Introduction
Methods
Literature search
Literature selection
Data extraction
Approaches to assess the quality and certainty of evidence
Methodological quality of SRs
Outcome-specific certainty of evidence of SRs
Rating the overall certainty of evidence and deriving conclusions
Overall certainty of evidence | Underlying criteria | Definition/explanation |
---|---|---|
Convincing | • At least one SR with or without MA of prospective studies available • If more than one SR with or without MA are available: all overall results must be consistent1 • In case of a positive or negative association, biological plausibility is given • All included SRs with or without MA must reach at least a “moderate” outcome-specific certainty of evidence2; in addition all included SRs must reach at least a methodological quality3 of “moderate” | There is high level of confidence that the true effect lies close to that of the estimate(s) of the effect |
Probable | • At least one SR with or without MA of prospective studies available • If more than one SR with or without MA are available, the majority of overall results must be consistent.1 • In case of a positive or negative association, biological plausibility is given • The majority4 of included SRs with or without MA must have reached at least a “moderate” certainty of evidence2; in addition all included SRs must reach at least a methodological quality3 of “moderate” | There is moderate confidence in the effect estimate(s): The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different |
Possible | • At least one SR with or without MA of prospective studies available • If more than one SR with or without MA are available, the majority of overall results must be consistent.1 • In case of a positive or negative association, biological plausibility is given • The majority4 of included SRs with or without MA must reach at least a “low” certainty of evidence2; in addition the majority4 of all included SRs must reach at least a methodological quality3 of “moderate” | Confidence in the effect estimate(s) is limited: The true effect may be substantially different from the estimate of the effect |
Insufficient | • No SR is available OR • The majority4 of included SRs with or without MA reach a “very low” certainty of evidence2; in addition the majority of all included SRs reach a methodological quality3 of “low” | There is very little confidence in the effect estimate (s): The true effect is likely to be substantially different from the estimate of effect |
Results
Characteristics of the included systematic reviews
Author, year | Study type, study period | Study population | Exposition | Protein intake | Outcome | Effect estimates | Heterogeneity estimators | NutriGrade rating | AMSTAR 2 rating |
---|---|---|---|---|---|---|---|---|---|
Blair 2020 | - SR without MA of RCTs - Published before 09/2018 - Study duration: 6–18 mos | - Both sexes - Healthy people - Mean age: 50–70.5 yrs | Whey protein | Supplements: 40 mg/d–40 g/d | NA | High | |||
2 RCTs | n = 148 | BMD | 2 out of 2 RCTs reported no changes to BMD | Low: 4.5 | |||||
