Bulimia nervosa is a recurrent binge–purge eating disorder that hides in plain sight: patients are typically at or above expected body weight, present with vague gastrointestinal or dental complaints, and rarely volunteer the diagnosis. DSM-5-TR locates it in Feeding and Eating Disorders alongside anorexia nervosa and binge-eating disorder, distinguished by recurrent compensatory behaviors and an undue influence of weight and shape on self-evaluation. The clinical priorities are three: confirm the diagnosis with a non-judgmental interview, screen aggressively for the medical sequelae of purging (hypokalemia, esophageal injury, dental erosion), and start cognitive behavioral therapy for eating disorders (CBT-E) — the only intervention with strong, replicated efficacy across age groups. Fluoxetine 60 mg daily is the sole FDA-approved pharmacotherapy and is an effective adjunct or alternative when therapy is unavailable. Mortality is lower than in anorexia nervosa but is driven by suicide and electrolyte-related cardiac events, both of which are preventable with attentive care.

Bulimia nervosa is more common than anorexia nervosa but consistently underdetected in primary care. Reliable lifetime prevalence estimates come from population surveys using DSM criteria; rates are rising in low- and middle-income countries as Western body-shape ideals globalize.^1-2

Prevalence and demographics

• Lifetime prevalence among women in high-income countries is approximately 1.0–1.5%, and 12-month prevalence 0.3–0.5%.^1-2
• Lifetime prevalence among men is roughly one-third that of women, with rising recognition in sexual and gender minority populations.^1,3
• Modal age of onset is 18–20 years, later than anorexia nervosa, with frequent onset after a period of dietary restriction or weight loss.^1-2
• Point prevalence among adolescents (ages 13–18) is approximately 0.9–1.3%, with female-to-male ratio of 3:1 to 5:1 in community samples.^2,4

Comorbidity

• Lifetime comorbid major depressive disorder occurs in 50–70% of cases; anxiety disorders in 50–80%.^1-2
• Substance use disorders co-occur in approximately 30–40%, with alcohol and stimulants overrepresented.^1-2
• Borderline personality disorder co-occurs in roughly 20–30%, and is associated with more severe impulsivity and self-harm.^2,5
• Lifetime suicide attempt prevalence is 25–35%; standardized mortality ratio is elevated approximately 2-fold relative to the general population.⁶

Risk factors

• Female sex, adolescence, family history of eating disorder, mood disorder, or substance use disorder.²
• Childhood obesity, early menarche, and parental comments about weight.^2,5
• Perfectionism, negative affectivity, and impulsivity as temperamental risk factors.^2,5
• Occupational exposure to weight-focused environments (ballet, modeling, wrestling, gymnastics).²

Bulimia nervosa is a disorder of disrupted appetite regulation layered on top of cognitive overvaluation of weight and shape. No single neurotransmitter system or gene accounts for it; current models integrate serotonergic dysregulation, reward-circuit alterations, and learned restriction-binge-purge cycles.^2,7

Neurobiology

• Serotonergic abnormalities are the most replicated finding: reduced central serotonin activity during the symptomatic state and persistent alterations in 5-HT1A and 5-HT2A receptor binding after recovery, consistent with the therapeutic effect of fluoxetine.^7-8
• Dopaminergic reward circuitry in the ventral striatum shows blunted response to food cues in the satiated state and exaggerated response during caloric restriction, mirroring binge vulnerability.⁷
• Functional MRI studies report altered insular and anterior cingulate activity to taste and interoceptive stimuli, supporting impaired interoception.⁷
• Structural imaging is largely unremarkable in uncomplicated bulimia nervosa, distinguishing it from the volume loss seen in low-weight anorexia nervosa.²

Genetics

• Twin-study heritability estimates range from 30% to 60% for bulimia nervosa and related binge-eating phenotypes.^2,7
• Genome-wide association studies have identified loci shared across eating disorders, with substantial genetic correlation between anorexia nervosa, bulimia nervosa, and psychiatric disorders including major depressive disorder and obsessive–compulsive disorder.⁷

Environmental contributors

• Dieting is the single most consistent proximal risk factor; restriction precedes binge onset in the majority of cases.^2,5
• Childhood sexual or physical abuse is a non-specific risk factor that approximately doubles odds of bulimia nervosa.^2,5
• Social media exposure to weight-focused content is associated with increased disordered-eating behaviors in adolescents, though causal evidence remains limited.²

