Nucleus of the evidence-appraisal cluster. How confident you are in an effect estimate and how strongly you should recommend acting on it are two separate judgments, and neither determines the other. GRADE calls the separation “the defining feature” of its approach and says grading systems that fail to make it create confusion. (Schünemann et al., n.d.)

Confidence: medium, and unevenly spread. The GRADE-mechanics core (what the handbook says sets strength, and its worked cases) is well extracted and supports more than that. The corpus-pairing section (Does the separation actually DO anything?) supports less: it rests on a small, non-independent set of WHO documents (one body).

Ordinary framing

Better evidence -> stronger recommendation. Grading systems that produce a single letter or star rating collapse the two, so a reader cannot tell whether a weak recommendation reflects thin evidence or a genuinely close call on well-established effects.

The reframing — certainty is ONE of four inputs to strength, and the mapping is asymmetric

The separation is not merely record two numbers: certainty is one of four determinants of strength, not a proxy for it (the other three are set out below, under What sets strength, once certainty is fixed). But the two off-diagonal cases are not symmetric, and the page must not flatten them:

  • High certainty -> weak (conditional) recommendation is what GRADE expects where values vary (GRADE states no base rate for it): panels facing important variability in values and preferences “are likely to offer a weak recommendation despite high quality evidence.” (Schünemann et al., n.d.)
  • Low or very low certainty -> strong recommendation is allowed but rare. GRADE “allows for” it where “the benefits of antibiotic therapy clearly outweigh the downsides in most patients independent of the quality assessment” — a clause about one worked example, not a general licensing condition. Separately, GRADE «In general» discourages strong recommendations when confidence in effect estimates for critical outcomes is low or very low, and names five paradigmatic situations in which they «may be warranted»; the handbook does not say whether the five are exhaustive. (Schünemann et al., n.d.)

Mechanism — what drives strength once certainty is fixed

Strength is set by the balance of desirable and undesirable consequences, the variability of values and preferences among those affected, and resource use. Where patients with differing values would predictably make different choices, panels are likely to offer a weak recommendation even on high quality evidence — because a strong recommendation asserts that most or all would choose the same way. (Schünemann et al., n.d.)

Worked cases the handbook gives:

  • High certainty, weak recommendation. Chemotherapy plus radiotherapy versus radiotherapy alone in locally advanced non-small-cell lung cancer: high quality evidence of a survival gain “of a few months” against added harm and burden — so values determine the choice.
  • High certainty, weak recommendation. Continuing long-term warfarin after a first unprovoked deep venous thrombosis: high quality trials show reduced recurrence at the cost of bleeding and inconvenience; “because patients with varying values and preferences will make different choices,” the recommendation is weak.
  • Low certainty, strong recommendation (GRADE’s counterfactual illustration, not a reported case). Rapid antibiotics in severe infection or sepsis has never been randomized against delay; on observational evidence rated low, panels “would be very likely to” make a strong recommendation. (Schünemann et al., n.d.)

Decision relevance

  • A weak recommendation is not a hedge about the science. It is a statement that the right action depends on the person’s values. Reading it as the evidence is poor inverts its meaning — and in two of GRADE’s three worked examples in §1.2 the evidence was high quality. (Not to be confused with the five paradigmatic situations of §6.3.2, which are the strong-on-low-certainty cases — none of those is high quality.) (Schünemann et al., n.d.)
  • Conversely, low certainty does not license inaction — acting through a WEAK recommendation is the default. GRADE says: «GRADE encourages panels to deal with their discomfort and to make recommendations even when confidence in effect estimate is low and/or desirable and undesirable consequences are closely balanced. Such recommendations will inevitably be weak, and may be accompanied by qualifications.» (Schünemann et al., n.d.)
  • The narrow exit is for a STRONG recommendation on low certainty. GRADE «In general» discourages a strong recommendation when certainty for critical outcomes is low or very low, and names «five paradigmatic situations in which strong recommendations may be warranted despite low or very low quality of evidence», framed as cases where «a panel would have a low level of regret if subsequent evidence showed that their recommendation was misguided» (Schünemann et al., n.d.) — a regret criterion, not simply a lopsided balance. The handbook does not say whether the five situations are exhaustive. Treating the exit as freely available inverts the rule as surely as ignoring it does.
  • The practical test when reading any recommendation: ask which of the two judgments produced its strength. If the source does not let you tell them apart, you cannot use its strength rating.

What sets strength, once certainty is fixed (chunk 03)

Strength is “the extent to which a guideline panel is confident that desirable effects of an intervention outweigh undesirable effects, or vice versa, across the range of patients for whom the recommendation is intended.” Four factors drive direction and strength: the balance of desirable and undesirable consequences, certainty in the effect estimates, confidence in values and preferences, and resource use. (Schünemann et al., n.d.)

  • Strong — the panel is confident the desirable effects outweigh the undesirable (or the reverse); “most or all individuals will be best served by the recommended course of action.”
  • Weak — the desirable effects probably outweigh the undesirable “but appreciable uncertainty exists”; “not all individuals will be best served,” so the individual’s circumstances and preferences need more careful weighing than usual.
  • Default under uncertainty is weak: where the panel is unsure whether the balance is clear, or the information bearing on the four factors is unavailable, a panel «should be more cautious and in most instances it would opt to make a weak recommendation». (Schünemann et al., n.d.)

On the balance itself: large relative effects pointing the same way favour a strong recommendation; large desirable effects accompanied by large undesirable ones favour a weak one. Large absolute effects favour strength more than small ones — and since baseline risk drives absolute effect, «The strength of recommendations and its direction, therefore, will likely differ in high- and low-risk groups» (Baseline Risk and the Relative-Absolute Split). (Schünemann et al., n.d.)

Two cautions GRADE states about its own output

  • “Strong recommendations are not necessarily high priority recommendations.” (Schünemann et al., n.d.) Strength expresses confidence in the direction of net benefit, not the size of the benefit or its rank against other things a person could do. An intervention hierarchy therefore cannot be read off recommendation strength — a confidently-correct small effect and a confidently-correct large one are both “strong.” (inferred from Schünemann et al., n.d.)
  • Recommendations are not dictates. “Even strong recommendations based on high-quality evidence will not apply to all circumstances and all patients,” and users “may reasonably conclude that following some strong recommendations based on the high quality evidence will be a mistake for some patients,” because no guideline “can take into account all of the often compelling unique features of individual patients.” Applying recommendations “by rote or in a blanket fashion” is explicitly disclaimed. (Schünemann et al., n.d.)