2 RCTs | n = 148 | Osteocalcin | 2 out of 2 RCTs reported unchanged osteocalcin | Low: 4.5 | |||||
Darling 2019 | - SR with MA of RCTs and observational studies - Published between 01/1976 and 02/2016 - Study duration: 38 d–2 yrs - Follow-up: 1–22 yrs | - Both sexes - Healthy people - Aged 20–82 yrs | Supplements in RCTs: 40 mg/d milk-based protein or 20.4–45 g/d protein or 18–40 g/d soy protein Cohort studies: total protein (mean): ≥ 46 g/d animal protein (mean): 17.6–29.0 g/d or 0.8–1.2 g/kg BW/d plant protein (mean): 12 g/d or 1.2 g/kg BW/d | Low | |||||
4 cohort studies | n = 157,046 | Total protein | Low trauma fractures | RR (95% CI), fixed effect 1.04 (0.93, 1.16) P = NP | I2 = 32% tau2 = 0.026 P = 0.22 | Very low: 2.0 | |||
4 cohort studies | n = 193,954 | Animal protein | Low trauma fractures | RR (95% CI), random effect 0.98 (0.76, 1.27) P = 0.87 | I2 = 46% tau2 = 0.031 P = 0.13 | Very low: 2.0 | |||
3 cohort studies | n = 154,167 | Plant protein | Low trauma fractures | RR (95% CI), fixed effect 0.97 (0.87, 1.09) P = 0.61 | I2 = 15% tau2 = 0.007 P = 0.31 | Very low: 2.0 | |||
2 RCTs | n = 255 | High vs low intake of total protein | Lumbar spine BMD | MD (95% CI), fixed effect 0.04 g/cm2 (0.00, 0.08) P = 0.24 | I2 = 0% tau2 = 0 P = 0.47 | Very low: 2.0 | |||
3 RCTs | n = 435 | High vs low intake of total protein | Femoral neck BMD | MD (95% CI), random effect 0.01 g/cm2 (− 0.03, 0.05) P = 0.07 | I2 = 68% tau2 = 0.001 P = 0.04 | Very low: 2.0 | |||
3 cohort studies | n = 11,112 | Total protein | Fracture risk | HR (95% CI), fixed effect 0.83 (0.67, 1.01) P = 0.59 | I2 = 35% tau2 = 0.048 P = 0.19 | Very low: 2.0 | |||
3 RCTs | n = 115 | Milk basic protein | Lumbar spine BMD | MD (95% CI), fixed effect 0.02 g/cm2 (0.00, 0.04) P = 0.08 | I2 = 0% tau2 = 0 P = 0.87 | Very low: 2.0 | |||
NP | NP | Soy protein | Lumbar spine BMD | MD (95% CI), random effect − 0.01 g/cm2 (− 0.07, 0.06) P = 0.82 | I2 = 51% P = 0.13 | Very low: (2.0) | |||
NP | NP | Soy protein | Femoral neck BMD | MD (95% CI), random effect 0.01 g/cm2 (− 0.06, 0.07) P = 0.87 | I2 = 74% P = 0.05 | Very low: 2.0 | |||
6 RCTs | n = 128 | Soy protein | BAP | MD (95% CI), random effect − 1.75 ng/ml (− 10.50, 7.01) P = 0.70 | I2 = 91% P = 0.00 | Very low: 2.7 | |||
Groenendijk 2019 | - SR with MA of cohort studies - Published before 11/2018 - Follow-up: 1–32 yrs | - Both sexes - Healthy people - Aged ≥ 50 yrs | High vs low protein | Cohort studies (mean intake) ≥ 0.93 g/kg BW/d vs < 0.95 g/kg BW/d or 19.5–29.1 E% vs 9.6–13.3 E% or 82.7 g/d vs 46.5 g/d | NA | High | |||
2 cohort studies | n = 1406 | Total body BMC | One cohort study found a significant higher total body BMC, whereas one cohort study found no association | Very low: 2.0 | |||||
4 cohort studies | n = 145,558 | Total body BMD | 3 out of 4 cohort studies found no association | Low: 4.0 | |||||
5 cohort studies | n = 153,841 | Total hip BMD | 3 out of 5 cohort studies found no association | Low: 4.0 | |||||
6 cohort studies | n = 4887 | Femoral neck BMD | Half of the cohort studies showed significantly higher femoral neck BMD and the other half showed no association | Low: 3.0 | |||||
