Articles
`
`Mortality in patients with Cushing’s disease more than
`10 years after remission: a multicentre, multinational,
`retrospective cohort study
`
`Richard N Clayton, Peter W Jones, Raoul C Reulen, Paul M Stewart, Zaki K Hassan-Smith, Georgia Ntali, Niki Karavitaki, Olaf M Dekkers,
`Alberto M Pereira, Mark Bolland, Ian Holdaway, Jorgen Lindholm
`
`Summary
`Background No agreement has been reached on the long-term survival prospects for patients with Cushing’s disease.
`We studied life expectancy in patients who had received curative treatment and whose hypercortisolism remained in
`remission for more than 10 years, and identifi ed factors determining their survival.
`
`Methods We did a multicentre, multinational, retrospective cohort study using individual case records from specialist
`referral centres in the UK, Denmark, the Netherlands, and New Zealand. Inclusion criteria for participants, who had
`all been in studies reported previously in peer-reviewed publications, were diagnosis and treatment of Cushing’s
`disease, being cured of hypercortisolism for a minimum of 10 years at study entry, and continuing to be cured with
`no relapses until the database was frozen or death. We identifi ed the number and type of treatments used to achieve
`cure, and used mortality as our primary endpoint. We compared mortality rates between patients with Cushing’s
`disease and the general population, and expressed them as standardised mortality ratios (SMRs). We analysed survival
`data with multivariate analysis (Cox regression) with no corrections for multiple testing.
`
`Findings The census dates on which the data were frozen ranged from Dec 31, 2009, to Dec 1, 2014. We obtained data
`for 320 patients with 3790 person-years of follow-up from 10 years after cure (female:male ratio of 3:1). The median
`patient follow-up was 11·8 years (IQR 17–26) from study entry and did not diff er between countries. There were no
`signifi cant diff erences in demographic characteristics, duration of follow-up, comorbidities, treatment number, or
`type of treatment between women and men, so we pooled data from both sexes for survival analysis. 51 (16%) of the
`cohort died during follow-up from study entry (10 years after cure). Median survival from study entry was similar for
`women (31 years; IQR 19–38) and men (28 years; 24–42), and about 40 years (IQR 30–48) from remission. The overall
`SMR for all-cause mortality was 1·61 (95% CI 1·23–2·12; p=0·0001). The SMR for circulatory disease was increased
`at 2·72 (1·88–3·95; p<0·0001), but deaths from cancer were not higher than expected (0·79, 0·41–1·51). Presence of
`diabetes, but not hypertension, was an independent risk factor for mortality (hazard ratio 2·82, 95% CI 1·29–6·17;
`p=0·0095). We noted a step-wise reduction in survival with increasing number of treatments. Patients cured by
`pituitary surgery alone had long-term survival similar to that of the general population (SMR 0·95, 95% CI 0·58–1·55)
`compared with those who were not (2·53, 1·82–3·53; p<0·0001).
`
`Interpretation Patients with Cushing’s disease who have been in remission for more than 10 years are at increased
`risk of overall mortality compared with the general population, particularly from circulatory disease. However,
`median survival from cure is excellent at about 40 years of remission. Treatment complexity and an increased
`number of treatments, refl ecting disease that is more diffi cult to control, appears to negatively aff ect survival.
`Pituitary surgery alone is the preferred treatment to secure an optimum outcome, and should be done in a centre
`of surgical excellence.
`
`Funding None.