Applied — WHO SFA 2023, and how the four determinants show up in a real rationale

WHO’s 2023 SFA guideline is GRADE applied by an adopting body, and it exhibits the certainty->strength linkage cleanly (Saturated Fat Intake and Replacement):

RecommendationCertaintyStrength
Reduce SFA to 10% energymoderatestrong
Reduce below 10%very lowconditional
Replace with PUFAmoderatestrong
Replace with MUFA / carbohydratelowconditional

The instructive case is the sub-10% recommendation. WHO issued it as conditional while recording that “no undesirable effects or other mitigating factors were identified that would argue against” it — so the conditionality is not a balance judgment at all. It is driven by certainty alone: “a conservative approach was therefore taken, leading to a conditional recommendation.” (World Health Organization, 2023)

  • CORRECTED 2026-07-26 — this is GRADE’s default rule, not a new pattern. It was previously written up here as a fifth pattern alongside GRADE’s own paradigm cases. That claim is refuted by a quotation this same page carries above: where a panel “is uncertain whether the balance is clear” or the information bearing on the four factors “is not available,” it «should be more cautious and in most instances it would opt to make a weak recommendation». WHO rated very low certainty and issued a conditional recommendation — the default executing normally. Retained below as a worked illustration of that rule, which is what it is good for; the paradigm-case framing is struck.
  • This is a fifth pattern alongside GRADE’s own paradigm cases: not high-certainty-weak (values vary) nor low-certainty-strong (balance lopsided), but low-certainty-conditional-despite-no-known-harm — where the panel could identify nothing against the action and still declined to recommend it strongly. Under GRADE’s four factors this is certainty operating almost alone, which the handbook permits but does not illustrate. (inferred from Schünemann et al., n.d.; World Health Organization, 2023)
  • Reading consequence: a conditional recommendation cannot be assumed to mean “the benefit is contested.” Here it means “we could not rule it out and could not confirm it” — a different message with a different action implication.

Limits

  • GRADE does not remove judgment, and says so. “Applying the GRADE approach does not minimize the importance of judgment or [suggest] that quality can always be objectively determined”; there will be cases of “legitimate disagreement about the interpretation of evidence.” Its claim is that judgments become explicit rather than implicit. (Schünemann et al., n.d.)
  • The framework governs guideline panels offering “recommendations for the management of typical patients.” (Schünemann et al., n.d.) Carrying a strength rating to a specific individual requires a transportability step the rating itself does not perform. (inferred from Schünemann et al., n.d.)
  • Source currency: this handbook (updated October 2013) is being phased out in favour of the GRADE Book (2024); §§1-3 are flagged in-source as already rewritten there. (Schünemann et al., n.d.)

A real process where appraisal and recommendation are done by DIFFERENT PEOPLE [2026-07-27]

This page treats certainty and strength as two judgements a guideline panel makes. The US Dietary Guidelines process separates them institutionally — and NASEM records that as atypical. The committee’s own view on conduct: «USDA and HHS should decide how the update is conducted, while making sure the process is transparent» (National Academies of Sciences Engineering and Medicine, n.d.). The atypicality is stated in the same paragraph:

«Multiple sets of standards exist that could serve as models, with the understanding that the process for developing DGA recommendations does not follow typical guidelines development processes because the experts assessing the evidence do not write the guidelines and recommendations themselves.» (National Academies of Sciences Engineering and Medicine, n.d.)

Why this matters beyond the DGA. The certainty/strength split is normally a distinction within one panel’s reasoning; here it is a handoff between two bodies, which creates a failure mode the framework above does not describe: the recommendation can depart from the appraisal silently, because no one person owns both.

NASEM’s response is not to close the seam but to make it auditable — Recommendation 2 asks that «the secretaries of USDA and HHS should provide the public with a clear explanation when the DGA omit or accept only parts of conclusions from the scientific report.» (National Academies of Sciences Engineering and Medicine, n.d.)

NASEM states the departure outright. «Additionally, how the writing team interpreted the DGAC Scientific Report and why some conclusions were modified or omitted when developing the DGA Policy Report has also not been clearly described. This lack of transparency resulted in suggestions that the process is being inappropriately influenced by the food sector, lobbyists, faddism, and the federal government» (National Academies of Sciences Engineering and Medicine, n.d.). So conclusions were modified or omitted, and the reasons were not clearly described. What this does NOT establish is that any specific recommendation was distorted — NASEM reports suggestions of influence and asks for transparency about the handoff; it does not document a case. (inferred from National Academies of Sciences Engineering and Medicine, n.d.) -> [[Which Objective Moved This Recommendation]]

A third axis the two-axis frame does not hold: how PRECISE is the number [2026-07-28]

NNR attaches no strength or certainty grade to any recommendation it issues — not one, across all ten chunks. [searched: "strong recommendation"/"conditional recommendation"/"GRADE approach"/"certainty of evidence" across all 10 NNR chunks — 1 hit total, and it is NNR describing WHO's grading of WHO's own sugars guideline, not grading its own]

Scope: the absence is of a grade on the RECOMMENDATION, not of evidence grading. The four-term search used GRADE vocabulary only. NNR rests its conclusions on qualified systematic reviews whose required protocol includes «7. Synthesis and grading of total strength of evidence following specific procedures» (Nordic Council of Ministers, 2023), and its health-effect text is «based on qualified systematic reviews (see Table 2), emphasizing conclusions with strong evidence (convincing and probable evidence)» (Nordic Council of Ministers, 2023). So the evidence beneath NNR is graded (convincing/probable); what NNR does not issue is a strength or certainty grade on its own recommendations.

That alone would only make NNR one more body issuing ungraded recommendations. What makes it worth a section is the reason NNR gives, which introduces a distinction this page did not hold:

«FBDGs are formulated more generally than the DRVs for nutrients, although the causal associations of foods and health outcomes can be stronger than for nutrients and health outcomes. As with DRVs, there are seldom precise calculations behind the quantitative FBDGs. The precise FBDGs are based on best scientific knowledge and most often decided as consensus among expert groups.» (Nordic Council of Ministers, 2023)

Read the first clause before the third, because it reverses the obvious reading. This is not a body confessing that its evidence is weak. NNR says the causal association behind a food-based guideline can be stronger than the nutrient-level one — and then says the number attached to it is a consensus rather than a calculation. Those are claims about two different things.

