Typing myalgic encephalomyelitis by infection at onset: A DecodeME study

Patient and Public Involvement

The DecodeME project grew out of the UK ME Research Collaborative (MERC), formerly known as the CFS/M.E. Research Collaborative or CMRC, which was first established in 2013. The MERC includes people with ME/CFS and carers within a Patient Advisory Group (PAG). As the project evolved in 2018–19, PPI was embedded in every discussion and workshop, resulting in the project becoming a co-production with its grant proposal, aims and outcomes being decided by researchers and PPI in equal measure. In 2020, PPI Steering Group members were selected from across diverse charities and organisations, and for their breadth of experience. In DecodeME, PPI representatives serve on each of its delivery groups, lead on marketing and communication (including social media), and contribute the majority (two of three) members of the decision-making body, the Management Group. People with lived experience of ME/CFS led the co-creation of a new DecodeME questionnaire, making substantial improvements in comprehension and accuracy, thereby boosting recruitment. The project’s name was suggested and decided by PPI members.

DecodeME’s genetics question (“What, if any, significant genetic differences are there between people with—and those without—ME/CFS?") was identified as a priority first by the MERC and its PAG, before being confirmed as a priority by a wider section of the patient community in the results of the Priority Setting Partnership for ME/CFS¹⁰. Established participant selection criteria were further refined with PPI throughout. PPI members, through their profound understanding of ME/CFS phenotypes, triggers, severity, symptom range, comorbidities and more, have improved the study’s adherence to our chosen case definition and thus further assured the relevance of genetic associations to ME/CFS lived experience.

A substantial minority of volunteer participants who trialled an initial questionnaire reported difficulties when answering its questions. We then created a new version implementing Canadian Consensus and IOM/NAM criteria^3,13 as well as criteria introduced in response to peer reviewers’ comments. This DecodeME questionnaire is freely available from the DecodeME website. As a co-production, PPI members advised and helped to create both our recruitment strategy and recruitment materials. Further description of DecodeME’s recruitment methods and PPI aspects can be found elsewhere¹⁴. Before study launch public awareness of DecodeME was enhanced using regular podcasts, webinars, blog posts and media interviews. These media channels will be used by PPI members and scientists to disseminate results to the international ME/CFS community. PPI team members maintain extensive input into reporting of the results of the questionnaire (including in this article), providing greater understanding and context, and ensuring accessibility. Our GWAS plan was co-created by researchers and PPI members.

The DecodeME study was reviewed and given a favourable opinion by the North West – Liverpool Central Research Ethics Committee (21/NW/0169). Potential participation bias due to internet use was mitigated by providing a paper questionnaire and providing participants with assistance in completing their online questionnaires. Team members were available to answer phone calls and emails during working hours.

Cohort

Participants were asked for their sex assigned at birth and about their conditions: “If a health professional has ever told you that you had any of the conditions below, please select all that apply. If the conditions don’t apply to you, please do not select any box.” Participants indicated whether each condition was Active (“If the condition has given you symptoms in the past 6 months”) or not active “If the condition has not given you symptoms in the past 6 months, either because it has died down or treatment has controlled it”). They were also asked about their symptoms: “In the last 6 months, have you had any of the symptoms below often, repeatedly, or constantly? Please mark any that apply. If none apply, leave all the boxes blank.” Questionnaire responses from participants who both consented to participate and self-reported being given a diagnosis of ME, CFS, ME/CFS or CFS/ME by a health professional (as of 19 December 2022) were analysed. Only those whose sex assigned at birth was male or female were analysed due to insufficient numbers of other identities. Participant ages were as of 19 December 2022. Further analyses will be undertaken for the full DecodeME cohort once the recruitment phase of the project is completed.

