How Common Are Eating Disorders? Incidence, Prevalence and Mortality Rates

Six month of blogging and I have yet to do a proper post on the prevalence of eating disorders. I think it is about time. I see all sorts of numbers thrown around, often depending on the purpose of the article and the author’s bias. Is it 1 in 1000, 1 in 100, 1 in 20 or maybe even 1 in 2? Who is right?

Well, it is a tricky question to answer.

The number depends on how the particular study was conducted. Here are some factors that may influence the final rates: the population being studied, the sample size, the definition of eating disorder, the methods used by researchers to identify and screen for individuals with eating disorders, the number of years over which data is collected, and so on. In other words, a lot! That’s why in order to get a better sense of the true numbers, I prefer to look at review articles summarizing several years worth of epidemiological studies.

For this post, I picked a relatively recent review, published in May 2012, by Smink, van Hoeken and & Hoek, summarizing the incidence, prevalence and mortality rates in eating disorders. Keep in mind, Smink et al limited their review to studies published in English. I want to include a lot of the incidence, prevalence and mortality figures primarily because I want to give you a sense of just how varied these numbers are, and just how much depends on the study parameters. (I think this is important to properly assess articles in the popular press, which often exaggerate or minimize the numbers by selecting studies where the findings suit their needs.)


The incidence rate is the number of new cases of a disorder in the population over a specified period. The incidence rate of eating disorders is commonly expressed in terms of per 100 000 persons per year (person-years).

Anorexia Nervosa

Community sample (includes those who may not seek/receive treatment):

  • 270 /100 000 persons per year in 15-19 year old Finnish female twins, during 1990-1998 (Keski-Rankonen et al., 2008)
  • 470 person-years for “broad AN” (includes AN without amenorrhea and ICD-10 atypical AN), using same sample as above (Keski-Rankonen et al., 2008)
  • 1204/100 000 person-years  for AN in females aged 15-18, in a study of 595 adolescents (ie, small sample size) (Isomaa et al., 2009)

General practice/primary care sample (earliest stage of detection by health care system):

  • in the UK: 4.7/100 000 person years from 1994-2000 and 4.2/100 000 person years from 1988-1993 (Currin et al., 2005)
  • in the Netherlands: 7.7/100 000 person years from 1995-1999 and 7.4/100 000 person years from 1985-1989 (van Son et al., 2006)
  • among children younger than 13, rate of AN was 1.1 per 100 000 person years in Britain (Nicholls et al., 2011)

Note the higher rates of AN when evaluated in a community sample versus general practice. There are lots of reasons for this trend. Saren has written about one contributing factor: the barriers minority groups face in seeking and receiving treatment.

Is anorexia on the rise?

The question of whether the incidence of AN is on the rise has been under debate. Long-term epidemiological studies are sensitive to minor changes in the absolute incidence numbers and in the methods used, for example, variations in registration policy, demographic differences between the populations, faulty inclusion of re-admissions, the specific methods of detection used or the availability of services…. Since 1970, the incidence of AN in Europe seems to have been rather stable… (van Son et al., 2006; Hoek., 2006; Theander, 1985; Kendell et al., 1973)

Bulimia Nervosa

Community sample

  • 200/100 000 person-years from 1975-1979 among Finnish female twins (peak age similar to AN, 16-20, versus 15-19 for AN) (Keski-Rankohen et al., 2009)
  • 438/100 00 person-years in 15-18 year old Finnish females for “broad BN” (all but one criteria) (Isomaa, et al., 2009)

General practice/primary care sample

  • in the Netherlands: 8.6/100 000 person-years in 1985-1989 and 6.1/100 000 in 1995-1999 (nation-wide study) (van Son et al., 2006)
  • in the UK: 12.2/100 000 person-years in 1993 and 6.6/100 000 person-years in 2000 (although incidence in 10-19 year old women remained stable during that time, around 40/100 000 person-years) (Currin et al, 2005)


EDNOS is a catch-all diagnosis, a diagnosis for everything that looks like an eating disorder but doesn’t fit the narrow and unscientific DSM-IV criteria for AN and BN. It is, by it is very nature, very heterogeneous, so take the studies with a grain of salt. 

