Research landscape
The shape of ME/CFS research, in numbers.
Every figure on this page is a direct count of the 6,211 public atlas records. Design tiers, PEM status and domain assignment are machine labels derived from abstracts — useful for orientation, not a substitute for reading the studies.
1 · Timeline
Studies per publication year
Stacked by study design tier. The rise after 2020 reflects both Long COVID research and broader indexing of recent literature.
The current year is incomplete.
Show the numbers
| Year | E0 | E1 | E2 | E3 | Total |
|---|---|---|---|---|---|
| 2026 | 31 | 10 | 110 | 55 | 206 |
| 2025 | 33 | 14 | 160 | 162 | 369 |
| 2024 | 36 | 21 | 111 | 137 | 305 |
| 2023 | 30 | 12 | 133 | 131 | 306 |
| 2022 | 19 | 7 | 124 | 119 | 269 |
| 2021 | 32 | 8 | 119 | 132 | 291 |
| 2020 | 31 | 8 | 100 | 78 | 217 |
| 2019 | 15 | 7 | 99 | 91 | 212 |
| 2018 | 16 | 5 | 101 | 88 | 210 |
| 2017 | 16 | 22 | 91 | 60 | 189 |
| 2016 | 14 | 13 | 89 | 72 | 188 |
| 2015 | 15 | 18 | 92 | 77 | 202 |
| 2014 | 16 | 9 | 63 | 67 | 155 |
| 2013 | 11 | 21 | 78 | 73 | 183 |
| 2012 | 5 | 10 | 89 | 81 | 185 |
| 2011 | 14 | 12 | 114 | 74 | 214 |
| 2010 | 5 | 14 | 92 | 70 | 181 |
| 2009 | 8 | 8 | 82 | 62 | 160 |
| 2008 | 8 | 7 | 70 | 55 | 140 |
| 2007 | 7 | 6 | 51 | 80 | 144 |
| 2006 | 5 | 5 | 58 | 63 | 131 |
| 2005 | 6 | 9 | 66 | 56 | 137 |
| 2004 | 3 | 16 | 66 | 58 | 143 |
| 2003 | 3 | 5 | 67 | 51 | 126 |
| 2002 | 6 | 7 | 65 | 66 | 144 |
| 2001 | 5 | 9 | 50 | 59 | 123 |
| 2000 | 2 | 2 | 57 | 37 | 98 |
| 1999 | 0 | 3 | 53 | 59 | 115 |
| 1998 | 0 | 7 | 69 | 48 | 124 |
| 1997 | 1 | 7 | 55 | 49 | 112 |
| 1996 | 0 | 7 | 52 | 55 | 114 |
| 1995 | 0 | 0 | 50 | 40 | 90 |
| 1994 | 0 | 2 | 48 | 48 | 98 |
| 1993 | 1 | 2 | 27 | 34 | 64 |
| 1992 | 0 | 2 | 24 | 49 | 75 |
| 1991 | 1 | 1 | 22 | 59 | 83 |
| 1990 | 0 | 2 | 10 | 16 | 28 |
| 1989 | 0 | 1 | 6 | 11 | 18 |
| 1988 | 0 | 1 | 6 | 16 | 23 |
| 1987 | 0 | 0 | 2 | 2 | 4 |
| 1986 | 0 | 0 | 1 | 1 | 2 |
| 1985 | 0 | 0 | 0 | 3 | 3 |
| 1984 | 0 | 0 | 1 | 1 | 2 |
| 1983 | 0 | 0 | 0 | 1 | 1 |
| 1982 | 0 | 0 | 0 | 0 | 0 |
| 1981 | 0 | 0 | 0 | 0 | 0 |
| 1980 | 0 | 0 | 1 | 2 | 3 |
Study design mix
How the 6,211 public studies in this atlas split by study-design tier. The tier describes the kind of study, assigned by machine from the abstract. It is not a quality rating and does not mean a finding has been replicated.
2 · Evidence map
Research domains × study design tier
Bubble area is the share of each domain’s studies at that level. Select a count to open those studies.
Lowest share of reviews or replicated studies (E0–E1): Metabolic Dysfunction, Immune Dysregulation, Neuroinflammation.
A study can belong to several domains; 15 domains in total.
| Research domain | E0Synthesis | E1Randomised | E2Observational | E3Other | Studies | PEM req. |
|---|---|---|---|---|---|---|
| Immune Dysregulation | 45 | 34 | 460 | 672 | 1211 | 6% |
| Post-Exertional Malaise | 57 | 16 | 316 | 359 | 748 | 43% |
| Diagnostics | 38 | 0 | 230 | 274 | 542 | 5% |
| Cognitive Impairment | 21 | 18 | 249 | 119 | 407 | 5% |
| Autonomic Nervous System | 17 | 11 | 209 | 126 | 363 | 3% |
| Biomarkers | 18 | 4 | 190 | 100 | 312 | 12% |
| Sleep Dysfunction | 21 | 19 | 157 | 84 | 281 | 2% |
| Pain and Sensitization | 14 | 13 | 126 | 102 | 255 | 2% |
| Interventions | 42 | 52 | 54 | 76 | 224 | 2% |
| Metabolic Dysfunction | 8 | 5 | 73 | 128 | 214 | 7% |
| Neuroinflammation | 11 | 1 | 34 | 137 | 183 | 8% |
| Endocrine Disruption | 10 | 5 | 55 | 64 | 134 | 0% |
| Long COVID Overlap | 8 | 1 | 42 | 28 | 79 | 4% |
| Gut Microbiome | 8 | 3 | 12 | 21 | 44 | 5% |
| Severe ME | 3 | 1 | 10 | 19 | 33 | 36% |
3 · Case definition
How often did cohorts require post-exertional malaise?