DSM-5-TR defines bulimia nervosa by recurrent binges paired with recurrent inappropriate compensatory behaviors, both occurring on average at least weekly for three months. The diagnosis cannot be made if the binge–purge behavior occurs exclusively during episodes of anorexia nervosa.⁹

Core DSM-5-TR criteria

• Recurrent episodes of binge eating, defined as eating an objectively large amount of food within a discrete period (typically under 2 hours) accompanied by a sense of loss of control.⁹
• Recurrent inappropriate compensatory behaviors to prevent weight gain: self-induced vomiting, laxative or diuretic misuse, fasting, or excessive exercise.⁹
• Both binges and compensatory behaviors occur on average at least once per week for three months.⁹
• Self-evaluation is unduly influenced by body shape and weight.⁹
• The disturbance does not occur exclusively during episodes of anorexia nervosa.⁹

Severity specifier (based on average weekly compensatory episodes):

• Mild: 1–3 episodes.⁹
• Moderate: 4–7 episodes.⁹
• Severe: 8–13 episodes.⁹
• Extreme: 14 or more episodes.⁹

Course specifiers

• In partial remission: full criteria previously met; some but not all criteria currently met for a sustained period.⁹
• In full remission: full criteria previously met; none currently met for a sustained period.⁹

ICD-11 differences

• ICD-11 (6B81) requires binge eating and compensatory behaviors at least once weekly for one month rather than three months, broadening diagnosis at earlier illness stages.¹⁰
• ICD-11 explicitly recognizes subjective binges (loss of control without objectively large quantity), which DSM-5-TR captures under Other Specified Feeding or Eating Disorder.¹⁰

Patients are typically normal weight or overweight, frequently distressed by perceived loss of control rather than weight per se, and often present after a partner or roommate discovers the behavior. The picture is one of secrecy, shame, and physical sequelae that may precede psychiatric disclosure.^2,5

Behavioral pattern

• Binges typically occur in the evening or at night, in secret, lasting under two hours, with caloric intake commonly 1,000–3,000 kcal per episode and occasionally far higher.^2,5
• Foods consumed during binges are characteristically high-carbohydrate or high-fat “forbidden” items the patient restricts at other times.^2,5
• Self-induced vomiting is the most common compensatory behavior (about 80–90% of patients); laxative misuse occurs in approximately 30%, diuretics in 10–15%, and driven exercise is increasingly recognized.^2,5
• The cycle is maintained by dietary restriction between binges, which itself increases binge vulnerability via hunger and cognitive disinhibition.^2,5

Cognitive and affective features

• Overvaluation of weight and shape — body image dominates self-worth disproportionately to other domains.^2,9
• Cognitive distortions: dichotomous thinking about food (“good” versus “bad”), catastrophizing about weight gain, and arbitrary food rules.^2,5
• High negative affect immediately preceding binges, with transient relief during binges followed by guilt and self-disgust after purging.^2,5

Physical signs

• Russell's sign: callus or scarring on the dorsum of the hand from self-induced vomiting.^2,5
• Parotid and submandibular gland hypertrophy, giving a “chipmunk” facial appearance.^2,5
• Dental erosion, particularly of the lingual surfaces of upper incisors (perimylolysis), and increased caries.^2,5
• Lanugo is uncommon in bulimia nervosa and, when present, suggests crossover with anorexia nervosa.²

Course red flags

• Ipecac use is rare but causes irreversible cardiomyopathy and is a medical emergency.^2,5
• Esophageal complaints (heartburn, hematemesis) raise concern for Mallory–Weiss tears or, rarely, Boerhaave syndrome.^2,5
• Rapid weight loss in a known bulimia nervosa patient should prompt re-evaluation for crossover to anorexia nervosa.²

The differential narrows quickly once weight status, the presence of true binges, and compensatory behavior are clarified. Medical mimics are uncommon but must not be missed in atypical presentations.^2,5

Other eating disorders

• Anorexia nervosa, binge-eating/purging subtype: same behaviors but at low weight (BMI below the DSM-5-TR threshold or, in adolescents, below expected for growth curve).⁹
• Binge-eating disorder: recurrent binges without compensatory behaviors; patients are more often overweight or obese.⁹
• Other Specified Feeding or Eating Disorder (OSFED): subthreshold frequency or duration, purging disorder without binges, or atypical anorexia at normal weight.⁹
• Avoidant/restrictive food intake disorder: restriction without body-image disturbance — does not feature binges.⁹