4 cohort studies | n = 1593 | Lumbar spine BMD | 2 cohort studies found no association, whereas one cohort study found significantly higher and one cohort study found significantly lower lumbar spine BMD | Very low: 2.0 | |||||
2 cohort studies | n = 434 | Osteocalcin | Both cohort studies showed no association | Low: 3.0 | |||||
2 cohort studies | n = 434 | NTX | Both cohort studies showed no association | Low: 3.0 | |||||
MA of 4 cohort studies | n = 144,580 | Hip fracture | Weighted HR (95% CI), fixed effect 0.89 (0.84, 0.94) P < 0.001 | I2 = 0% P = 0.61 | Very low: 2.5 | ||||
Wright 2019 | - SR with MA of RCTs - Published before 01/2019 - Study duration: 3 mos–2 yrs | - Both sexes - People, which actively losing weight (BMI 31.4 ± 2.5 kg/m2) - Aged 49 ± 10 yrs | High protein diet: dietary protein ≥ 25 En% and/or ≥ 1 g/kg BW/d | 1.01–1.69 g/kg BW/d vs 0.67–0.97 g/kg BW/d | MD (95% CI), random effect | Low | |||
8 RCTs | n = 488 | Total body BMD | 0.006 g/cm2 (0, 0.012) P = 0.0163 | I2 = 69% tau2 = 0 P = 0.001 | Low: 5.65 | ||||
2 RCTs | n = 184 (only postmenopausal women) | Total hip BMD | 0 g/cm2 (− 0.026, 0.026) P = 0.497 | I2 = 0% tau2 = 0 P = 1.00 | Low: 5.5 | ||||
3 RCTs | n = 272 | Femoral neck BMD | 0 g/cm2 (− 0.016, 0.016) P = 0.50 | I2 = 0% tau2 = 0 P = 1.00 | Low: 5.75 | ||||
4 RCTs | n = 322 | Lumbar spine BMD | 0.017 g/cm2 (0.001, 0.033) P = 0.0187 | I2 = 0% tau2 = 0 P = 0.72 | low: 5.25 | ||||
7 RCTs | n = 408 | Total body BMC | 4.28 g (− 11.49, 20.04) P = 0.393 | I2 = 0% tau2 = 0 P = 0.94 | Low: 5.4 | ||||
2 RCTs | n = 135 | Femoral neck BMC | − 0.05 g (− 0.16, 0.05) P = 0.848 | I2 = 0% tau2 = 0 P = 0.78 | Low: 5.62 | ||||
2 RCTs | n = 135 | Lumbar spine BMC | 0.40 g (− 1.19, 2.00) P = 0.310 | I2 = 0% tau2 = 0 P = 0.81 | Low: 5.25 | ||||
Shams-White 2018 | - SR without MA of RCTs - Published before 11/2016 - Study duration: 6–24 mos | - Peri- or postmenopausal women - Generally healthy people - Mean age: 50.6–73.1 yrs | Supplements: 18–40 g/d | Low | |||||
2 RCTs | n = 393 | Isoflavone-poor soy protein vs animal protein | Lumbar spine BMD | Both RCTs found no significant differences in the net changes in lumbar spine BMD | NA | Moderate: 5.5 | |||
2 RCTs | n = 146 | Isoflavone-poor soy protein vs animal protein | Femoral neck BMD | Both RCTs found no significant differences in the net changes in femoral neck BMD | NA | Moderate: 5.5 | |||
2 RCTs | n = 146 | Isoflavone-poor soy protein vs animal protein | Total body BMD | Both RCTs found no significant differences in the net changes in total body BMD | NA | Moderate: 5.5 | |||
2 RCTs | n = 91 | Isoflavone-poor soy protein vs animal protein | BAP | Both RCTs found no significant differences in the net changes in BAP | NA | Low: 5.25 | |||
2 RCTs | n = 133 | Isoflavone-poor soy protein vs animal protein | NTX | Both RCTs found no significant differences in the net changes in NTX | NA | Low: 5.25 | |||
Shams-White 2017 | - SR with MA of RCTs - Published between 1946 and 10/2016 - Study duration: 6–24 mos | - Both sexes - Healthy people - Aged ≥ 18 yrs | Higher protein vs lower protein | 10–45 g/d vs. 0–2.1 g/d or 25–30 E% vs. 15–18 E% or 1.4 g/kg BW/d vs. 0.8 g/kg BW/d | Pooled net change (95% CI), random effect | Low | |||