`
`Introduction
`The predominant cause of endogenous Cushing’s
`syndrome is Cushing’s disease, which is characterised by
`increased secretion of adrenocorticotropic
`(ACTH)
`hormone from the anterior pituitary gland.1–4 If untreated,
`Cushing’s disease leads to markedly increased mortality
`from circulatory diseases. Findings from population-
`based studies and patients with adrenal incidentalomas
`suggest that higher concentrations of cortisol within
`normal reference ranges might be linked to increased
`cardiovascular mortality (appendix).5–9
`
`Several cohort studies10–18 have examined whether
`mortality in Cushing’s disease is aff ected by restoration
`of normal cortisol concentrations, or not, with variable
`results. Findings from meta-analyses have shown that
`patients with Cushing’s disease who have
`initial
`remission of hypercortisolism have
`a
`reduced
`standardised mortality ratio (SMR) compared with that
`of patients with persistent disease, although overall
`SMR (2·5) was still increased. Considerable heterogeneity
`in SMR exists between studies, and with respect to
`defi nition of remission, timing of remission (which
`
`www.thelancet.com/diabetes-endocrinology Vol 4 July 2016
`
`Lancet Diabetes-Endocrinol 2016;
`4: 569–76
`Published Online
`June 2, 2016
`http://dx.doi.org/10.1016/
`S2213-8587(16)30005-5
`See Comment page 551
`Department of Medicine
`(Prof R N Clayton MD) and
`Health Services Research Unit
`(Prof P W Jones PhD), Keele
`University, Stoke on Trent, UK;
`Centre for Childhood Survivor
`Studies, Institute of Applied
`Health Research,
`(Prof R C Reulen MD), and
`Centre for Endocrinology,
`Diabetes and Metabolism,
`Birmingham Health Partners,
`University of Birmingham,
`Birmingham, UK
`(Z K Hassan-Smith PhD,
`N Karavitaki PhD); Oxford
`Centre for Diabetes,
`Endocrinology, and
`Metabolism, University of
`Oxford, Oxford, UK (G Ntali MD,
`N Karavitaki); Department of
`Medicine, University of Leeds,
`Leeds, UK
`(Prof P M Stewart MD);
`Department of Medicine,
`Division of Endocrinology,
`Leiden University Medical
`Centre, Leiden, Nederlands
`(Prof O M Dekkers PhD,
`Prof A M Pereira MD); Institute
`of Clinical Epidemiology,
`Aarhus, Denmark
`(Prof O M Dekkers); Department
`of Endocrinology, University of
`Auckland, Auckland, New
`Zealand (M Bolland PhD,
`Prof I Holdaway MD);
`Department of Endocrinology,
`Aalborg, Denmark
`(Prof J Lindholm DMSci)
`Correspondence to:
`Prof Richard N Clayton,
`Department of Medicine,
`Keele University, Stoke-on-Trent,
`ST5 5BG, UK
`r.n.clayton@keele.ac.uk
`
`See Online for appendix
`
`569
`
`Teva Pharmaceuticals USA, Inc. v. Corcept Therapeutics, Inc.
`PGR2019-00048
`Corcept Ex. 2067, Page 1
`
`

`

`Articles
`
`Research in context
`
`Evidence before this study
`We searched the MEDLINE database for English-language
`articles with the search terms “meta-analysis”, “mortality”, and
`“Cushing’s disease” and identifi ed two articles. Findings from a
`published meta-analysis of mortality outcomes in individuals
`with treated Cushing’s disease in initial remission after
`treatment (the defi nition of which varies between studies)
`shows an increased standardised mortality ratio (SMR) of 2·5
`but with marked heterogeneity (SMR range 0·3–10·0).
`
`Added value of this study
`We did a retrospective cohort study of 320 individuals who
`have been cured of Cushing’s disease for more than 10 years
`using various treatments, and who remain in remission.
`
`These patients have an overall SMR of 1·6 and an SMR from
`circulatory disease of 2·7. We noted a stepwise reduction in
`survival with increasing number of treatments, but patients
`cured by pituitary surgery alone had no increase in SMR
`(0·95). There was no heterogeneity between countries, and
`median survival from time of cure was 40 years (IQR 30–48)
`from remission.
`
`Implications of all the available evidence
`For Cushing’s disease, treatment complexity and increasing
`the number of treatments negatively aff ects patient survival.
`Pituitary surgery alone is the preferred treatment to secure an
`optimum outcome, and should be done in a centre of surgical
`excellence.