Psychiatric mimics

• Major depressive disorder with appetite increase: weight gain and overeating without loss of control or compensatory behavior.²
• Borderline personality disorder: impulsive overeating may resemble binges but lacks the cognitive overvaluation of shape and weight; the two often co-occur.^2,5
• Body dysmorphic disorder: preoccupation focuses on non-weight features (skin, facial structure); compensatory behaviors are absent.⁹

Medical mimics

• Kleine–Levin syndrome: episodic hypersomnia with hyperphagia, no compensatory behavior, characteristic neurologic course.²
• Hypothalamic lesions (craniopharyngioma, tumor): hyperphagia with neurologic or endocrine findings.²
• Prader–Willi syndrome: childhood-onset hyperphagia with characteristic dysmorphic and developmental features.²
• Gastroparesis and gastroesophageal reflux: postprandial vomiting that is involuntary and lacks intentionality.²

Substance-induced presentations

• Cannabis use: increased appetite with episodic overeating but without self-evaluation by shape and weight.²
• Stimulant misuse (including misused weight-loss drugs): appetite suppression and weight loss that may coexist with bulimia nervosa rather than replacing it.²

Assessment combines a focused eating-disorder history, a medical evaluation tailored to purging method, and validated rating scales. The interview should be matter-of-fact: patients minimize symptoms, and shame is a major barrier to disclosure.^2,5

History — mandatory elements

• Frequency, duration, and triggers of binges and each compensatory behavior, asked about by name (vomiting, laxatives, diuretics, fasting, exercise, ipecac, diet pills, thyroid hormone misuse).^2,5
• Weight history, including highest and lowest adult weights and recent trajectory.^2,5
• Dietary pattern between binges — restriction is the modifiable driver of relapse.^2,5
• Menstrual history; amenorrhea suggests crossover with anorexia nervosa or significant medical compromise.²
• Comorbid mood, anxiety, substance use, and personality features; screen for suicide and self-injury at every visit.^2,5-6
• Trauma history, asked with appropriate timing and consent.²

Physical examination

• Vital signs: orthostatic blood pressure and pulse; bradycardia or orthostasis suggest dehydration or low weight.^2,5
• Oropharynx and teeth: erosion, caries, parotid enlargement.^2,5
• Hands: Russell's sign on the dorsum.^2,5
• Skin: lanugo, calluses, signs of self-injury.^2,5
• Cardiac exam, given the risk of arrhythmia from electrolyte disturbance.^2,5

Laboratory and diagnostic workup

• Basic metabolic panel: hypokalemia, hypochloremic metabolic alkalosis (vomiting) or non–anion-gap acidosis (laxatives).^2,5
• Magnesium and phosphate; hypomagnesemia is common and exacerbates refractory hypokalemia.^2,5
• CBC, liver function tests, lipase if abdominal pain.^2,5
• TSH to rule out thyroid contribution and to detect thyroid hormone misuse.^2,5
• Urinalysis (specific gravity, pH) and urine pregnancy test in patients of reproductive potential.^2,5
• EKG: look for QT prolongation, U waves, and arrhythmia in any patient with electrolyte disturbance or syncope.^2,5
• Bone density (DXA) if there is a history of anorexia nervosa, prolonged amenorrhea, or significant weight loss.²

What NOT to order routinely

• Brain imaging is not indicated in typical presentations; reserve for focal neurologic signs, atypical age of onset, or suspected hypothalamic lesion.¹³
• Endoscopy is not part of routine workup; obtain for hematemesis, dysphagia, or suspected Boerhaave syndrome.²

Validated rating scales

• Eating Disorder Examination (EDE) and EDE-Q: gold-standard semi-structured interview and self-report; quantify restraint, eating concern, shape concern, and weight concern.¹¹
• SCOFF questionnaire: 5-item screening tool; 2 or more positive responses has sensitivity ~85% and specificity ~75% for an eating disorder in primary care.¹²
• Eating Attitudes Test (EAT-26): widely used screening instrument with strong psychometric performance.²⁴
• PHQ-9 and GAD-7 to capture comorbid depression and anxiety; suicide risk screen at every visit.¹³