5 RCTs | n = 989 | Lumbar spine BMD | 0.52% (0.06, 0.97) P = NP | I2 = 0% P = 0.58 | Low: 5.0 | ||||
7 RCTs | n = 1205 | Total hip BMD | 0.30% (− 0.02, 0.62) P = NP | I2 = 0% P = 0.54 | Low: 4.4 | ||||
6 RCTs | n = 1169 | Femoral neck BMD | − 0.14% (− 0.60, 0.32) P = NP | I2 = 0% P = 0.95 | Low: 4.4 | ||||
8 RCTs | n = 494 | Osteocalcin | 0.06 ng/ml (− 0.49, 0.60) P = NP | I2 = 27% P = 0.21 | Low: 4.4 | ||||
5 RCTs | n = 370 | CTX | 47.72 ng/L (− 27.34, 122.78) P = NP | I2 = 61.3% P = 0.04 | Low: 4.0 | ||||
Wu 2015 | - SR with MA of cohort studies - Published before 07/2014 and 10/2016 - Follow-up: NP | - Both sexes - Aged 18–89 yrs | Total protein: 50.11–> 98 g/d vs < 40.75–< 68 g/d or 20% increased (g/d) vs normal Animal protein: > 20.6–> 80 g/d vs < 13.6–< 51 g/d or 1 g/kg BW/d increased vs normal Plant protein: > 13.27–> 26.2 g/d vs < 4.98–18 g/d or 1 g/kg BW/d increased vs normal | Adjusted RR (95% CI), random effect | High | ||||
3 cohort studies | n = 215,493 | Total protein | All fractures | 0.99 (0.97, 1.02) P = NP | I2 = 0% P = 0.57 | Very low: 3.5 | |||
2 cohort studies | n = 36,342 | Animal protein | All fractures | 0.79 (0.32, 1.96) P = NP | I2 = 70% P = 0.07 | Very low: 3.5 | |||
3 cohort studies | n = 60,745 | Plant protein | All fractures | 0.77 (0.52, 1.12) P = NP | I2 = 86% P = 0.001 | Very low: 3.5 | |||
6 cohort studies | n = 270,011 | Total protein | Hip fractures | 0.89 (0.82, 0.97) P = NP | I2 = 0% P = 0.44 | Low: 4.0 | |||
4 cohort studies | n = 161,393 | Animal protein | Hip fractures | 1.04 (0.70, 1.54) P = NP | I2 = 52% P = 0.08 | Very low: 2.5 | |||
3 cohort studies | n = 121,606 | Plant protein | Hip fractures | 1.00 (0.53, 1.91) P = NP | I2 = 57% P = 0.01 | Very low: 2.5 | |||
2 cohort studies | n = 230,480 | Total protein | Limb fractures | 1.05 (0.81, 1.37) P = NP | I2 = 90% P = 0.01 | Very low: 2.5 | |||
2 cohort studies | n = 87,765 | Plant protein | Limb fractures | 0.94 (0.40, 2.22) P = NP | I2 = 86% P = 0.001 | Very low: 2.5 | |||
Pedersen 2014 | - SR without MA of RCTs and cohort studies - Published between 01/2000 and 01/2012 - Study duration: 63 d–3 yrs - Follow-up: 1–7 yrs | - Both sexes - Healthy people - Aged 55–92 yrs | High vs low protein | Cohorts: 82.7–> 87 g/d vs 46.5–< 66 g/d 1.2–> 1.6 g/kg BW/d vs 0.84–1.1 g/kg BW/d RCTs: 1.2 g/kg BW/d vs 1.1 g/kg BW/d or 20–24 E% vs 14–16 E% | NA | High | |||
1 RCT, 3 cohort studies | n = 2543 | Total protein | BMD | “The evidence is assessed as suggestive regarding a positive association between protein intake and BMD.” | Moderate: 4.5 | ||||
1 RCT, 3 cohort studies | n = 1621 | Total protein | Bone loss | 3 out of 4 studies found no statistically significant associations between total protein and bone loss | Low: 3.0 | ||||
1 RCT, 2 cohort studies | n = 1529 | Animal protein | 2 out of 3 studies found no statistically significant associations between total protein and bone loss | Low: 3.0 | |||||
Plant protein | None of the studies found significant associations between total protein and bone loss | Low: 3.0 | |||||||