`
`Panel: Criteria for defi nition of cure or remission in Cushing’s disease by study centre
`
`Birmingham UK (15)
`•
`Initial: morning plasma cortisol less than 50 nmol/L between 4 days to 6 weeks after
`surgery
`• Follow-up: normal UFC or suppression of plasma cortisol to less than 50nmol/L after
`o/n dexamethasone
`
`Stoke on Trent UK (13)
`•
`Initial: normal UFC plus plasma cortisol less than 100 nmol/L after o/n dexamethasone
`• Follow-up: normal UFC × 2 or plasma cortisol less than 50 nmol/L after o/n
`dexamethasone
`
`Oxford UK (17)
`•
`Initial: undectable morning plasma cortisol within days after surgery
`• Follow-up: one or more of the following—normal plasma cortisol after o/n
`dexamethasone, normal UFC, or normal mean plasma cortisol from fi ve samples taken
`between 0800–1800 h
`
`Auckland (14)
`•
`Initial: information not provided
`• Follow-up: normal UFC or normal plasma cortisol after o/n dexamethasone
`
`Leiden (11)
`•
`Initial: information not provided
`• Follow-up: plasma cortisol less than 100 nmol/L after o/n dexamethasone and normal
`UFC (<220 nmoL per day)
`
`Aalborg (10)
`•
`Initial: subnormal plasma cortisol after synthetic ACTH 250 μg with or without UFC
`less than 50 nmoL per day
`• Follow-up: UFC less than 250 nmoL/day
`
`The presence of adrenal insuffi ciency requiring glucocorticoid treatment=cure. Follow-up occurred at various regular intervals
`according to local protocols. UFC=24 h urine free cortisol. o/n dexamethasone=plasma cortisol measured between
`0800–1000 h after dexamethasone 1 mg or 2 mg given at 2300 h the previous evening. ACTH=adrenocorticotropic hormone.
`Normal refers to the local reference range for the centre, which varied by centre according to the assays used, and which
`themselves changed during the course of the follow-up.
`
`initial
`immediately after
`was based on remission
`treatment, with no information provided on relapse or
`recurrence of hypercortisolism during follow-up that
`could clearly drive mortality. In an attempt to eliminate
`these issues, minimise participant heterogeneity, and
`measure the long-term survival for Cushing’s disease in
`remission, we examined mortality in patients who had
`survived 10 years or more from the time of curative
`treatment and were in continued remission up until the
`census date or their death.
`
`Methods
`Study design and participants
`We did a multicentre, multinational, restrospective cohort
`study using individual case records from specialist
`referral centres in the UK (Birmingham, Oxford, and
`Stoke on Trent), Denmark (Aalborg), the Netherlands
`(Leiden), and New Zealand (Auckland). The patients
`included were all part of previous studies reported in
`peer-reviewed publications,10,11,13,14,15,17 and were managed at
`tertiary referral centres using extant best-practice
`management techniques to national and international
`standards of the time, so are likely to refl ect national data.
`Unlike for acromegaly, there are no national registries or
`datasets for Cushing’s disease. Inclusion criteria were
`diagnosis and treatment of Cushing’s disease, cured of
`hyper cortisolism for a minimum of 10 years at study
`entry, and continued cure with no relapses of hyperc-
`ortisolism until database was frozen or death. Defi nitions
`of cure from each centre are in the panel.
`The studies from Birmingham, Oxford, Stoke on Trent,
`New Zealand, and Denmark were approved and
`registered as audits by the respective institutions so did
`not require ethical approval. The study from the
`Netherlands had ethical approval.
`
`diff ered between studies), and persistence of Cushing’s
`disease (appendix).13,19 A major limitation of these
`individual studies was that the division of the cohorts13,19
`
`Procedures
`Our primary endpoint was mortality. Follow-up for the
`present analyses started after 10 years of cure. We
`
`570
`
`www.thelancet.com/diabetes-endocrinology Vol 4 July 2016
`
`Teva Pharmaceuticals USA, Inc. v. Corcept Therapeutics, Inc.
`PGR2019-00048
`Corcept Ex. 2067, Page 2
`
`

`

`Articles
`
`identifi ed the number and type of treatments used to
`achieve cure, with cure being defi ned after the last
`treatment. In the case of radiotherapy, if this was given
`prophylactically to reduce the risk of Nelson’s syndrome
`after bilateral adrenalectomy, we used the date of the
`adrenalectomy as the cure date. The diagnostic criteria
`for Cushing’s disease and defi nition of remission could
`not be standardised and relied on those used by the
`individual centres which have all been published in their
`respective publications10,11,13,14,15,17 (appendix). Similarly,
`treatment methods varied between centres and with
`time. Initial diagnosis and treatment was from as early as
`1958 (Stoke-on-Trent) through to 1997 (Leiden). In
`addition to data for demographic characteristics and
`primary endpoint status of alive or dead and cause of
`death (from National Registry Offi ces or patient records),
`we obtained information on initial treatment, subsequent
`treatments and dates
`thereof, radiotherapy usage,
`glucocorticoid and thyroxine replacement, and treatment
`for hypertension or diabetes mellitus.