Significance testing. Logistic regression analysis for Figure 6, for example, was of the form: OnsetType ~ age + sex + symptoms + intercept. The analyses conducted were: (i) for 80 symptoms against age and sex – Figure 4; (ii) for each symptom (n=80; and, age and sex) against severity – Figure 5; (iii) for each of the 5 onset types against 8 fatigue symptoms plus age and sex – Figure 6A; (iv) for each of the 5 onset types against 72 non-fatigue symptoms (plus age and sex) – Figure 6B; (v) for each of the 5 onset types against 5 illness courses, relative to ‘Fluctuating’, the majority response – Figure 6C; and (vi) for each of the 5 onset types against 34 comorbidities (active and inactive) plus age and sex – Figure 7. Analyses used the glm function in R version 4.2.2. Only p-values that survive Bonferroni correction for multiple tests per analysis are shown.

Demographics

The DecodeME study continues to recruit individuals living in the UK aged 16y and above, with its oldest participant aged over 90y old. Male participants tended to be older than females (median 52y and 48y respectively; p < 2.2×10^-16, Wilcoxon rank sum test). The median age of study participants was 49y, similar to those in previous USA-based studies^18–20. Only 3.3% of 17,074 participants did not self-report their ethnicity as White, far fewer than the 18.3% in England and Wales who identify as non-White (https://www.ethnicity-facts-figures.service.gov.uk/). Participants indicated the duration of their ME/CFS illness by selecting from a set of predefined ranges, for example between 5 and 10 years, or over 10 years, since onset of symptoms. Most (61.3%) DecodeME participants have had ME/CFS for over 10y, and 81.5% over 5y (Figure 3). Together, study participants have experienced over 1.3 ×10⁵ years of ME/CFS symptoms.

Participants who started their illness within the last 1–3y or 0.5–1y numbered 1,287 or 354, respectively. These numbers are 57% and 21% fewer, per year, than the study’s 1,634 participants from the 3–5y recruitment interval. This paucity of participants with illness duration within 3 years may reflect how long it usually takes to receive a clinical diagnosis in the UK (median 13–24 months).

To examine the incidence of ME/CFS, we considered 3,150 participants reporting ME/CFS onset within the last 5y. Their median age was 40y, overlapping the peak age (40–59y) of ME/CFS diagnosis in UK primary care²¹. On average, those with glandular fever onset <5y ago were a decade younger than all others (median ages 30.5y and 41y, respectively; p < 2.2×10^-16, Wilcoxon rank sum test). This means that for at least half of our glandular fever onset participants, ME/CFS onset occurred after the age of 25y. This is a decade after peak incidence of glandular fever in the UK between 15 and 19y old²². This difference in peak incidence is consistent with adolescents being less likely, than older people, to develop ME/CFS after glandular fever.

Sex- and age-bias of ME/CFS comorbidities and symptoms

In the DecodeME cohort, females outnumber males by over five-to-one (83.5% females; 16.5% males). This is among the highest female-bias among those with ME/CFS yet reported internationally^{3,9,21,23–27}. Despite this strong bias, the substantial number of males participating in DecodeME (N=2,827) allowed the study to reveal previously unreported sex-biases in comorbidities or symptoms.

Females with ME/CFS reported more comorbidities and symptoms than males in the DecodeME questionnaire. Two-thirds (66.7%) of females, but a half (52.7%) of males, reported at least one active comorbidity; similarly 39.2% of females and 28.6% of males reported at least one inactive comorbidity. Female participants reported, on average, more symptoms than males (42 versus 36).

To test more formally for an association between age and sex and each symptom we used logistic regression using the Bonferroni correction to adjust for multiple tests (Methods). This identified 62 of 80 symptoms as significantly female-biased, and 61 as biased towards younger age (Figure 4). Female-bias is evident across all symptom types (Figure 4). Females were significantly more likely to report fatigue ‘often, repeatedly, or all the time’ (p=6×10^-4; age p=1.1×10^-13), and more likely to report post-exertional malaise after physical or mental activity (p=3×10^-4; age p=4.2×10^-7).

Figure 4. Most symptoms are strongly associated with female sex at birth and younger age.