AxisQuestionHeld by
Certaintyhow confident are we that the effect is real and about this big?GRADE, WHO
Strengthhow strongly should we recommend acting on it?GRADE, WHO
Precision of the thresholdis the specific number derived, or agreed?NNR, explicitly

The third is independent of the first two, and that is the finding. NNR’s own procedure is curve-first: «Then, a dose-response curve should be considered in a meta-analysis or qSR. If a dose-response curve can be established, a quantitative FBDG may be formulated. If no adequate dose-response curve can be established, in general, a qualitative FBDG may still be formulated (Arnesen et al., 2020a; Christensen et al., 2020).» (Nordic Council of Ministers, 2023) Its whole-grain number sits on such curves: «Dose-response curves show that risk reduction for all-cause mortality is observed for intakes up to 50-60 g/day of whole grains. Higher intakes (i.e. 90 g/day) confer even greater risk reduction for coronary heart disease, type 2 diabetes and colorectal and breast cancer (Reynolds et al., 2019).» (Nordic Council of Ministers, 2023) So for curve-based FBDGs such as whole grains, what the consensus passage describes is where on an established curve the round number is placed, not whether a curve exists. NNR also names an adequacy-based exception: «Some food groups may have a quantitative FBDG, even without an established dose-response relationship with a health outcome (e.g., dairy). In such cases, the quantitative FBDG is based on the significance of the food group for nutritional adequacy and the intake of specific nutrients.» (Nordic Council of Ministers, 2023) There the number rests on nutritional adequacy, not on a health-outcome curve. Certainty about a relationship does not transmit to precision about a cutpoint — and a reader who sees only a confident recommendation attached to a round number will assume it does. (inferred from Nordic Council of Ministers, 2023)

This is consistent with a rule the corpus already carries. The dose-response section of the telos holds that where a threshold appears in guidance, the first hypothesis is that it marks the edge of the evidence, not a feature of the curve. NNR’s statement that the numbers are «most often decided as consensus among expert groups» fits that rule, but it is not the same claim: NNR describes consensus placing a point on an established curve, not a threshold sitting at a data edge. SACN’s 30 g remains the worked instance of the edge-of-evidence case.

What this does NOT establish. NNR is describing how quantitative FBDGs are set in general, including its own; it is not conceding that any particular NNR number is arbitrary, and it explicitly grounds them in «best scientific knowledge». Consensus-derived is not evidence-free — it is un-graded, which is a different and narrower claim.

Where this leaves the recommending bodies the wiki holds that issue no recommendation grade. SACN issues no strength axis (Scientific Advisory Committee on Nutrition, 2015); NNR attaches no grade to its recommendations, though its underlying qSRs grade evidence strength (convincing/probable), and it explains why its quantitative FBDG numbers are consensus rather than calculation (not why it does not grade); ADA 2019 grades nothing (below). NASEM is a different kind of body — a process reviewer that issues no dietary recommendations of its own and wants GRADE’s properties without the brand. None of these is evidence for the others — several bodies declining to grade recommendations is a fact about the field’s norms, not a converging judgment about grading’s value, and no [E-independent] is claimed.

Self-critique of the NNR addition [run 2026-07-28, before commit]

  • Quote-truncation check: RUN, and it caught the error before the prose was written. The usable sentence is «there are seldom precise calculations behind the quantitative FBDGs». Quoted alone it reads as a confession of weak evidence. The sentence before it says the opposite — food-level causal associations «can be stronger» than nutrient-level ones. The full passage is quoted and the paragraph explicitly instructs the reader to take the first clause first. This is the same failure shape as the four defects logged in the 2026-07-27 session, caught by running the check rather than by noticing.
  • Absence claim: SCOPED, not asserted. NNR attaches no grade to its recommendations is backed by a stated four-term search across all ten chunks with the single hit explained (NNR describing WHO). The earlier form of this claim on the source page carried no search scope. Corrected 2026-10-28: this bullet and the closing paragraph previously read NNR grades nothing; the GRADE-vocabulary search could not see NNR’s qSR-protocol grading of evidence strength (convincing/probable), so the claim is narrowed to recommendations.
  • Over-claim check on the third axis. The claim is that precision of the threshold is independent of certainty and strength. NNR states the certainty/precision split; the three-axis table and the independence claim are the wiki’s and are tagged. A weaker reading is available — that NNR is merely restating low certainty in other words — and the first clause of the quote is what rules it out, which is why the quote could not be truncated.
  • Laundered independence: explicitly denied in the body. Several bodies that do not grade their recommendations are not several witnesses; the closing paragraph says so rather than leaving the count to imply corroboration.
  • Residual: NNR’s passage is about quantitative FBDGs generally. Applying it to NNR’s own numbers is warranted by «As with DRVs» and by NNR issuing no recommendation grades (it grades evidence strength inside its SRs), but NNR never says “our numbers are consensus” in the first person.

Does the separation actually DO anything? The corpus’s own pairings [2026-07-28; re-scoped 2026-10-28]

This page holds GRADE’s claim that certainty is one of four determinants of strength and not a proxy for it, and that panels facing important variability in values and preferences are likely to issue the off-diagonal case — high certainty, weak recommendation. The corpus holds enough graded recommendations to check whether that separation shows up in practice. In the table below it mostly does not, but it does show once: most tabled pairs sit on the diagonal, and the one high-certainty pair in the table is conditional, for a reason WHO states that is not certainty.

Read the sample bound FIRST, because it decides what this can mean. Every pair in this table comes from four of the five WHO GRADE guidelines the wiki holds (SFA/TFA 2023, Total Fat 2023, Non-Sugar Sweeteners 2023, Sugars 2015) — one body, so the four are not independent witnesses. The fifth, WHO Sodium 2012, is held and GRADE-based but attaches no certainty grade to any recommendation, so it yields no pairs (see Bodies that do not label one of the axes below). The PA evidence-profile annex carries per-outcome certainty but no recommendation strengths (it is the annex, not the guideline), so it contributes no pairs. This is one body’s house practice, not a finding about guideline-making.

The wiki holds other graded guidelines that this table does not cover; those named here are examples, not a complete list. The table is limited to WHO so that one body’s method is held constant. NutriRECS 2019 (GRADE; discussed below) pairs a weak recommendation with low certainty, so its single pair would sit on the diagonal. Endocrine Society vitamin D 2024 (GRADE scale) and EASL MASLD 2024, ESC 2021, ESC-EAS 2025, EAU 2026 and ACC-AHA 2026 (other scales) are held, graded and not tabulated; see Held tests nobody has run yet below.

SourceRecommendationStrengthOverall certaintyWHO’s stated reason for the strength
WHO SFA 2023reduce SFA to 10%Estrongmoderate—
WHO SFA 2023further reduce to <10%Econditionalvery lowcertainty (conservative approach)
WHO SFA 2023replace with PUFAstrongmoderatecertainty («moderate certainty overall»)
WHO SFA 2023replace with MUFA (plant)conditionallow—
WHO SFA 2023replace with carbohydrate from fibre-containing foodsconditionallow—
WHO TFA 2023reduce TFA to 1%Estrongmoderatecertainty + no mitigating factors
WHO TFA 2023further reduce to <1%Econditionallowcertainty (conservative approach)
WHO TFA 2023replace TFA with PUFA or plant MUFAconditionalnone single (PUFA very low, MUFA moderate)thin disease-outcome evidence; LDL not disease (conservative approach)
WHO Total Fat 2023limit total fat to <=30%E (adults)conditionalhighsubstitution reduces net benefit
WHO NSS 2023do not use NSS for weight control / NCD riskconditionallowuncertain balance, possible reverse causation (conservative approach)
WHO Sugars 2015free sugars <10%Estrongmoderate (observational, caries)—
WHO Sugars 2015free sugars <5%Econditionalvery low (ecological)—
WHO Sugars 2015reduce free sugars throughout the lifecoursestrongnone stated (mixed: weight low-moderate, caries very low-moderate)—

(— = reason not extracted to this page.)