`
`Statistical analysis
`We summarised continuous data as medians (IQR) and
`discrete data as proportions. We analysed survival data
`with multivariate analysis (Cox regression; factors
`associated with survival were examined individually,
`including age at baseline as a covariate) and displayed
`them in Kaplan-Meier plots. Signifi cance was taken as p
`less than 0·05 (Wilcoxon rank sum test), and CIs, with
`no correction for multiple testing, are presented. We only
`included age at study entry in the models; other age
`variables were given for information. There was no
`signifi cant departure
`from proportional hazards
`assumptions for any of the variables.
`We obtained reference mortality rates from the Human
`Mortality Database for
`the relevant countries. We
`calculated the expected numbers of deaths (needed for the
`calculation of SMRs) with standard cohort techniques
`described in detail by Breslow and Day.20 Briefl y, we
`derived expected numbers by aggregating the person-
`years in each sex, age (1-year bands), and calendar year
`(1-year bands) strata, then multiplying the person-years by
`the corresponding mortality rate for the general population
`of the relevant country. We then added the expected
`numbers across the strata and calculated SMRs as the
`sum of observed number of deaths divided by the sum of
`expected number of deaths. All cause population mortality
`rates were available for: Denmark from 1901–2011, New
`Zealand 1948–2008, Netherlands 1901–2012, and UK
`1922–2013. Tests for linear trends and heterogeneity of
`SMRs were based on likelihood ratio tests comparing the
`deviance between relevant Poisson regression models. We
`did all analyses with Stata version 13.1.
`
`Role of the funding source
`There was no funding source for this study. The
`corresponding author had full access to all the data in the
`
`study and had fi nal responsibility for the decision to
`submit for publication. The fi nal text was approved by all
`coauthors.
`
`Results
`Census dates (ie, the last date on which information was
`available for all patients from a specifi c centre) at which
`the data were frozen were: Dec 31, 2013 (Birmingham),
`Dec 31, 2009 (Oxford), Jan 12, 2014 (Stoke-on-Trent),
`Jan 1, 2014 (Leiden), Jan 1, 2012 (Aalborg), and Jan 1, 2012
`(Auckland) or earlier. We obtained data for 320 patients
`with 3790 person-years of follow-up from 10 years after
`cure (female to male ratio of 3:1). No patients were lost to
`follow-up. There were no signifi cant diff erences between
`women and men with respect to age at study entry, cure,
`
`Age at study entry (years)
`Age at cure (years)
`Age at census date (years)
`Diabetes*
`Hypertension*
`Receiving glucocorticoid
`replacement therapy treatment†
`Receiving thyroxine treatment‡
`
`Women
`(n=246)
`
`46 (38–56)
`36 (28–46)
`58 (51–69)
`24 (10%)
`125 (51%)
`158 (64%)
`
`Men
`(n=74)
`
`43 (32–52)
`33 (22–42)
`53 (46–62)
`7 (9%)
`39 (53%)
`52 (70%)
`
`95 (39%)
`
`28 (38%)
`
`Data are median (IQR) or n (%). *Defi ned as receiving treatment at census date or
`death. †Information missing for seven women and two men. ‡Information missing
`for 28 women and ten men.
`
`Table 1: Baseline characteristics of the cohort
`
`Pituitary surgery only
`Pituitary surgery plus bilateral
`adrenalectomy
`Pituitary surgery plus radiotherapy*
`Bilateral adrenalectomy only
`Bilateral adrenalectomy plus
`radiotherapy
`Radiotherapy only*
`Radiotherapy plus bilateral
`adrenalectomy plus pituitary surgery
`Pituitary surgery as fi rst treatment
`Bilateral adrenalectomy as fi rst
`treatment
`Radiotherapy as fi rst treatment
`Number of treatment methods
`One
`Two
`Three or more
`
`Women
`(246)
`
`144 (59%)
`40 (16%)
`
`20 (8%)
`7 (3%)
`15 (6%)
`
`16 (7%)
`4 (2%)
`
`204 (83%)
`22 (9%)
`
`Men
`(74)
`
`45 (61%)
`12 (16%)
`
`5 (7%)
`2 (3%)
`5 (7%)
`
`5 (6·8%)
`0 (<1%)
`
`62 (84%)
`7 (9%)
`
`20 (8%)
`
`5 (7%)
`
`160 (65%)
`54 (22%)
`32 (13%)
`
`47 (64%)
`17 (23%)
`10 (14%)
`
`Data are n (%). *These patients were given treatment with metyrapone, with or
`without aminoglutethamide, for diff erent periods while awaiting the eff ects of
`radiotherapy.