The question asked was: “In the last 6 months, have you had any of the symptoms below often, repeatedly, or constantly? Please mark any that apply.” Sex-biased (X-axis) and/or Age-biased (Y-axis) associations in a logistic regression analysis (Symptom ~ age + sex + intercept) are shown as data points. Data points within the blue-shaded areas are not significant after accounting for 82 tests (p<0.05/82, or |Z|<3.427. Only one symptom (“Feeling easily annoyed or irritable”) was male-biased; 3 symptoms (sensitivities to chemicals or medicine, or bladder problems) were associated with older age. Results for 80 symptoms are shown.

Severity, comorbidities and symptoms

Participants were asked: “How severe is your illness?” with answer options matching severity definitions from the UK’s National Institute for Health and Care Excellence (NICE) guidelines (2021). Most DecodeME participants’ severity levels are categorised as Mild or Moderate, but Severe and Very Severe individuals are also represented (Figure 3). Severity categories were consistent with participants’ reports of their comorbidities and symptoms (see below).

Being female, increasing age and being over 10y from ME/CFS onset are each separately associated with severity in the DecodeME cohort (sex: p=4.5×10^-4; age: p<2.2×10^-16; years since ME/CFS onset: p=1.6×10^-6). These results are from a comparison of those with mild ME/CFS (34%) against the remaining 66% with moderate, severe or very severe illness. Testing for all 68 co-occurring (active and inactive) comorbidities, and including both age and sex as covariates in the model, 6 active comorbidities were significantly associated with severity. In order of decreasing significance these were: fibromyalgia (p<2×10^-16), clinical depression (p<2×10^-16), irritable bowel syndrome (p=5.7×10^-12), mast cell activation syndrome (p=1.8×10^-11), diabetes (p=9.5×10^-10) and sleep apnoea (p=5.2×10^-8). Severity was also associated with a single inactive comorbidity, hypothyroidism (p=1.6×10^-5).

Testing all symptoms simultaneously with sex and age, showed strong and independent association between ME/CFS severity and 18 factors including fatigue, age, difficulty remaining standing, and sleep problems (Figure 5). Finally, participants describing their illness as relapsing and remitting were significantly less likely to report their illness as moderate, severe or very severe than those reporting fluctuating symptoms (p<2.2×10^-16).

Figure 5. Questionnaire responses that significantly associate with ME/CFS symptom severity.

Z-scores are shown for symptoms that significantly associate with severity (p<0.05 after Bonferroni correction for 82 tests, including age and sex). Here severity is defined by self-report of moderate or severe or very severe symptoms versus self-report of mild symptoms (see Figure 3). Responses to questions 14 and 15 (Q14, Q15) are significantly associated with mild symptoms. Responses relate to DecodeME Questionnaire questions (e.g. question 10, Q10).

The type of infectious or non-infectious disease onset does not explain these strong and pervasive sex-biases because across the five onset types proportions of females were not significantly different (83.1%-84.5%; χ² = 1.707, df = 4, p = 0.79).

ME/CFS characteristics by onset type

A feature that strongly distinguished among the five onset types was longevity of participants’ ME/CFS symptoms. Participants reporting an infection at onset were more likely to have had ME/CFS symptoms for over 10y than those reporting no infection at onset (67.0% vs. 43.6%). This is despite their similar ages (medians 50y and 46y, respectively).

The statistical significance of this difference is strong. When testing for association between those with an infection prior to ME/CFS onset and duration (<10y vs. >10 years since time of onset), age and sex, only association with duration was significant (p = 4×10^-67). This relative paucity of participants not reporting an infection prior to onset of their ME/CFS over 10y ago is unexpected, and not easily explained by historic variation in ME/CFS triggers because association with age was not significant in this analysis (p > 0.05). When analysed separately, each onset type was not associated with participants’ sex, when including age and ME/CFS duration over 10y in the analysis.