(World Health Organization, 2023) (World Health Organization, 2015) (World Health Organization, 2015)

TFA rows, in WHO’s words:

  • «1. WHO recommends that adults and children reduce trans-fatty acid intake to 1% of total energy intake (strong recommendation).» (World Health Organization, 2023)
  • «2. WHO suggests further reducing trans-fatty acid intake to less than 1% of total energy intake (conditional recommendation).» (World Health Organization, 2023)
  • «These systematic reviews found that lower TFA intake reduced the risk of CVDs. The overall certainty in the evidence for recommendation 1 was moderate and for recommendation 2 was low.» (World Health Organization, 2023)
  • «▶▶ Recommendation 1 was assessed as strong because evidence of overall moderate certainty from different study types assessing both risk factors and incidence of CVDs suggested reduced risk of all- cause mortality, CVDs and coronary heart disease with lower TFA intake (in a dose-dependent manner with respect to all-cause mortality and coronary heart disease).» (World Health Organization, 2023)
  • «▶▶ Recommendation 2 was assessed as conditional because, although there is evidence from different study types from each of the systematic reviews suggesting reduced risk of all-cause mortality, CVDs and coronary heart disease with TFA intakes of less than 1% of total energy intake, the evidence is more limited than for intakes greater than 1% of total energy intake and therefore there is less confidence in it (low certainty evidence overall).» (World Health Organization, 2023) For both TFA recommendations 1 and 2 the rationale goes on to say no undesirable effects or mitigating factors argued against the action; for recommendation 2 it then states that a conservative approach was taken, leading to the conditional grade (paraphrased; same locus).
  • «3. WHO recommends replacing trans-fatty acids in the diet with polyunsaturated fatty acids or monounsaturated fatty acids primarily from plant sources (conditional recommendation).» (World Health Organization, 2023)
  • «▶▶ Recommendation 3 is based on very low certainty evidence overall for replacing TFA with poly­un­ saturated fatty acids and moderate certainty evidence overall for replacing TFA with monounsaturated fatty acids from plant sources. Evidence comes from two systematic reviews that assessed the effects of lower compared with higher TFA intake via replacement nutrient analysis.» (World Health Organization, 2023)
  • «▶▶ Recommendation 3 was assessed as conditional because evidence for disease outcomes comes only from a limited number of observational studies; most of the evidence is from RCTs with LDL cholesterol as an outcome. The evidence for LDL cholesterol is of high certainty. However, although LDL cholesterol is a well-established biomarker for measuring the effects of interventions on CVD risk, and is considered by many to be a causal factor for atherosclerosis and coronary heart disease, it is not a physical manifestation or confirmation of disease. Therefore, a conservative approach was taken, leading to a conditional recommendation.» (World Health Organization, 2023)

Total Fat row, in WHO’s words — the recommendation, its certainty, and the stated reason for the conditional grade (read end to end; the rationale names no other reason):

  • «1. To reduce the risk of unhealthy weight gain, WHO suggests that adults limit total fat intake to 30% of total energy intake or less (conditional recommendation)» (World Health Organization, 2023b)
  • «▶▶ This recommendation is based on evidence of high certainty from a systematic review of RCTs of dietary fat reduction in adults in which weight loss was not an explicit goal.» (World Health Organization, 2023b)
  • «▶▶ The recommendation was assessed as conditional because some individuals who reduce their fat intake might replace some of the energy from dietary fat with energy from foods that are undesirable from a dietary quality perspective (e.g. free sugars), reducing the net benefit.» (World Health Organization, 2023b)
  • The same paragraph goes on: «It is therefore important to consider this recommendation in the context of other WHO dietary recommendations, including those on free sugars and carbohydrates, which provide guidance on carbohydrate quality.» and «The evidence did not suggest any undesirable effects with respect to serum lipids, blood pressure or quality of life from lower total fat intake, but rather of small benefits or no effect (all high certainty evidence, except for quality of life, which was assessed as low certainty evidence).» (World Health Organization, 2023b) Neither sentence reverses the stated reason; the second confirms the conditional grade is not about direct harms. That it is not about certainty comes from the basis sentence above («evidence of high certainty»).
  • Total Fat’s recommendation 2 (fat quality, strong) is not tabled: WHO places its evidence-to-recommendation information in the SFA/TFA guideline, so it would duplicate rows already above.

NSS row, in WHO’s words:

  • «WHO suggests that non-sugar sweeteners not be used as a means of achieving weight control or reducing the risk of noncommunicable diseases (conditional recommendation).» (World Health Organization, 2023c)
  • «The overall certainty in the available evidence for an effect of NSS intake on outcomes in adults was assessed as low.1 GRADE assessments for each outcome can be found in Annex 6, GRADE evidence profiles 1 and 2.» (World Health Organization, 2023c)
  • «▶▶ Because of lack of certainty about the overall balance of desirable and undesirable effects associated with long-term NSS use for reducing NCD risk, including the possibility that reverse causation1 may have contributed to one or more of the associations observed between long-term NSS use and risk of disease in prospective observational studies, a conservative approach was taken, leading to a conditional recommendation.» (World Health Organization, 2023c)

Eleven pairs carry a single overall certainty grade; ten sit on the diagonal and one does not. The four strong recommendations all pair with moderate certainty; six of the seven conditional ones pair with low or very low. The exception is WHO Total Fat: high certainty, conditional — the case GRADE expects where values vary. WHO’s stated reason is substitution: people who cut fat may replace it with free sugars, reducing the net benefit. That is the balance determinant, reached through what people eat instead (an implementation consideration), not certainty and not variability in values. Two rows carry no single certainty grade and cannot be placed: the strong lifecourse sugars recommendation (mixed very-low-to-moderate evidence) and the conditional TFA replacement (very low for PUFA, moderate for plant MUFA, conditional because disease-outcome evidence is thin). The rare case GRADE discourages — low certainty, strong — occurs zero times in the table.