  • 6.5/100 000 persons per year in a Spanish population-based study (using Public Health Registry to identify EDs diagnosed by mental health professionals) (Larranaga et al., 2012)
  • 1.2/100 000 persons per year among children 13 or younger, in a British study (Nicholls et al., 2011)


The prevalence can be expressed as point prevalence, one year prevalence rate and lifetime prevalence. The point prevalence is the prevalence at a specific point in time, e.g. January 1 of a specific year. The one-year prevalence rate is the point prevalence plus annual incidence rate (the number of new cases in the following year). The lifetime prevalence is the proportion of people that had the disorder at any point in their life. The prevalence is the most useful measure for planning health care facilities, as it indicates the demand for care.

Anorexia Nervosa

  • 1.2% in a cohort of twins born from 1935-1958 (Bulik et al., 2006)
  • 1.9% in female twins aged 28-39 (Wade et al., 2006)
  • 2.2% in sample of women born between 1975-1979; 0.24% in a sample of men from the same cohort (Keski-Rahkonen et al., 2007)
  • 0.6% in a sample of 496 adolescent girls, followed for 8 years (Stice et al., 2009)
  • 0.39% among adolescent girls (point prevalence), in Portugal (Machado et al., 2007)
  • large population samples: 0.9% in adult females, 0.3% among males (Hudson et al., 2007) 0.3% in 1318 year old females (Swanson et al., 2011)

I just want to point out, many of the above studies followed twins, and so, I’d suspect, given that we know first-degree relatives have higher rates of EDs than the general population, that the numbers coming out from twin studies may be inflated (and you can see, they are higher than those that are taking random population samples).

Well, who would’ve thought, we are underdiagnosing males:

… many recent community-based studies have found that AN is more common among males than previously thought. AN may be even more frequently underdetected in males than in females (Keski-Rahkonen et al., 2008)

Bulimia Nervosa

  • 1.7% lifetime prevalence in female twins born from 1975-1979, 2.3% if reduce frequency of bingeing and purging to 1x/week (Keski-Rahkonen et al., 2009)
  • 2.9% in twin cohort of women from 28-39 years old (Wade et al., 2006)
  • large scale population studies: 0.9-1.5% among women, and 0.1-0.5% among men (Hudson et al., 2007; Preti et al., 2009)
  • 1.6% in a sample of 496 adolescent girls, followed for 8 years (Stice et al., 2009)

Next time you see headlines about an increase in EDs, ask yourself: is that just because the DSM relaxed its criteria? (to reduce the number of cases that fall into EDNOS):

Trace and colleagues assessed the impact of reducing the binge eating frequency on the lifetime prevalence of BN in a large population sample of female Swedish twins. The lifetime prevalence of BN increased from 1.2 % for a minimum of 8 binges per month (DSM-IV) to 1.6 % for at least 4 binges per month (proposal DSM-5) (Trace et al., 2011)

Bulimia, on the decrease?

The decrease in occurrence of BN over time found in the incidence studies is supported by a US study of university students in which the point prevalence of BN among women significantly decreased from 4.2 % in 1982, to 1.3 % in 1992 and 1.7 % in 2002 (Keep et al., 2006). In another US study among female students the point-prevalence of probable cases of BN remained relatively stable between 1990 and 2004 (Crowther et al., 2008).


Again, because different researchers use different definitions of EDNOS, and many don’t even assess for it, take the numbers with a grain of salt.

  • 2.4%, point prevalence in nation-wide community sample in young females (Machado et al., 2007)


One could describe the mortality rate as an incidence rate in which the event being measured is death. Mortality rates are often used as one of the indicators of illness severity… The crude mortality ratio (CMR) is the number of deaths within the study population over a specified period. The standardized mortality ratio (SMR) is the ratio of observed deaths in the study population to expected deaths in the population of origin.