There were significant differences between the 5 ME/CFS onset types and 4 fatigue symptoms (Figure 6A), 16 other symptoms (Figure 6B) and 3 different types of illness course (Figure 6C). Those with glandular fever onset were significantly more likely than others to report swollen or tender glands and viral infections with long recovery periods within the last 6 months, and to experience relapsing and remitting symptoms (relative to ‘Fluctuating’, the majority response). Others with COVID-19 infection at ME/CFS onset preferentially reported a tight feeling in the chest, sensitivity to alcohol and a feeling of burning in the lungs. Participants with other types of infection onset more frequently reported feeling mentally fatigued, feeling fatigued less than half the time, and difficulties remaining standing, and less frequently reported feeling more sleepy than is normal, having worsening symptoms (relative to ‘Fluctuating’), unusual changes in appetite and mood swings.

Figure 6.

Associations of symptoms or age to 5 ME/CFS onset types: (A) Fatigue symptoms (10 tests), (B) Non-fatigue symptoms (74 tests), and (C) Illness course descriptions (7 tests). These were considered in a logistic regression model of the form OnsetType ~ age + sex + symptoms/descriptions and an intercept. A covariate is only shown if it survived Bonferroni multiple testing correction (p<0.05) per regression for one or more symptom/description. Significant associations are indicated with an asterisk (*); their Z-scores lie outside of non-significant values, bounded by the red dashed lines, after Bonferroni multiple testing correction. The z-score (Y-axis) is the effect-size estimate in standard deviation units.

Participants reporting an infectious onset (when compared to those who did not) were also significantly more likely to report: improving symptoms, relapsing/remitting, or recovered (relative to ‘Fluctuating’) symptoms, and less likely to report worsening symptoms (again, relative to ‘Fluctuating’). They were more likely, among other things, to report viral infections with long recovery periods, fewer viral infections than they used to get, and having a pale face. Other symptoms that were significantly more likely to be reported by participants without an identified infection at onset were fatigue more than half the time, reduced libido, and unusual changes in appetite. They were also less likely to report symptoms common during infection: flu-like feelings, and swollen or tender glands.

Those with an infection prior to onset of ME/CFS more frequently reported symptoms typical of infection in the last 6 months, whereas those reporting no infection at onset less frequently indicated these symptoms. This was unexpected because of the long time-lag between onset (mostly >10y ago) and participants’ recent questionnaire responses. Those with an infection prior to onset of ME/CFS frequently reported symptoms typical of infection in the last 6 months, whereas those reporting no infection at onset infrequently indicated these symptoms. Even though most participants report a long interval between their onset of ME/CFS (mostly >10y ago) and their recent symptoms characteristic of infection, our results cannot distinguish between whether these recent symptoms are a natural consequence of their ME/CFS onset, for example because of viral persistence in some individuals²⁸, or else they are independent of onset.

In our last analysis, we tested for association between participants’ onset type and their comorbidities, age and sex. Only younger age, rather than any comorbidity, was significantly associated with glandular fever onset (Figure 7). Among all onset types, only coronavirus disease 2019 (COVID-19 caused by SARS-CoV-2) infection was significantly associated active Mast Cell Activation Syndrome (MCAS), i.e. MCAS symptoms within the previous 6 months. COVID-19 related onset was also negatively associated with active fibromyalgia. Onset with another infection was positively associated with inactive Shingles or active Lyme disease, and negatively associated with fibromyalgia or clinical depression. Onset without reported infection at onset was significantly associated with recent clinical depression symptoms; and, onset with unknown infection status was significantly associated with active fibromyalgia as a comorbidity (Figure 7).

Figure 7. Associations of comorbidities or age to 5 ME/CFS onset types.

Thirty-four comorbidities were considered in a logistic regression model of the form OnsetType ~ age + sex + comorbidities and an intercept. A covariate is only shown if it survived Bonferroni multiple testing correction (p<0.05) for one or more onset type. Active and inactive comorbidities were considered independently: Active, if the condition has given symptoms in the past 6 months, or Inactive, if the condition has not given symptoms in the past 6 months, either because it has died down or treatment has controlled it. The z-score (Y-axis) is the effect-size estimate in standard deviation units.