Minor grumble - we have no idea how most people with ME/CFS die - we don't even know what age they die at or what old age is for people with ME/CFS. Suicide may be a significant risk in ME/CFS but we know so little about the whole patient population that saying things like suicide is one the three leading causes of death is unjustified when the samples are so low.
Elliot and Jason quote two studies -
1) Roberts, Wessely, Chalder, Chin-Kuo & Hotopf (boo hiss etc but the numbers look right in this study)
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(15)01223-4/fulltext
"
Results
We identified 2147 cases of chronic fatigue syndrome in CRIS with 17 deaths. Of them, 1533 patients were women of whom 11 died, and 614 were men of whom six died. Eight deaths were from malignant neoplasm, five from suicide, and four from other causes.
There was no significant difference in age-standardised and sex-standardised mortality ratios for all-cause mortality (SMR 1·14; 95% CI 0·5–1·85; p=0·67) or cancer-related mortality (1·39; 0·60–2·73; p=0·45). This remained the case when stratified by sex, and when those deaths from external causes were removed from the analysis. However, there was a significant increase in suicide mortality with an SMR of 6·85 (95% CI 2·22–15·98; p=0·002;
table 1). Although the suicide-specific SMR was significantly increased compared with the general population, if there had been two fewer deaths by suicide, this result would have been non-significant, although the effect size (SMR>4) would still be indicative of a strong effect.
Table 1 shows detailed SMRs for the study cohort."
and
2) Jason, Corradi, Gress, Williams & Torres-Harding
https://sci-hub.st/10.1080/07399330600803766 which uses National CFIDS Foundation Memorial List a cohort of just 144 deaths.
"Table 1 presents the reason for death among the 144 individuals where this information was available. The three leading causes of death were heart failure, suicide, and cancer, accounting for 59.6% of cases. In regard to gender, 74.4% of this sample were female and 25.6% were male, and this difference was significant at the p < .01 level using a "