The diagonal is GRADE’s own expected default, not evidence against the separation. GRADE says: «Typically, a strong recommendation is associated with high, or at least moderate, confidence in the effect estimates for critical outcomes. If one has high confidence in effects on some critical outcomes (typically benefits), but low confidence in effects on other outcomes considered critical (often long-term harms), then a weak recommendation is likely warranted.» (Schünemann et al., n.d.) And the sample can barely test the off-diagonal GRADE expects where values vary: only one tabled pair (Total Fat) has HIGH overall certainty, so the high-certainty column has a denominator of one — and that one is off the diagonal.

What this does and does not license

  • It is NOT WHO collapsed certainty into strength. If the other three determinants — balance of benefits and harms, variability in values and preferences, resource use — happen not to vary much across a set of recommendations, then a mostly-diagonal mapping is the correct output, not a defect. Total Fat shows the condition does not always hold even for a population nutrient threshold: the balance determinant moved it off the diagonal through substitution.

  • What it DOES establish is a reader-side problem, and that is the decision-relevant part. From the label alone you cannot tell which reason produced a conditional:

    • conditional because the evidence is thin -> more research could move it, and the reader’s own values are not what is in question;
    • conditional because of the balance or values (Total Fat’s substitution; NutriRECS’s values, below) -> more research on the effect will NOT move it, and what the reader eats instead, or what the reader values, is what decides it.

    Those license different actions, and the label is identical in both cases. Where the mapping is diagonal, the strength label carries no information the certainty grade did not already carry — so in the tabled WHO sample, reading only the certainty would lose nothing on ten of the eleven single-graded pairs, and would miss the one that matters (Total Fat).

  • The practical rule: read the was assessed as strong / conditional because sentence, not the label. Every WHO 2023 rationale quoted on this page has one (SFA sub-10%, TFA 1-3, Total Fat; NSS states an equivalent a conservative approach was taken sentence). Reasons like less confidence or a conservative approach was taken mean the evidence is thin; reasons like substitution, net benefit or values mean the outcome is decided by the balance or by the person. With that sentence in hand, from the output alone you cannot tell is true of the label but not of the document.

  • The rationale text is where the distinction survives. WHO SFA’s strength rationale for the strong PUFA recommendation cites «moderate certainty overall» — i.e. it reaches for certainty to justify strength rather than for the other three determinants. WHO Total Fat’s rationale reaches for the balance instead.

(World Health Organization, 2015, inferred from 2023a, 2023b, 2023c)

The falsifier, stated in advance — and partly fired

A guideline whose recommendations vary in values/preferences should produce the off-diagonal — screening decisions, drugs carrying burden or side-effects, interventions with real opportunity cost.. The original fallback — if the off-diagonal appears elsewhere, this pattern is WHO-and-nutrition-specific — no longer stands: the off-diagonal has already appeared inside WHO nutrition guidance (Total Fat), through the balance determinant rather than values. The narrower claim that survives: in WHO nutrition guidelines, strength usually tracks certainty, and the exception is announced in the rationale sentence. The open test is whether a guideline with genuinely values-variable recommendations produces high-certainty conditional recommendations driven by values, which the WHO sample does not show. Recorded as a prediction rather than a conclusion, because a 13-row single-body sample is the size at which a pattern is most tempting and least supported.

Held tests nobody has run yet (G-gap, not a result). The guidelines below are among those held and graded (examples, not a complete list). None of their pairs has been tabulated or counted here.

Endocrine Society vitamin D 2024 (Demay), on the same GRADE scale as the table. «The Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) methodology was used to assess the certainty of evidence and guide recommendations.» (Demay et al., 2024) It pairs a strength digit with a ⊕ certainty on its recommendations, and some are conditional on moderate-certainty evidence — for example «In the general population aged 75 years and older, we suggest empiric vitamin D supplementation because of the potential to lower the risk of mortality. (2 | ⊕⊕⊕◯)» (Demay et al., 2024), where the strength digit 2 is «2—Conditional recommendation for or against» and ⊕⊕⊕◯ is «Moderate ⊕⊕⊕O We are moderately confident in the effect estimate; the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.» (Demay et al., 2024) By this page’s diagonal (strong with moderate or high, conditional with low or very low), that pair sits off the diagonal. For this recommendation Demay states the reason for the grade, and the reason is certainty: «In the absence of high overall certainty of evidence, the panel issued a conditional recommendation in this regard.» (Demay et al., 2024) The sentences before it set the direction, not the strength: the panel judged that the balance of effects probably favours supplementation (same locus). So this cell is consistent with strength tracking certainty. For this recommendation, Demay places the strong/conditional line at high certainty rather than at moderate: the same axis on a different cut point, not an instance of the separation. Among its moderate-certainty conditional recommendations, Demay gives a certainty reason for this one only; its other certainty reasons sit on low- or very-low-certainty, on-diagonal cells [searched with srcgrep: "In the absence of high" 1/4 Demay chunks; "overall certainty" 2/4 chunks, 4 hits]. The open test is the other conditional recommendations on moderate certainty. The one against routine supplementation at ages 50-74 explains the against (evidence of little or no benefit, high certainty per outcome though the recommendation’s overall rating is moderate) but not the conditional grade, and adds a circumstances consideration: «There was therefore no compel- ling rationale to recommend empiric vitamin D in this age group, especially since supplementation would involve costs (admittedly minor) and inconvenience. Importantly, most of the recent trials were completed in populations that were meeting their DRI and did not have low vitamin D status at baseline. Given the well-established harmful consequences of very low vitamin D status on skeletal health and calcium homeostasis, the panel judges that some subgroups in this age group could rationally choose to take vitamin D supplementation, especially if they are not expected to have adequate vitamin D status via sun exposure (dark complexion, housebound, clothing style) or reliable IOM-recommended intake via diet, supplements or ingestion of vitamin D–fortified foods.» (Demay et al., 2024) The one for high-risk prediabetes gives no reason for the conditional grade in its justification [read: Recommendation 10 justification, chunk 02].