Bulimia nervosa is treated outpatient in the majority of cases, with CBT-E and fluoxetine as the two evidence-based pillars. Severity of medical compromise, comorbidity, and prior treatment response drive choices between modalities and levels of care.^2,13-14

Pharmacotherapy

• Strong evidence supports fluoxetine 60 mg daily as first-line pharmacotherapy and the only FDA-approved agent for bulimia nervosa; the 60 mg dose is superior to 20 mg for binge–purge reduction.^13,15-16
• Evidence suggests other SSRIs (sertraline, citalopram) are reasonable second-line options when fluoxetine is not tolerated, though direct comparative trials are limited.^13-14
• Topiramate, dosed 100–250 mg daily, reduces binge–purge frequency with moderate-quality evidence but is limited by cognitive adverse effects and teratogenicity.^14,17
• DANGER: Bupropion is contraindicated in active bulimia nervosa because of an elevated seizure risk demonstrated in early clinical trials with bulimic patients.^13,18
• Tricyclic antidepressants (imipramine, desipramine) have demonstrated efficacy but are rarely used because of overdose lethality in a population with elevated suicide risk.¹³
• Monoamine oxidase inhibitors should be avoided given dietary tyramine restrictions that are unsafe in patients with unpredictable eating.¹³

Psychotherapy

• Strong evidence supports CBT-E (enhanced cognitive behavioral therapy) as the first-line psychotherapy; 20 sessions over 20 weeks for non-underweight patients, with approximately 40–60% achieving full remission at end of treatment.^14,19-20
• Evidence suggests interpersonal psychotherapy (IPT) achieves outcomes comparable to CBT-E at long-term follow-up but works more slowly; it is a reasonable alternative when CBT-E is unavailable or unacceptable.^14,19
• Family-based treatment (FBT) is recommended for adolescents with bulimia nervosa, with evidence supporting modest superiority over individual therapy in this age group.^14,21
• Dialectical behavior therapy is an option for patients with comorbid emotion dysregulation, borderline features, or self-injury, though evidence is more limited.¹⁴
• Guided self-help based on CBT principles is an evidence-based step in stepped-care models, particularly when access to specialist therapy is limited.^14,19

Neuromodulation

• Limited evidence suggests repetitive transcranial magnetic stimulation (rTMS) to dorsolateral prefrontal cortex reduces food craving and binge frequency in small trials; not currently recommended outside research settings.²⁵
• ECT has no established role in bulimia nervosa itself and is reserved for severe comorbid mood disorder unresponsive to pharmacotherapy.²

Adjunctive

• Nutritional rehabilitation focused on regular structured eating (three meals plus two to three snacks daily) is foundational and reduces binge frequency by counteracting restriction.^14,19
• Dental referral for evaluation and protective measures (fluoride, custom mouthguards) in any patient with self-induced vomiting.^2,5
• Treatment of comorbid mood, anxiety, or substance use disorders proceeds in parallel.¹⁴
• Higher levels of care (partial hospitalization, residential, inpatient) are warranted for medical instability, severe electrolyte derangement, suicidality, failure of outpatient treatment, or severe comorbid psychiatric illness.¹⁴

Bulimia nervosa is associated with substantial medical morbidity from purging behaviors and significant treatment-related harms in pharmacotherapy. The evidence base has well-recognized gaps, particularly in men, sexual and gender minority populations, and long-term outcomes.^2,5,14

Medical harms of the disorder

• Electrolyte disturbance: hypokalemia, hypochloremic metabolic alkalosis from vomiting, non–anion-gap metabolic acidosis from laxatives; risk of cardiac arrhythmia and sudden death.^2,5
• Dental erosion, caries, and parotid hypertrophy from chronic exposure to gastric acid.^2,5
• Esophagitis, Mallory–Weiss tears, and rarely Boerhaave syndrome.^2,5
• Menstrual irregularity, infertility risk, and adverse pregnancy outcomes (preterm birth, miscarriage).^2,5
• Cathartic colon and rebound edema with laxative withdrawal.^2,5
• Ipecac-induced cardiomyopathy, irreversible and potentially fatal.^2,5

