comorbidity/2026-08-22/edges

Source: data/comorbidity/2026-08-22/edges.parquet. Source vintage: Package vintage 2026-10-01; source path partition 2026-08-22; inspect record and rule clocks separately. Grain: Packaged source record; inspect schema and linked world method before aggregation. Snapshot: 7a53af6a9beb58928d4a64dc92db09c9db65433f9909f1a10d16e302adaf86ba.

Source records. NULL means missing; empty text is labelled separately.
seq Int64	outcome Utf8	stratum Utf8	from_condition Utf8	to_condition Utf8	measure_type Utf8	point Float64	low Float64	high Float64	interval_type Utf8	interval_caveat Utf8	population Utf8	ascertainment Utf8	ascertainment_means Utf8	n Utf8	follow_up Utf8	heterogeneity Utf8	pmid Utf8	grade Utf8	grade_means Utf8	quote Utf8	is_not Utf8	selection_warning Utf8	__row_id Int64
0 Filter to this value	cardiovascular_event Filter to this value	unstratified Filter to this value	MASLD / NAFLD Filter to this value	fatal or non-fatal cardiovascular event Filter to this value	HR Filter to this value	1.45 Filter to this value	1.31 Filter to this value	1.61 Filter to this value	ci95 Filter to this value	empty text	Adults aged 18 and over, mean age 53 years (SD 7), across 36 longitudinal observational studies. NAFLD was diagnosed by imaging, by International Classification of Diseases codes, OR by liver biopsy, so the pooled population's ascertainment is not uniform and the estimate belongs to none of the three alone.	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	5,802,226 individuals; 99,668 incident fatal and non-fatal CVD events Filter to this value	median 6.5 years (IQR 5.0 to 10.2) Filter to this value	I-squared 86.18 percent Filter to this value	34555346 Filter to this value	pooled_meta_analysis Filter to this value	Pooled across studies in a systematic review, with a confidence interval the authors computed. Read the heterogeneity beside it: a pooled estimate over I-squared above 75 percent is a central tendency of things that disagree.	NAFLD was associated with a moderately increased risk of fatal or non-fatal CVD events (pooled random-effects HR 1.45, 95% CI 1.31-1.61; I2=86.18%).	NOT a causal effect and NOT a treatment effect. The authors state the risks were independent of age, sex, adiposity, diabetes and other cardiometabolic risk factors, which is an adjustment and not a randomisation. It is also NOT an estimate for a screened population: a third of the contributing studies ascertained by biopsy or by diagnosis code, both of which require somebody to have already been suspected.	empty text	0
1 Filter to this value	cardiovascular_event Filter to this value	higher_fibrosis Filter to this value	MASLD / NAFLD, higher fibrosis stage Filter to this value	fatal or non-fatal cardiovascular event Filter to this value	HR Filter to this value	2.5 Filter to this value	1.68 Filter to this value	3.72 Filter to this value	ci95 Filter to this value	empty text	The subset of the same 36 studies that reported by severity. Severity here is chiefly fibrosis stage, which means the studies contributing to this estimate are the ones that staged fibrosis, which in practice means the biopsied ones.	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	a subset of the 5,802,226; the abstract does not state it separately Filter to this value	median 6.5 years (IQR 5.0 to 10.2), for the parent analysis Filter to this value	I-squared 73.84 percent Filter to this value	34555346 Filter to this value	pooled_meta_analysis Filter to this value	Pooled across studies in a systematic review, with a confidence interval the authors computed. Read the heterogeneity beside it: a pooled estimate over I-squared above 75 percent is a central tendency of things that disagree.	This risk markedly increased across the severity of NAFLD, especially the stage of fibrosis (pooled random-effects HR 2.50, 95% CI 1.68-3.72; I2=73.84%).	NOT stackable on the 1.45 above. Both are hazard ratios against the same comparator - people without NAFLD - so this REPLACES that figure for a staged population rather than multiplying it. Multiplying them would report a hazard of 3.6 that nobody measured.	The staged subset is the biopsied subset. A screening programme that finds fibrosis non-invasively is not finding the same people this figure was measured in.	1
2 Filter to this value	incident_type_2_diabetes Filter to this value	unstratified Filter to this value	MASLD / NAFLD Filter to this value	incident type 2 diabetes Filter to this value	HR Filter to this value	2.19 Filter to this value	1.93 Filter to this value	2.48 Filter to this value	ci95 Filter to this value	empty text	Adults across 26 of 33 observational studies with at least one year of follow-up, in which NAFLD was diagnosed by imaging or by biopsy. 30.8 percent of the pooled cohort had NAFLD.	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	501,022 individuals in the full review; 27,953 incident diabetes cases Filter to this value	median 5 years (IQR 4.0 to 19) Filter to this value	I-squared 91.2 percent Filter to this value	32938692 Filter to this value	pooled_meta_analysis Filter to this value	Pooled across studies in a systematic review, with a confidence interval the authors computed. Read the heterogeneity beside it: a pooled estimate over I-squared above 75 percent is a central tendency of things that disagree.	Patients with NAFLD had a higher risk of incident diabetes than those without NAFLD (n=26 studies; random-effects HR 2.19, 95% CI 1.93 to 2.48; I2 =91.2%).	NOT a bidirectional edge, despite the relationship being bidirectional in the literature. This measures NAFLD to incident diabetes only. The reverse direction is a different estimate from different studies and is not in this corpus. I-squared of 91.2 percent also means the contributing studies disagree substantially, so the interval understates the uncertainty a reader should carry.	empty text	2