EASL MASLD 2024, on the Oxford Centre for Evidence-based Medicine levels rather than GRADE. It attaches a level of evidence and a strength to its recommendations (European Association for the Study of the Liver, 2024) — for example «In the general population, non-pharmacological measures should be recommended to prevent the development of MASLD and its complications, including hepatocellular carcinoma, and preventive measures should be reinforced in high-risk groups (LoE 3, strong recommendation, strong consensus).» (European Association for the Study of the Liver, 2024) EASL states its own mapping, and how it handles recommendations without supporting literature: «The levels of evidence were developed by applying the Oxford Centre for Evidence-based Medicine system (Table 1).6 The strength of recommendations reflects the quality (grade) of underlying evi- dence (Table 2).The committee memberssubmitted contributions to specific PICO questions, which were integrated into the final document and discussed between the panel members to achieve unanimous consensus for each of the recommendations. The Wilson and Jungner criteria served as a framework for evaluating the appropriateness and effectiveness of screening programs and the implementation of population-based screening initiatives.7 In cases where the committee determined guidance to be necessary despite a lack of available supporting literature, a recommenda- tion was developed based on expert opinion and consensus.» (European Association for the Study of the Liver, 2024) (The sentences that follow concern Delphi agreement thresholds, not the strength mapping.) EASL’s in-text table references are swapped relative to its captions: the quote points levels to Table 1 and strength to Table 2, but by caption Table 1 is the grades of recommendation and Table 2 the levels of evidence. The tags below follow the captions. Its grades-of-recommendation table nonetheless lists criteria beyond evidence for a grade — consistency of studies, risk-benefit ratio, individual preferences, ethical obligations and feasibility (European Association for the Study of the Liver, 2024). Its levels table puts level 1 in a high band, levels 2-3 in an intermediate band and levels 4-5 in a low band, where any estimate of effect is uncertain (European Association for the Study of the Liver, 2024); so the LoE 3 example above sits in EASL’s middle band. EASL also pairs lower levels of evidence with strong recommendations. The candidates are its level-5 strong recommendations, in the low band — for example «In the absence of a formal demonstration of histological improvement in large, well conducted, phase III trials, glucagon-like peptide 1 receptor agonists (GLP1RA) cannot currently be recommended as MASH-targeted therapies (LoE 5, strong recommendation, strong consensus).» (European Association for the Study of the Liver, 2024) Whether these strong-on-low-evidence cells are the case GRADE discourages, or simply a different scale’s mapping, is unchecked.

ACC-AHA 2026, ESC 2021, ESC-EAS 2025 and EAU 2026. ACC-AHA attaches a class of recommendation and a level of evidence: «The Class of Recommendation (COR) indicates the strength of recommendation and encompasses estimated benefit in proportion to risk. The Level of Evidence (LOE) rates the quality of scientific evidence supporting the in­ tervention based on the type, quantity, and consistency of data from clinical trials and other sources (Table 3).9» (Blumenthal et al., 2026) ESC 2021 does the same, and EAU 2026 (level of evidence plus a strength rating) and ESC-EAS 2025 (class plus level) are also held and graded. [searched with srcgrep: Class of recommendation — ESC 2021 5/11 chunks, ESC-EAS 2025 3/4; Level of evidence — ESC 2021 6/11, ESC-EAS 2025 3/4, EAU 2026 3/6; Strength rating — EAU 2026 4/6] For EASL, ESC 2021, ESC-EAS 2025, EAU 2026 and ACC-AHA, the scales are not established as the same quantity as GRADE certainty, and none of their pairs has been checked against the diagonal. (Blumenthal et al., 2026; inferred from European Association for the Study of the Liver, 2024; European Association of Urology et al., 2026; European Society of Cardiology, 2021; Mach et al., 2025)

The falsifier can’t run against a NON-grading body — ADA 2019 [2026-09-16]

The AWAITS on ADA is cashed, and the result is that the cell grid does not exist to test. ADA’s 2019 nutrition-therapy consensus report is a major body issuing values-laden, individualized recommendations — exactly where the off-diagonal (high-certainty conditional) was predicted — but it grades nothing: no per-outcome certainty rating, no strength grade, no EtD table. Its own method statement: «the recommendations presented are the informed, expert opinions of the authors after consensus was reached through presentation and discussion of the evidence» (Evert et al., 2019).

So ADA neither confirms nor falsifies the mostly-diagonal WHO pattern — it sidesteps the grid. Where WHO maps certainty x strength cell-by-cell, ADA dissolves the population recommendation into individualization: «there is not an ideal percentage of calories from carbohydrate, protein, and fat for all people with or at risk for diabetes; therefore, macronutrient distribution should be based on individualized assessment» (Evert et al., 2019). That is the values/preferences determinant doing maximal work — it doesn’t produce a weak graded recommendation, it produces no population macronutrient percentage, handing the weighting to the person -> The Weighting Problem - Why Population Guidance Is Ill-Posed and Individual Advice Is Not. ADA still gives a population-level direction on carbohydrate: «Reducing overall carbohydrate in- take for individuals with diabetes has demonstrated the most evi- dence for improving glycemia and may be applied in a variety of eating patterns that meet in- dividual needs and preferences.» (Evert et al., 2019)

The narrowing this forces: the certainty x strength cell grid is a property of bodies that grade. A consensus body enacts the same underlying logic (high values-variability -> defer to the individual) through a different instrument (individualization) that has no cells to inspect. So the mostly-diagonal finding stays scoped to graded guidance; a body’s absence of an off-diagonal cell can mean it graded and found stable values, OR that it never graded at all. NASEM (still AWAITS) remains the live test of a body that grades AND has genuinely values-variable recommendations; the held graded guidelines (e.g. Endocrine Society vitamin D 2024 on the same GRADE scale; EASL, ESC, ACC-AHA on other scales) have not yet been checked (see the G-gap above). (inferred from Evert et al., 2019)

The prediction is partly met — NutriRECS, where VALUES set the strength [2026-07-29]

The corpus also holds a non-WHO, non-nutrient-threshold GRADE-graded recommendation (not the only held non-WHO graded one; see the G-gap above) that meets the prediction partly — at the rationale level, not the label level. NutriRECS 2019 issued a weak recommendation to continue red/processed meat — and the strength is driven not only by low certainty but explicitly by the values/preferences determinant, which none of the WHO rationales quoted on this page invokes. (Johnston et al., 2019) The engine is a dedicated systematic review of those values (Valli 2019): «Conclusion: Low-certainty evidence suggests that omnivores are attached to meat and are unwilling to change this behavior when faced with potentially undesirable health effects.» (Valli et al., 2019) NutriRECS combines that with a small, uncertain benefit to reach weak — and states it would stay weak even granting a small causal effect. So here the strength axis carries information the certainty axis does not: two of GRADE’s other three determinants (balance of benefits/harms; variability in values) are doing visible work -> Should Adults Reduce Red and Processed Meat.

This is still not the pure off-diagonal (NutriRECS’s certainty is also low, so certainty and strength remain aligned in level). It is consistent with a subtler reading: as soon as a body reviews values explicitly and takes the individual-decision standpoint, the values determinant becomes the audible driver of strength. It does NOT show that the WHO diagonal is a property of population nutrient thresholds as such — WHO Total Fat is a population nutrient threshold that went conditional on high certainty through the balance determinant. The reader-side problem sharpens: NutriRECS’s weak is values-driven (the reader’s values decide; more research on the health effect would not move it), whereas WHO’s conditional on free sugars <5%E is thinness-driven (more research could) — identical labels, different action-implications. Qualifier: the values evidence is itself low-certainty (Valli’s own conclusion, above), so more research will not move it holds only to the extent that the values estimate is right — better values research could still move NutriRECS’s weak.