`
`Table 2: Treatments categorised by sex
`
`www.thelancet.com/diabetes-endocrinology Vol 4 July 2016
`
`571
`
`Teva Pharmaceuticals USA, Inc. v. Corcept Therapeutics, Inc.
`PGR2019-00048
`Corcept Ex. 2067, Page 3
`
`

`

`Less than 10% (29/320) had bilateral adrenalectomy or
`radiotherapy (10%; 25/320) as fi rst treatment and these
`participants were diagnosed and treated in the early years
`of the study (mostly before 1975). Only 13% (43/320) of
`the cohort required three or more treatments to achieve
`cure. The median patient follow-up was 11·8 years (IQR
`17–26) from study entry and did not diff er between
`countries.
`51 (16%) participants died during follow-up from study
`entry (10 years after cure), and the proportions were
`similar for women and men. The median age at death for
`both sexes combined was 68 years (IQR 60–78). Median
`(50%) survival for women was 31 years (IQR 19–38) from
`study entry and 28 years for men (IQR 24–42; non-
`signifi cant diff erence) (fi gure 1). The median age at death
`was 69 years (IQR 60–78) for women and 66 years (59–77)
`for men (table 3), and the median time to death from
`10 years after cure was 14 years (IQR 9–20) for women, 16
`(8–25) for men. Among patients who died, we noted no
`sex-related diff erences in age at cure, age at study entry,
`age at death, or duration of follow-up (table 3). However,
`when data for the sexes were combined and the deceased
`group compared with the alive group, we noted
`diff erences: the median age at cure for patients who were
`alive was 34 years (IQR 27–43) years versus 42 years
`(32–55) for those who were dead, p<0·0001. The median
`age at study entry for live patients was 44 years (IQR
`33–54) years versus 52 years (42–66) for those who were
`dead, p<0·0001. The median duration of follow-up from
`study entry for live patients was 11 years (IQR 6–15)
`versus 14 (9–20) for those who were dead, p=0·008.
`The causes of death were predominantly circulatory
`(n=30), including cardiovascular and cerebrovascular
`disease, ruptured aortic aneurysm, and pulmonary
`embolism, in both women and men but all nine cancer
`deaths occurred in women. Only two deaths from sepsis
`were recorded (table 3), despite 68% (210/311) of the
`cohort being on glucocorticoid replacement.
`The standardised mortality ratio (SMR) for all cause
`mortality was 1·61 (95% CI 1·23–2·12) indicating a
`61% increase in mortality risk (table 4). In absolute terms
`this translates to fi ve excess deaths per 1000 individuals
`per year beyond those which would be expected on the
`basis of general population mortality rates. There was
`evidence for heterogeneity (pheterogeneity=0·02) between
`countries, with the Netherlands showing a low absolute
`observed risk and New Zealand a high one (table 4). This
`heterogeneity disappeared after adjustment for pituitary
`surgery only (appendix) with fewer patients from New
`Zealand having pituitary surgery. Although men did not
`have an increased SMR and women did, this sex-related
`diff erence was not signifi cant. The SMR for circulatory
`disease was 2·72 (95% CI 1·88–3·95; p<0·0001) with no
`heterogeneity between countries (p=0·22). There was no
`signifi cant variation in circulatory SMR by time since
`study entry (ie, duration of follow-up). SMR for cancer
`(0·79, 95% CI 0·41–1·51; p=0·41) was not increased.
`
`Female
`Male
`
`(n=74)
`
`Articles
`
`100
`
`75
`
`50
`
`25
`
`Survival (%)
`
`50
`
`00
`
`(n=246)
`
`40
`
`11
`
`20
`30
`Time after 10 years
`
`83
`
`37
`13
`
`10
`
`147
`44
`
`0
`
`0
`
`Number at risk
`Female
`Male
`
`246
`74
`
`Figure 1: Kaplan–Meier survival curves stratifi ed by sex
`
`Number of deaths
`Age at cure years (deceased)
`Age at entry years (deceased)
`Age at death years
`Follow–up to death* (years)
`Duration of follow–up for live
`participants* (years)
`Causes of death
`Cardiovascular†
`Cancer‡
`Pulmonary embolism
`Disseminated sepsis
`Other or unknown§
`
`Women
`(246)
`
`41 (17%)
`42 (34–56)
`52 (44–66)
`69 (60–78)
`14 (9–20)
`11 (6–16)
`
`18 (44%)
`9 (22%)
`2 (5%)
`2 (5%)
`10 (24%)
`
`Men
`(74)
`
`10 (14%)
`43 (23–52)
`53 (33–62)
`66 (59–77)
`16 (8–25)
`11 (6–15)
`
`8 (80%)
`0
`2 (20%)
`0
`0
`
`Data are n (%) or median (IQR). *From study entry; ie, 10 years after remission of
`hypercortisolism, add 10 years to each for total follow–up from time of remission.
