Our findings show that the frequency of central AI is muchhigherthan that of primary AI in a Japanese hospital. Given thatprimaryAI and central AI occur in a 1:2 ratio based on paststudiesconducted outside Japan , it is surprising that no casesofprimary AI were diagnosed while more than 20 cases of centralAIwere diagnosed during the same period in this study.Nonetheless,this result may be supported by past studies reportingthatAddison’s disease is much less common in Japan than inothercountries [11,13,22,23].
In the present study, the most common cause of newlydiagnosedcentral AI was glucocorticoids. This result supportsindirectevidence [1,4-7,19-21,36-39] that the most common cause ofcentralAI is probably glucocorticoids because the prevalence ratesofchronic glucocorticoid use and biochemical AI amongglucocorticoidusers are high [21,36-39]. However, AI due topituitary tumoursaccounted for less than one-fourth of all cases ofnewly diagnosedcentral AI excluding glucocorticoid-induced AI inthis study. Thisresult is not consistent with the results of paststudies thatreported that more than two-thirds of central AIexcludingglucocorticoid-induced AI was caused by pituitarytumours[3,7,15,17,18]. Given that these past studies [3,15,17,18]wereconducted outside Japan and idiopathic isolated ACTH deficiencymaybe more common in Japan than in other countries [24,25],ourfinding suggests that pituitary tumours are a relativelylesscommon cause of central AI in Japan. However, our studyhaslimitations due to the small sample size and single-centrestudydesign. Furthermore, given that few studies have focused ontheaetiology of newly diagnosed central AI and most paststudiestargeted hypopituitarism rather than hypoadrenalism[15,17,18],further studies are needed to investigate the aetiologyof newlydiagnosed central AI in Japan and other countries.
In the present study, approximately 20% of all newly diagnosedAIwas diagnosed more than one year after onset. Our findingisconsistent with that of past studies showing that diagnosticdelayof AI is common [2,3]. Moreover, the time to correct diagnosisfromthe onset was significantly shorter for patientswithglucocorticoid-induced AI than for those with central AI duetoother causes in this study. This result supports anecdotalevidence[41,46] that glucocorticoid-induced AI is easier todiagnose thancentral AI due to other causes. To improve thisdiagnostic delay,further studies investigating associated factorsare warranted.
With regard to the features of central AI atpresentation,excluding glucocorticoid-induced AI, our findings areconsistentwith those of past studies [3,29], showing that the mostcommonsymptoms are fatigue, asthenia, and gastrointestinalsymptoms; itis noteworthy that one  of the two previous studieshad recallbias because the survey was administered a long timeafter thediagnosis was made. However, abdominal pain and diarrhoeaasgastrointestinal symptoms were relatively less common inthisstudy, although other gastrointestinal symptoms were common.Thiswas also consistent with the findings of past studies thatshowedthat abdominal pain and diarrhoea were less common inpatients withcentral AI than in those with primary AI [3,29]. Inthe presentstudy, a substantial proportion of patients with centralAI hadhypoglycaemic episodes and hyponatremia before diagnosis.Theseresults are also consistent with the results of paststudiesinvestigating the clinical features of isolated ACTHdeficiency[24,30,32] or central AI .
To the best of our knowledge, this study was the firsttoinvestigate the features of glucocorticoid-induced AI at thetimeof presentation because past studies on glucocorticoid-inducedAItargeted asymptomatic rather than asymptomatic AI inglucocorticoidusers [21,36-39]. There was no difference in clinicalfeaturesbetween glucocorticoid-induced AI and central AI due toothercauses. However, fatigue and asthenia were less common inpatientswith glucocorticoid-induced AI than those with central AIdue toother causes. Moreover, hypoglycaemic episodes andhyponatremiawere also less frequent in patients withglucocorticoid-induced AIthan in those with central AI due to othercauses. Given that thetime to correct diagnosis was shorter inpatients withglucocorticoid-induced AI than in those with centralAI of othercauses, these symptoms and findings may develop if AI isnottreated promptly after onset.
Strengths and weaknesses
To the best of our knowledge, this was the first studytodetermine the aetiology of newly diagnosed AI and investigatetheclinical features of central AI, includingglucocorticoid-inducedAI. To avoid including diseases other thanAI, we included onlycases of AI that were confirmed by hormonetests. Moreover, becausenone of the hormone tests, including theinsulin tolerance test,correctly classifies all patients with AI[1,7,40], only patientswho were symptomatic and treated by hormonereplacement wereincluded.
However, our findings should be interpreted cautiously.First,our study was a retrospective observational study. Therefore,thedata extracted for use in this study might not beaccurate.Moreover, AI that was not coded by the attendingphysicians wasmissed in this study. Second, we excluded AI that wasdiagnosedclinically without confirmation by hormone tests. However,none ofthe 13 excluded patients who had AI treated by hormonereplacementtherapy without confirmation by hormone tests hadprimary AI.Furthermore, even if the 5 excluded patients with AI whoweretreated by hormone replacement therapy after surgery forpituitarytumours who lacked AI symptoms and confirmation fromhormone testswere included in this study, AI due to pituitarytumours accountedfor less than half of the cases of central AI.Third, our study wasa single-centre study with a small sample size.Therefore, furthermulticentre studies should be performed toinvestigate theaetiology of newly diagnosed AI. Fourth, the use ofunprescribedcorticosteroid-like agents  is not routinelyassessed in ourhospital. Moreover, the assessment of the effect oftopical andintra-articular glucocorticoid use on AI might beinsufficient.Therefore, the frequency of drug-induced AI mightbeunderestimated. Fifth, given the limited sensitivity of the250-µgcorticotropin stimulation test for central AI, particularlyin theearly phase after onset [47,48], the frequency of central AImighthave been underestimated. Sixth, pituitary hormones otherthancorticotropin were not routinely investigated by laboratorytestswithout clinical suspicion of deficiency. Therefore,deficienciesof other pituitary hormones, which could affect thesymptoms andsigns of patients, might have beenunderestimated.
Implications for clinical practice
In the present study, primary AI was rare and pituitarytumoursaccount for less than one-fourth of the cases of central AI.In areview article published 80 years ago, Sheehan stated that thetruetype of central AI due to tumours is rare . Ourfindingssuggest that pituitary tumours may not be common causes ofcentralAI in Japan, unlike in other countries. However, given thatfewstudies have focused on the aetiology of newly diagnosed centralAI and that most past studies conducted in countries outsideofJapan [15-18] targeted hypopituitarism and nothypoadrenalism,further studies are warranted to investigate theaetiology of newlydiagnosed central AI in Japan and othercountries.