3 Filter to this value	incident_type_2_diabetes Filter to this value	higher_fibrosis Filter to this value	MASLD / NAFLD, higher fibrosis severity Filter to this value	incident type 2 diabetes Filter to this value	HR Filter to this value	3.42 Filter to this value	2.29 Filter to this value	5.11 Filter to this value	ci95 Filter to this value	empty text	The five studies in the same review that reported by liver fibrosis severity. Filter to this value	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	5 studies; the abstract does not state the individual count Filter to this value	median 5 years (IQR 4.0 to 19), for the parent analysis Filter to this value	I-squared 44.6 percent Filter to this value	32938692 Filter to this value	pooled_meta_analysis Filter to this value	Pooled across studies in a systematic review, with a confidence interval the authors computed. Read the heterogeneity beside it: a pooled estimate over I-squared above 75 percent is a central tendency of things that disagree.	This risk markedly increased across the severity of liver fibrosis (n=5 studies; random-effects HR 3.42, 95% CI 2.29 to 5.11; I2=44.6%).	NOT stackable on the 2.19 above, for the same reason as the cardiovascular pair: both are against people without NAFLD. Filter to this value	Five studies, and the ones that staged fibrosis are the ones that biopsied. Does not transport to a screened population. Filter to this value	3
4 Filter to this value	liver_specific_mortality Filter to this value	unstratified Filter to this value	MASLD / NAFLD Filter to this value	liver-specific mortality Filter to this value	IRR Filter to this value	1.94 Filter to this value	1.28 Filter to this value	2.92 Filter to this value	range Filter to this value	The published uncertainty is a RANGE across contributing studies, not a confidence interval. It does not have the coverage properties a CI has and must not be reported as one.	86 studies from 22 countries, excluding studies that enrolled exclusively morbidly obese, exclusively diabetic or paediatric populations.	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	8,515,431 Filter to this value	not stated in the abstract Filter to this value	not stated for this figure Filter to this value	26707365 Filter to this value	range_across_studies Filter to this value	The published uncertainty is a RANGE across contributing studies, not a confidence interval. It does not have the coverage properties a CI has and must not be reported as one.	Incidence risk ratios for liver-specific and overall mortality for NAFLD were 1.94 (range, 1.28-2.92) and 1.05 (range, 0.70-1.56).	NOT a 95 percent confidence interval. The paper publishes a RANGE across contributing studies for this figure, which has none of the coverage properties of a CI and must not be reported as one.	empty text	4
5 Filter to this value	all_cause_mortality Filter to this value	unstratified Filter to this value	MASLD / NAFLD Filter to this value	all-cause mortality Filter to this value	IRR Filter to this value	1.05 Filter to this value	0.7 Filter to this value	1.56 Filter to this value	range Filter to this value	The published uncertainty is a RANGE across contributing studies, not a confidence interval. It does not have the coverage properties a CI has and must not be reported as one.	the same 86 studies from 22 countries Filter to this value	mixed Filter to this value	More than one of the above within a single pooled estimate. The pooled figure is over a population whose ascertainment is not uniform, so it cannot be read as belonging to any one of them.	8,515,431 Filter to this value	not stated in the abstract Filter to this value	not stated for this figure Filter to this value	26707365 Filter to this value	range_across_studies Filter to this value	The published uncertainty is a RANGE across contributing studies, not a confidence interval. It does not have the coverage properties a CI has and must not be reported as one.	Incidence risk ratios for liver-specific and overall mortality for NAFLD were 1.94 (range, 1.28-2.92) and 1.05 (range, 0.70-1.56).	This is the NULL RESULT in the corpus and it is registered precisely because it is inconvenient. The range spans 1.00, so this meta-analysis did not find NAFLD associated with all-cause mortality. A savings argument that quotes the liver-specific 1.94 and omits this one has selected its evidence.	empty text	5

Evidence receipt r21

Permanent saved result

Secondary typed Arrow export