`†Included cerebrovascular–related deaths (n=6) and ruptured aortic aneurysm (n=3).
`‡Cancer types were lung (n=2), nasopharyngeal (n=1), bladder (n=1), metastatic
`uterine (n=1), malignant small bowel (n=1), pancreatic (n=1), meningioma (n=1),
`and metastatic primary unknown (n=1). §Included acute asthma (n=1), fi brosing
`alveolitis (n=1), chronic obstructive pulmonary disease or bronchopneumonia (n=2),
`acute bronchitis (n=1), old age (n=1), and unknown (n=2)
`
`Table 3: Summary statistics for deaths categorised by sex
`
`or census date (table 1), so data from the sexes were
`pooled
`for survival analysis and predictors. The
`proportions with diabetes, hyper tension, or receiving
`glucocorticoid replacement therapy were also similar
`between the sexes. We noted no diff erences in treatment
`methods between women and men (table 2). More than
`80% (266/320) of the cohort were initially given trans-
`sphenoidal pituitary surgery with 60%
`(189/320)
`requiring only this treatment to achieve a long-term cure.
`However, 40% (128/320) required additional treatment.
`
`572
`
`www.thelancet.com/diabetes-endocrinology Vol 4 July 2016
`
`Teva Pharmaceuticals USA, Inc. v. Corcept Therapeutics, Inc.
`PGR2019-00048
`Corcept Ex. 2067, Page 4
`
`

`

`Articles
`
`Participants (%) Observed/expected SMR (95% CI)
`
`p value
`
`AER*
`
`320 (100%)
`320 (100%)
`320 (100%)
`
`135 (42%)
`75 (23%)
`36 (11%)
`74 (23%)
`··
`
`Cause of death
`Overall
`Cancer
`Circulatory
`Country
`England
`New Zealand
`Netherlands
`Denmark
`Pheterogeneity
`Sex
`Male
`Female
`Pheterogeneity
`Year of cure
`<1985
`≥1985
`Pheterogeneity
`Follow-up (years)
`10–14
`15–19
`20–24
`25+
`Ptrend
`Pituitary surgery
`143 (45%)
`No
`177 (55%)
`Yes
`··
`Pheterogeneity
`Glucocorticoid replacement therapy
`No
`101 (32%)
`Yes
`210 (66%)
`Missing
`9 (3%)
`··
`Pheterogeneity
`
`74 (23%)
`246 (77%)
`··
`
`89 (28%)
`23·1 (72%)
`··
`
`56 (18%)
`85 (27%)
`86 (27%)
`93 (29%)
`··
`
`51·0/31·7
`9·0/11·4
`28·0/10·3
`
`15·0/9·3
`20·0/7·3
`4·0/6·0
`12·0/9·1
`··
`
`10·0/9·0
`41·0/22·6
`··
`
`28·0/15·9
`23·0/14·7
`··
`
`8·0/7·8
`13·0/8·4
`15·0/7·1
`15·0/8·4
`··
`
`35·0/13·6
`16·0/17·1
`··
`
`10·0/9·9
`39·0/19·6
`
`1·61 (1·23–2·12)
`0·79 (0·41–1·51)
`2·72 (1·88–3·95)
`
`1·62 (0·98–2·69)
`2·73 (1·76–4·23)
`0·66 (0·25–1·77)
`1·33 (0·75–2·33)
`··
`
`1·11 (0·60–2·06)
`1·81 (1·33–2·46)
`··
`
`1·76 (1·17–2·54)
`1·57 (1·04–2·35)
`··
`
`1·03 (0·51–2·06)
`1·55 (0·90–2·67)
`2·12 (1·28–3·51)
`1·79 (1·08–2·96)
`··
`
`2·58 (1·85–3·59)
`0·94 (0·57–1·53)
`··
`
`1·01 (0·54–1·89)
`1·99 (1·45–2·72)
`
`··
`
`··
`
`0·0001
`0·41
`<0·0001
`
`0·06
`<0·0001
`0·41
`0·33
`0·02
`
`0·75
`<0·0001
`0·14
`
`0·003
`0·032
`0·68
`
`0·94
`0·12
`0·003
`0·03
`0·17
`
`<0·0001
`0·80
`0·0005
`
`0·96
`<0·0001
`
`0·04
`
`5
`0
`5
`
`4
`14
`0
`4
`··
`
`1
`6
`··
`
`7
`4
`··
`
`0
`4
`12
`12
`··
`
`11
`0
`··
`
`0
`8
`
`··
`
`There was a tendency for increased SMR with longer
`follow-up (>20 years) although this did not reach
`signifi cance. We calculated SMR by decade of treatment
`but no signifi cant trend was observed (ptrend=0·17). We
`examined SMR by year of diagnosis before and after
`1985, this year being chosen as about midway from the
`earliest patient entry (1957) and the close of databases
`(average 2012); results were not signifi cantly diff erent
`(table 4, appendix).
