (from food only), μg RE/day Means (Range) T1(ref): 138 (29–199) T2: 265 (200–349) T3: 538 (351–2320) n per tertile T1: 62 T2: 62 T3: 63 Exposure assessment: Semi‐quantitative FFQ at baseline	Cases of osteoporosis n, (%) T1: 6 (9.7) T2: 3 (4.8) T3: 8 (12.7)	Model 1: Age, weight, height, years since menopause, current tobacco use, regular alcohol intake, exercise habits, supplement use, and total energy intake Model 2: Model 1 + intakes of calcium, magnesium, potassium, and vitamin D	OR (95% CI) for osteoporosis per tertile of preformed vitamin A Model 1: T1(ref): 1 T2: 0.57 (0.12, 2.78) T3: 1.49 (0.36, 6.22) Model 2: T1(ref): 1 T2: 0.34 (0.06, 1.84) T3: 0.91 (0.19, 4.29)

Abbreviations: ADL, activities in daily living; BMD, bone mineral density; BMI, Body mass index; CC, case‐cohort; DXA, dual X‐ray absorptiometry; FEV, forced expiratory volume; FFQ, food frequency questionnaire; HRT, hormone replacement therapy; IU, international units; NA, not available; NR, Not reported; NCC, nested case–control; PC, prospective cohort; RAE, retinol activity equivalents; RE, retinol equivalents.

Note: Unless otherwise reported, values are mean ± SD. Preformed vitamin A refers to retinol intake, total vitamin A refers to retinol and β‐carotene intakes.

^a n = 2032 in the full cohort for which baseline covariates are presented, n = 1221 for dietary records.

^b As per study author likely to be preformed vitamin A.