Among ME/CFS, post-infectious and overlapping-population studies from the last five years, 17% are labelled as having required PEM for inclusion.
“Unclear” usually means the abstract does not state the criteria — not that PEM was absent. Criteria that do not require PEM (e.g. Oxford, Fukuda) can include people without ME/CFS as defined today. Why criteria matter
4 · Relationships
Which domains are studied together
Number of studies assigned to both domains, and what share of the smaller domain that represents.
Overlap means the same studies address both areas. It does not show a biological or causal link.
- Immune Dysregulation and Post-Exertional Malaise121 studies · 16% of the smaller domain
- Biomarkers and Immune Dysregulation112 studies · 36% of the smaller domain
- Neuroinflammation and Immune Dysregulation79 studies · 43% of the smaller domain
- Pain and Sensitization and Sleep Dysfunction46 studies · 18% of the smaller domain
- Metabolic Dysfunction and Immune Dysregulation42 studies · 20% of the smaller domain
- Autonomic Nervous System and Cognitive Impairment40 studies · 11% of the smaller domain
- Sleep Dysfunction and Cognitive Impairment40 studies · 14% of the smaller domain
- Biomarkers and Post-Exertional Malaise39 studies · 13% of the smaller domain
- Pain and Sensitization and Cognitive Impairment37 studies · 15% of the smaller domain
- Metabolic Dysfunction and Post-Exertional Malaise30 studies · 14% of the smaller domain
- Neuroinflammation and Post-Exertional Malaise25 studies · 14% of the smaller domain
- Neuroinflammation and Biomarkers25 studies · 14% of the smaller domain
- Biomarkers and Metabolic Dysfunction24 studies · 11% of the smaller domain
- Gut Microbiome and Immune Dysregulation23 studies · 52% of the smaller domain
5 · Paradigm
Research paradigm over time
Biomedical, psychosomatic and neutral framing of studies by year (machine label; 3-year rolling average).
Research paradigm over time
How the share of biomedical vs. psychosomatic vs. neutral research has changed year by year within the 6,211 public studies in this atlas. Values are a 3-year rolling average of the paradigm mix by publication year, not raw counts.
This describes the atlas’s corpus composition over time, not the overall ME/CFS research field. The “biomedical” bucket is a classifier-assigned label for studies with explicitly biomedical framing; many biologically relevant observational and review studies sit in “neutral” instead. Studies without a publication year in 1990–2026 are not plotted. How we classify paradigms →
Data table
| Year | Biomedical % | Psychosomatic % | Neutral % |
|---|---|---|---|
| 1990 | 50 | 25 | 25 |
| 1991 | 53 | 21 | 26 |
| 1992 | 55 | 20 | 25 |
| 1993 | 62 | 16 | 22 |
| 1994 | 61 | 19 | 20 |
| 1995 | 61 | 20 | 20 |
| 1996 | 56 | 22 | 22 |
| 1997 | 60 | 21 | 19 |
| 1998 | 58 | 21 | 22 |
| 1999 | 58 | 19 | 22 |
| 2000 | 54 | 21 | 25 |
| 2001 | 53 | 24 | 23 |
| 2002 | 54 | 24 | 22 |
| 2003 | 53 | 25 | 21 |
| 2004 | 55 | 23 | 22 |
| 2005 | 55 | 23 | 21 |
| 2006 | 58 | 21 | 20 |
| 2007 | 57 | 23 | 20 |
| 2008 | 56 | 24 | 20 |
| 2009 | 57 | 22 | 21 |
| 2010 | 60 | 20 | 20 |
| 2011 | 63 | 18 | 19 |
| 2012 | 61 | 21 | 18 |
| 2013 | 62 | 22 | 16 |
| 2014 | 57 | 24 | 19 |
| 2015 | 58 | 21 | 21 |
| 2016 | 56 | 21 | 23 |
| 2017 | 62 | 18 | 20 |
| 2018 | 63 | 17 | 20 |
| 2019 | 66 | 15 | 20 |
| 2020 | 65 | 14 | 21 |
| 2021 | 67 | 12 | 21 |
| 2022 | 69 | 10 | 21 |
| 2023 | 68 | 10 | 22 |
| 2024 | 69 | 11 | 20 |
| 2025 | 69 | 10 | 21 |
| 2026 | 71 | 10 | 19 |
Data: public ME/CFS Atlas records, recomputed hourly. Download the underlying records from the evidence explorer (CSV/JSON export). Methodology