1 cohort study | n = 946 | Total protein | Fracture risk | This cohort study found an inverse association | NA | ||||
1 cohort study | n = 1035 | Animal protein | Fracture risk | This cohort study found a positive association | NA | ||||
1 cohort study | n = 807 | Total, animal and plant protein | Falls | This cohort study found no significant associations between falls and total, animal and plant protein | NA | ||||
Pedersen 2013 | - SR without MA of cohort studies - Published between 01/2000 and 12/2011 - Follow-up: 2–15 yrs | - Both sexes - Healthy and/or people with prehypertension or stage 1 hypertension - Aged 14–69 yrs - n = NP for all studies | High vs low protein | 17.4-30.8 En% vs 5.6–19.9 En% or > 82.7 g/d vs 46.5–65 g/d or > 1.2 g/kg BW/d vs 0.84–1.1 g/kg BW/d | NA | Moderate | |||
3 cohort studies | n = 2210 | Total protein | Bone loss | 2 out of 3 cohort studies found no significant associations between total protein and bone loss | Low: 3.0 | ||||
3 cohort studies | n = 40,908 | Total protein | Fracture risk | None of the cohort studies found significant associations | Moderate: 5.0 | ||||
1 cohort study | n = 1,035 | Animal protein | Fracture risk | This cohort study found a positive association | NA | ||||
1 cohort study | n = 1,035 | Plant protein | Fracture risk | This cohort study found no significant associations | NA | ||||
Santesso 2012 | - SR with MA of RCTs - Published before 08/2011 - Study duration: 63–180 d | - Both sexes - Healthy and/or people with overweight, obesity hyperlipidemia, hypertension or MS - Aged 26–54 yrs | High protein diet vs low protein diet | Median: 27 En% (range: 16–45 En%) vs. 18 En% (range: 5–23 En%) | MD (95% CI), random effect | High | |||
3 RCTs | n = 208 | Total body BMD | 0.00 g/cm2 (− 0.02, 0.03) P = 0.82 | I2 = 0% tau2 = 0.00 P = 0.74 | Very low: 3.0 | ||||
2 RCTs | n = 180 | Lumbar spine BMD | 0.60 g/cm2 (− 0.34, 1.54) P = 0.21 | I2 = 67% tau2 = 0.45 P = 0.05 | Very low: 3.0 | ||||
1 RCT | n = 130 | Hip BMD | 0.61 g/cm2 (− 0.13, 1.34) P = 0.11 | I2 = 0% tau2 = 0.00 P = 0.79 | Very low: 3.0 | ||||
Darling 2009 | - SR with MA of RCTs and cohort studies - Published between 1974 and 07/2008 - Study duration: 38 d–6 mos - Follow-up: 1–22 yrs | - Both sexes - Healthy people - Aged 35–91 yrs | Higher vs lower protein | Supplements in RCTs: 40 mg/d milk-based protein or 20.4 g/d protein Cohort studies: median: 79.6 g/d | Moderate | ||||
3 cohort studies | n = 120,849 | Total protein | Hip fractures | RR (95% CI), random effect 0.75 (0.47, 1.20) P = 0.23 | I2 = 20.4% P = 0.28 | Low: 4.5 | |||
3 cohort studies | n = 157,737 | Animal protein | Hip fractures | RR (95% CI), random effect 0.83 (0.54, 1.30) P = 0.42 | I2 = 48.3% P = 0.14 | Very low: 3.5 | |||
2 cohort studies | n = 117,950 | Plant protein | Hip fractures | RR (95% CI), random effect 1.21 (0.82, 1.79) P = 0.34 | I2 = 2.0% P = 0.31 | Very low: 3.5 | |||
3 RCTs | n = 110 | Protein supplementation | Lumbar spine BMD | MD (95% CI), fixed effect 0.02 g/cm2 (0.00, 0.04) P = 0.04 | I2 = 0% P = 0.62 | Very low: 3.5 |