Anorexia Nervosa

  • in meta-analysis of 35 published studies, CMR was 5.1 deaths per 1000 persons per year (or 0.51% per year; one in five committed suicide (Arcelus et al., 2011)
  • SMR was 5.86 with follow-up of 14 years (Arcelus et al., 2011)
  • meta-analysis of SMRs for 6-12 year follow-up studies was 9.6, in studies for 20-40 year follow-up: 3.7 (Arcelus et al., 2011)

Duration of follow-up is inversely correlated with SMR: the longer the follow-up, the higher the mortality in the population, thus, the ratio of deaths due to AN versus general population decreases. Other factors that influence mortality numbers: age, severity of patients, study period, the reporting of AN as cause of death in death certificates (may be over- and under-reported). 

Bulimia Nervosa

  • in a meta-analysis of 12 studies, mortality rate: 1.74 per 1000 persons per year (0.17 patients de a year); SMR was 1.93 (Arcelus et al., 2011)
  • in a sample of 1885 patients, 35/906 individuals with BN (3.9%) died in a mean follow-up of 19 years (not sure what the population average is though, it is kind of irrelevant without a comparison to the non-ED population) (Crow et al., 2009)
  • out of the 35 deaths, suicide accounted for 23% (Crow et al., 2009)


  • in a meta-analysis of 6 studies, mortality rate: 3.31 per 1000 persons per year, SMR: 1.92 (Arcelus et al., 2011)

“Any elevated mortality risk of EDNOS could be partly explained by the assertion that EDNOS sometimes reflects the earlier stages of AN” Or frankly, that the notion that EDNOS is “less severe” is faulty. It may be less severe, but, it need not be.


The authors conclude by summarizing the key points in this review article. Namely, it appears that overall, the incidence of anorexia nervosa has been stable, but, there’s been an increase among 15-19 year old female. But, whether this reflects an earlier age of onset or earlier detection is not clear. It seems that bulimia nervosa might have actually decreased since the early 90’s. Eating disorders, particularly anorexia nervosa, have high mortality rates.

We are still likely underdiagnosing males with eating disorders. And, a lot of people with eating disorders do not get the access to treatment they need and deserve, for whatever reason (lack of available treatment, denial of illness, lack of affordable treatment, stigma, etc..)

The take home message, if I can call it that, is mainly that the incidence, prevalence and mortality rates (as well as all other rates, such as recovery, relapse, etc..) depend in large part on the study parameters. 

Are we studying twins? Inpatient female sample? Outpatient? General population? 500 people or 500,000? At one point, over a year, or over 2 decades? What criteria are we using to assess eating disorders? DSM-III? DSM-IV? DSM-5? Official diagnosis or self-assessment? Are we including sub-threshold EDs? Are we studying adolescents? Children under 13? Young adults or individuals in their 40’s and 50’s? Are we studying minorities? What’s the access to treatment like in that country? What about the record keeping (great or sloopy)? Can we even trust the primary documents the researchers are working off (like a registry, for example)? 

Well, you get the point.

I’ve blogged about mortality rates in EDs in more depth before here (Arcelus et al., 2011) and here (Crow et al., 2009). These two papers were covered in this review. Check out those posts for more discussion on the factors that influence the mortality rates observed in various studies.

Please note that I’ve excluded binge eating disorder (BED) in this post, though it was mentioned in the article. My choice to do here, and more-or-less through-out the blog is not because I don’t think BED is a problem, or should be included in the ED umbrella. Of course it should be! I don’t blog about it because, in part, I don’t know too much about it and because there is already another blog by Dr. Sumati Gupta, that focuses specifically on binge eating and bulimia.


Smink, F.R., van Hoeken, D., & Hoek, H.W. (2012). Epidemiology of eating disorders: incidence, prevalence and mortality rates. Current Psychiatry Reports, 14 (4), 406-14 PMID: 22644309


Tetyana is the creator and manager of the blog.