Bodies that do not label one of the axes on their recommendations

  • SACN issues no strength axis whatsoever — its bespoke instrument grades evidence and stops, so the certainty/strength question cannot even be posed of it.
  • WHO Sodium 2012 issues strength without a single overall certainty grade (World Health Organization, 2012) — every recommendation is «strong», the word conditional does not appear in the document, and no certainty rating is attached to any recommendation line. Certainty is stated per outcome, not per recommendation: the Remarks say «The evidence regarding the relationship between sodium intake and blood pressure was of high quality, whereas the evidence regarding sodium intake and all-cause mortality, cardiovascular disease, stroke and coronary heart disease was of lower quality.» (World Health Organization, 2012) The strength-considerations annex likewise lists quality of evidence by outcome. (World Health Organization, 2012)
  • EASD-DNSG 2023 is the MIRROR — certainty without a formal strength axis [2026-09-16]. It grades every recommendation’s certainty with GRADE’s ⊕ notation (very low / low / moderate / high) but issues no strong/conditional label: «This certainty of evidence determines how recommendations are worded. However, there is some variability in wording given the diversity in recommendations made.» (Diabetes and Nutrition Study Group of the EASD, 2023). Strength is folded, loosely (EASD concedes the wording varies), into the verb («should» / «can be» / «not recommended»), never graded. So WHO-Sodium and EASD point opposite ways — a strength label with certainty stated only per outcome vs a per-recommendation certainty grade with no strength label — which shows the two axes can be reported separately: real bodies label each without the other. What EASD loses by folding strength into wording is the very information this page is about — a reader cannot tell whether a moderate-certainty «not recommended» is a close values call or a firm balance judgement. -> Was GRADE Actually Used (inferred from Diabetes and Nutrition Study Group of the EASD, 2023)

Note what the second one means for the table above: WHO’s own practice is not constant across its guidelines. The 2012 sodium guideline states certainty per outcome but attaches no overall certainty to a recommendation; the 2023 SFA guideline states one per recommendation. So the diagonal pattern can only be read off WHO guidelines that report a per-recommendation certainty, and the corpus can see the reporting format changing — which is a further reason, beyond sample size, to refuse to generalise it.

Self-critique [run 2026-07-28, before commit]

  • The sample bound is stated before the table, not after it. Eleven single-graded cells from four documents by one body is the kind of evidence that reads as a law. Corrected 2026-10-28: this section first read seven cells, two documents, zero off-diagonal; it had under-sampled the held WHO guidelines (TFA recommendations, Total Fat, NSS were left out), and the omitted Total Fat recommendation is the high-certainty conditional the section said never occurred. The section is structured so a reader cannot reach the pattern without first reading what it rests on.
  • Steelmanned the null reading, and it survives. A diagonal mapping is the correct output when the other three determinants do not vary is not a hedge — it is a rival explanation that the evidence here cannot rule out, and it is given its own bullet rather than buried.
  • Falsifier named in advance, so this cannot be quietly retained if contradicted — and it has already fired once, inside WHO, which is why the WHO-and-nutrition-specific fallback was rewritten. Per the AWAITS convention, the remaining test artifact is linkable and will surface via reconcile.py awaits-arrived when ingested.
  • Type check: A, and single-body. This is an emergent observation across sources, present in none of them — but the sources are not independent (one body, overlapping committees), so no [E-independent] and the claim is about a practice, not about the world.
  • Nucleus-coverage discipline observed. This section was written because the missing sources had something to say on this page’s question; the facet sources that do not (Naude, Goldenberg, Moynihan, Te Morenga, Willett) were left out of sources: rather than rolled up — which is what the generator’s own guard requires.
  • Residual: the strongest unverified step is that WHO’s strength ratings across the four guidelines were each set independently. If one guideline group applied one implicit rule across them, the ten diagonal cells are closer to one observation than to ten.

A body that has NO strength axis at all — USPSTF replaces it with certainty x magnitude [2026-07-31]

This page holds GRADE’s separation of certainty from strength. USPSTF, a major guideline body not using GRADE, dissolves the strength axis entirely and builds the recommendation from a different pair.

The output is a letter grade A/B/C/D/I read off a grid whose two axes are certainty and magnitude of net benefit — «The recommendation grade for a preventive service is derived from separate assessments of the certainty of evidence for a service and the magnitude of net benefit the service offers». Magnitude of net benefit is «substantial, moderate, small, or zero/negative». There is no strong/weak(conditional) rating anywhere. -> Net Benefit and the USPSTF Recommendation Grid (US Preventive Services Task Force, 2022)

Where GRADE’s four strength-determinants go under USPSTF:

GRADE strength determinantWhere it lands in USPSTF
balance of desirable/undesirable effectsinside net benefit — certainty is «likelihood that the… net benefit… is correct»; the balance is the graded object, not a separate weighing
certainty in the effect estimateone grid axis (3-level certainty of net benefit)
values / preferencesgrade C — «recommends selectively offering… based on professional judgment and patient preferences» — + shared decision-making, not a determinant that weakens a recommendation
resource usefinancial cost excluded; opportunity cost folded into harms («Harms may also be considered in the form of opportunity costs»)

(US Preventive Services Task Force, 2022) -> Which Objective Moved This Recommendation

The decision-relevant contrast with this page’s whole thesis. GRADE’s central move is that a weak recommendation can mean high-certainty, values vary. USPSTF cannot express that: values-variability is not a strength lever; it surfaces as grade C (a small net benefit, at least moderately certain) plus shared decision-making. So the GRADE off-diagonal (high certainty -> weak) has no USPSTF analogue — the two systems place the identical judgement (this depends on the person’s values) in structurally different outputs. A GRADE “conditional” and a USPSTF “C” are not the same quantity: GRADE conditional can be high-certainty-values-vary or thin-evidence; USPSTF “C” is specifically a confidently-small net benefit. -> GRADE vs USPSTF - Two Appraisal Systems (Schünemann et al., n.d.; inferred from US Preventive Services Task Force, 2022)

And the fourth-state handling differs at the recommendation level, not just the outcome level. GRADE permits (rarely) a strong recommendation on low certainty; USPSTF forecloses it — low certainty of net benefit is always I, «unable to assess the magnitude of net benefit». -> The Insufficient-Evidence Statement (US Preventive Services Task Force, 2022)

Self-critique of the USPSTF section [run 2026-07-31, before commit]

  • The central new claim — “GRADE conditional != USPSTF C” — is INFERRED and tagged. Verified sound: a GRADE conditional can be high-certainty-values-vary or thin-evidence; a USPSTF “C” is specifically small net benefit at ≥ moderate certainty. The two label-sets are not co-extensive, so treating them as equivalent would be the error this page’s whole thesis warns against.
  • “No USPSTF analogue of the GRADE off-diagonal” — checked, held. USPSTF has no values-variability strength lever; a high-certainty-values-vary case routes to its net-benefit magnitude (any of A/B/C) plus shared decision-making, never to a distinct “weak-despite-high-certainty” output. So the off- diagonal has no single USPSTF cell — defensible, and stated as INFERRED.
  • Independence: denied. GRADE and USPSTF share the EBM/certainty-first lineage; the determinant- placement table is a type-C catalogue of one body’s design against another’s, not type-E corroboration.
  • No fabricated equivalence. Each cell of the determinant-placement table is sourced to USPSTF’s own text (grade C, cost exclusion, net-benefit definition); GRADE’s four determinants are this page’s existing extracted content.