Treatment-related harms

• Fluoxetine: GI upset, insomnia, sexual dysfunction, modest weight loss; FDA black-box warning for suicidality in patients under age 25.^16,18
• Topiramate: cognitive slowing, word-finding difficulty, paresthesias, kidney stones, metabolic acidosis, teratogenicity (oral cleft).^17-18
• SSRI–MAOI interactions and serotonin syndrome risk if combined.¹⁸
• Psychotherapy carries risks of partial response, dropout (20–30%), and symptom substitution in a minority.^14,19

Limitations of the evidence base

• Most RCTs enroll predominantly white, female, adult patients; generalizability to men, racial/ethnic minorities, and gender-diverse patients is limited.^2,14
• Long-term outcome data beyond 5 years are sparse for both psychotherapy and pharmacotherapy.¹⁴
• Comparative effectiveness of CBT-E versus pharmacotherapy alone versus combination remains incompletely characterized.^14,19
• Publication bias likely overstates pharmacotherapy effects in older trials.¹⁴

Clinical features and treatment principles hold across populations, but specific modifications are required. The most consequential are in adolescents, pregnant patients, and patients with comorbid type 1 diabetes.^2,14

Adolescents

• Family-based treatment is the preferred psychotherapy with evidence supporting superiority over individual CBT in this age group.^14,21
• Fluoxetine is used off-label in adolescents with bulimia nervosa; the SSRI suicidality warning applies and warrants close monitoring.^14,18
• Growth and pubertal development must be tracked; crossover to anorexia nervosa is common in adolescent presentations.²

Pregnancy and perinatal

• Symptoms often improve during pregnancy but relapse postpartum, with elevated risk of postpartum depression.^2,14
• Bulimia nervosa in pregnancy is associated with hyperemesis gravidarum, miscarriage, preterm birth, and postpartum mood disorders.²
• Fluoxetine has the largest reproductive safety dataset among SSRIs; topiramate is contraindicated due to oral-cleft teratogenicity.^17-18

Geriatric

• Late-onset bulimia nervosa is uncommon; new-onset purging in an older adult warrants medical workup (malignancy, GI disease) before psychiatric attribution.²
• Medication choice should account for renal function, anticholinergic burden, and bone health.^2,18

Comorbid medical illness

• DANGER: Type 1 diabetes mellitus with bulimia nervosa carries the risk of insulin omission for weight control (sometimes called diabulimia); this dramatically accelerates retinopathy, nephropathy, and mortality and warrants urgent multidisciplinary intervention.^2,22
• Inflammatory bowel disease and celiac disease may overlap with restrictive and purging behaviors; coordinate with gastroenterology.²

Comorbid substance use

• Alcohol and stimulant use disorders are common and amplify medical risk from electrolyte disturbance.^2,5
• Diuretic and laxative misuse may be hidden within stimulant or weight-loss product use; ask specifically.^2,5

Cultural considerations

• Bulimia nervosa occurs across cultures; presentation may differ, with less explicit weight/shape language in some non-Western settings.^1-2
• Sexual and gender minority youth, particularly transgender adolescents, have elevated rates of bulimia nervosa and benefit from gender-affirming care alongside eating-disorder treatment.³

Bulimia nervosa has a more favorable prognosis than anorexia nervosa, with substantial recovery rates over the medium term and lower mortality, but chronicity and relapse are common. Suicide is the most important driver of excess mortality.^2,6,23

Natural history

• Approximately 50–70% of patients achieve full remission over 5–10 years; another 20–30% have a chronic, fluctuating course.^2,23
• Diagnostic crossover is common: roughly 20–30% of patients with bulimia nervosa cross over to or from anorexia nervosa or other eating disorders over time.^2,23
• Median duration of illness in untreated cases exceeds 5 years.^2,23

Predictors of outcome

• Better outcomes: shorter duration of illness, earlier age of diagnosis, fewer comorbidities, good early therapy response.^2,14,23
• Worse outcomes: longer illness duration, comorbid borderline personality disorder, substance use, severe purging frequency, history of anorexia nervosa.^2,14,23

Mortality

• Standardized mortality ratio is approximately 1.5–2.0 relative to age-matched general population, lower than anorexia nervosa but still elevated.⁶
• Suicide accounts for a substantial proportion of deaths; medical deaths are largely electrolyte-related cardiac events.⁶

Most bulimia nervosa care is outpatient, but several scenarios require urgent escalation. The two highest-acuity questions at presentation are medical stability and suicide risk.¹⁴