References

Blumenthal, R. S., Morris, P. B., Gaudino, M., Johnson, H. M., Anderson, T. S., Bittner, V. A., Blankstein, R., Brewer, L. C., Cho, L., de Ferranti, S. D., Gianos, E., Gluckman, T. J., Gradney, K. F., Isiadinso, I., Lloyd-Jones, D. M., Marrs, J. C., Martin, S. S., McLain, K. H., Mehta, L. S., … Young, J. (2026). 2026 ACC/AHA/AACVPR/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA Guideline on the Management of Dyslipidemia: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation, 153(17). https://doi.org/10.1161/cir.0000000000001423
Demay, M. B., Pittas, A. G., Bikle, D. D., Diab, D. L., Kiely, M. E., Lazaretti-Castro, M., Lips, P., Mitchell, D. M., Murad, M. H., Powers, S., Rao, S. D., Scragg, R., Tayek, J. A., Valent, A. M., Walsh, J. M. E., & McCartney, C. R. (2024). Vitamin D for the Prevention of Disease: An Endocrine Society Clinical Practice Guideline. The Journal of Clinical Endocrinology &amp; Metabolism, 109(8), 1907–1947. https://doi.org/10.1210/clinem/dgae290
Diabetes and Nutrition Study Group of the EASD. (2023). Evidence-based European recommendations for the dietary management of diabetes. Diabetologia, 66(6), 965–985. https://doi.org/10.1007/s00125-023-05894-8
European Association for the Study of the Liver. (2024). EASL–EASD–EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). Journal of Hepatology, 81(3), 492–542. https://doi.org/10.1016/j.jhep.2024.04.031
European Association of Urology, Skolarikos, A., Davis, N. F., Gambaro, G., Geraghty, R., Jung, H., & et al. (2026). EAU Guidelines on Urolithiasis.
European Society of Cardiology. (2021). 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice. European Heart Journal, 42(34), 3227–3337. https://doi.org/10.1093/eurheartj/ehab484
Evert, A. B., Dennison, M., Gardner, C. D., Garvey, W. T., Lau, K. H. K., MacLeod, J., Mitri, J., Pereira, R. F., Rawlings, K., Robinson, S., Saslow, L., Uelmen, S., Urbanski, P. B., & Yancy, W. S. (2019). Nutrition Therapy for Adults With Diabetes or Prediabetes: A Consensus Report. Diabetes Care, 42(5), 731–754. https://doi.org/10.2337/dci19-0014
Johnston, B. C., Zeraatkar, D., Han, M. A., Vernooij, R. W. M., Valli, C., El Dib, R., Marshall, C., Stover, P. J., Fairweather-Taitt, S., Wójcik, G., Bhatia, F., de Souza, R., Brotons, C., Meerpohl, J. J., Patel, C. J., Djulbegovic, B., Alonso-Coello, P., Bala, M. M., & Guyatt, G. H. (2019). Unprocessed Red Meat and Processed Meat Consumption: Dietary Guideline Recommendations From the Nutritional Recommendations (NutriRECS) Consortium. Annals of Internal Medicine, 171(10), 756–764. https://doi.org/10.7326/m19-1621
Mach, F., Koskinas, K. C., Roeters van Lennep, J. E., Tokgözoğlu, L., Badimon, L., Baigent, C., Benn, M., Binder, C. J., Catapano, A. L., De Backer, G. G., Delgado, V., Fabin, N., Ference, B. A., Graham, I. M., Landmesser, U., Laufs, U., Mihaylova, B., Nordestgaard, B. G., Richter, D. J., … Shek, A. (2025). 2025 Focused Update of the 2019 ESC/EAS Guidelines for the management of dyslipidaemias. European Heart Journal, 46(42), 4359–4378. https://doi.org/10.1093/eurheartj/ehaf190
National Academies of Sciences Engineering and Medicine. (n.d.). Redesigning the Process for Establishing the Dietary Guidelines for Americans. https://nap.nationalacademies.org/catalog/24883/redesigning-the-process-for-establishing-the-dietary-guidelines-for-americans
Nordic Council of Ministers. (2023). Nordic Nutrition Recommendations 2023: Integrating Environmental Aspects. https://pub.norden.org/nord2023-003/
Schünemann, H., Brożek, J., Guyatt, G., & Oxman, A. (n.d.). GRADE Handbook: for grading quality of evidence and strength of recommendations. https://gradepro.org/handbook/
Scientific Advisory Committee on Nutrition. (2015). Carbohydrates and Health. https://www.gov.uk/government/publications/sacn-carbohydrates-and-health-report
US Preventive Services Task Force. (2022). U.S. Preventive Services Task Force Procedure Manual. https://www.uspreventiveservicestaskforce.org/uspstf/sites/default/files/inline-files/procedure-manual-2022.pdf
Valli, C., Rabassa, M., Johnston, B. C., Kuijpers, R., Prokop-Dorner, A., Zajac, J., Storman, D., Storman, M., Bala, M. M., Solà, I., Zeraatkar, D., Han, M. A., Vernooij, R. W. M., Guyatt, G. H., & Alonso-Coello, P. (2019). Health-Related Values and Preferences Regarding Meat Consumption: A Mixed-Methods Systematic Review. Annals of Internal Medicine, 171(10), 742–755. https://doi.org/10.7326/m19-1326
World Health Organization. (2012). Guideline: Sodium intake for adults and children. https://www.who.int/publications/i/item/9789241504836
World Health Organization. (2015). Guideline: Sugars intake for adults and children. https://www.who.int/publications/i/item/9789241549028
World Health Organization. (2023a). Saturated fatty acid and trans-fatty acid intake for adults and children: WHO guideline. https://www.who.int/publications/i/item/9789240073630
World Health Organization. (2023b). Total fat intake for the prevention of unhealthy weight gain in adults and children: WHO guideline.
World Health Organization. (2023c). Use of non-sugar sweeteners: WHO guideline. https://www.who.int/publications/i/item/9789240073616