Indications for medical hospitalization

• Serum potassium below 3.0 mEq/L, or any potassium below 3.5 mEq/L with EKG changes or symptoms.^2,5,14
• Significant dehydration, orthostasis unresponsive to oral rehydration, or hypovolemia.^2,5
• Hematemesis or suspected esophageal rupture.^2,5
• QTc prolongation above 500 ms or any clinically significant arrhythmia.^2,5
• Suspected ipecac toxicity.^2,5

Indications for psychiatric hospitalization

• Active suicidal ideation with plan or intent, or recent suicide attempt.¹⁴
• Severe comorbid psychiatric illness preventing outpatient engagement.¹⁴
• Failure of intensive outpatient or partial hospitalization with continued medical compromise.¹⁴

Suicide risk markers

• Lifetime attempt prevalence is 25–35%; risk is elevated by comorbid mood, borderline personality, and substance use disorders.⁶
• Screen for suicide at every visit; access-to-means counseling around medications (especially tricyclics, if prescribed for other indications) is standard.⁶

Acute agitation

• Agitation is uncommon in uncomplicated bulimia nervosa; when present, evaluate for electrolyte disturbance, substance intoxication or withdrawal, and comorbid psychiatric illness before pharmacologic management.²

Several questions remain unsettled in bulimia nervosa research and practice. Most reflect gaps in long-term and comparative-effectiveness data rather than fundamental disagreement about first-line care.^14,19

• Combination therapy: whether CBT-E plus fluoxetine is reliably superior to either alone remains debated; some trials show added benefit, others do not.^14,19
• DSM-5-TR versus ICD-11 threshold: the once-weekly-for-one-month ICD-11 criterion captures patients DSM-5-TR places under OSFED, with unclear clinical implications.¹⁰
• Diagnostic stability: the high crossover rate between eating disorders has prompted calls for a transdiagnostic framework, reflected in CBT-E's design but not in current DSM categories.^19,23
• Pharmacogenomics: routine pharmacogenomic testing to guide SSRI choice is not currently recommended for bulimia nervosa.¹³
• Glucagon-like peptide-1 (GLP-1) receptor agonists: case reports and small studies have raised both interest in their potential to reduce binge eating and concern about facilitating restrictive behaviors; safety in bulimia nervosa is not established.¹³
• Digital and remote-delivered CBT-E: evidence is promising for non-inferiority to in-person therapy but heterogeneous; access and engagement vary substantially.¹⁴
• Psychedelic-assisted therapy: preliminary investigation only; no current role outside research.¹³

• Bulimia nervosa requires binges plus compensatory behaviors averaging at least once weekly for three months per DSM-5-TR.⁹
• Patients with bulimia nervosa are typically normal weight or overweight; if low weight, the diagnosis is anorexia nervosa, binge-eating/purging subtype.⁹
• Fluoxetine 60 mg daily is the only FDA-approved pharmacotherapy for bulimia nervosa; the 60 mg dose outperforms 20 mg.^15-16
• Bupropion is contraindicated in active bulimia nervosa because of elevated seizure risk.^13,18
• CBT-E is first-line psychotherapy; family-based treatment is preferred for adolescents.^14,19,21
• Self-induced vomiting causes hypokalemic, hypochloremic metabolic alkalosis; laxative misuse causes non–anion-gap metabolic acidosis.^2,5
• Russell's sign is a callus on the dorsum of the hand from self-induced vomiting.^2,5
• Parotid hypertrophy and dental erosion of the lingual surfaces of upper incisors are characteristic physical signs.^2,5
• Ipecac use causes irreversible cardiomyopathy.^2,5
• SCOFF questionnaire score of 2 or more is a positive screen for an eating disorder.¹²
• Lifetime suicide attempt prevalence in bulimia nervosa is 25–35%.⁶
• Standardized mortality ratio is approximately 1.5–2.0, lower than anorexia nervosa.⁶
• ICD-11 requires once-weekly behavior for one month rather than DSM-5-TR's three months.¹⁰
• Insulin omission for weight control in type 1 diabetes (diabulimia) dramatically accelerates microvascular complications and mortality.²²
• Approximately 20–30% of patients cross over between eating disorder diagnoses over time.^2,23

No external funding. No conflicts of interest declared. Peer-review status: pending.