`Half the patients in the cohort survived for 25 years or
`more from study entry (ie, 35 years from cure; fi gure 1).
`Presence of diabetes had a mortality hazard ratio (HR) of
`2·82 (95% CI 1·29–6·17; p<0·0096); for hypertension
`HR was 1·59 (0·77–3·31, p=0·08). Patients not taking
`glucocorticoid treatment had an SMR no diff erent from
`that of the general population, compared with patients
`taking gluco corticoids who had a signifi cantly diff erent
`SMR (table 4). 57% (108/190) of patients having pituitary
`surgery as
`the only
`treatment were
`receiving
`glucocorticoid
`replacement
`therapy, whereas
`the
`proportion who received other or additional treatments
`was 79% (102/130), a highly signifi cant diff erence
`(p<0·0001). When we applied multivariate analysis, the
`eff ect of being on glucocorticoids was no
`longer
`signifi cant, either alone (HR 1·57, 95% CI 0·78–16·0;
`p=0·21) or with pituitary surgery in the model (HR 1·31,
`0·63–2·69; p=0·47). Hypothyroidism was not a risk for
`mortality. Being on sex hormones was not a risk factor
`(HR 1·46, 95% CI 0·54–3·97; p=0·46), but data were only
`available for 51% (162/320) of the cohort. There was an
`association between mortality and number of treatments,
`with higher risk noted with a higher number of
`treatments (fi gure 2). HRs were 1·77 (95% CI 0·93–3·38)
`for
`two versus one
`treatments
`(p=0·08), and
`2·6 (1·15–5·87) for three versus one treatments (p=0·02).
`Median survival time from study entry for one treatment
`was 33 years (95% CI 26–38), for two was 27 years
`(19–28), and for three was 21 years (17–21). When
`considered by centre, Stoke-on-Trent, Oxford, and
`Auckland had the lowest proportions of patients receiving
`one treatment and the highest receiving three treatments
`(appendix). There was a signifi cant diff erence (p<0·001
`Anova) in the number of treatments between centres.
`Patients given pituitary surgery as a fi rst and only
`treatment had an SMR that did not diff er from that of the
`general population (0·94, 95% CI 0·57–1·53) versus
`those who did not (2·58, 1·85–3·59) the diff erence being
`highly signifi cant (p<0·0005). There was no heterogeneity
`between countries in terms of the number of patients
`having pituitary surgery (appendix). Moreover, patients
`cured by pituitary surgery only had a longer median
`survival of 31 years (95% CI 26–38) versus 24 years
`(21–28) if radiotherapy had been required at any time
`(p=0·03). Median survival for patients requiring bilateral
`adrenalectomy at any time was not reached, although for
`75% survival it was 17 years (95% CI 12–21) versus
`26 years (20–28) for pituitary surgery only (p=0·1).
`
`Table 4: Standardised mortality ratios (SMRs) with date of entry as date of ‘cure’ plus 10 years
`
`Discussion
`For patients who have been cured of Cushing’s disease
`for 10 years or more, treatment complexity and an
`increased number of treatments, refl ecting disease that
`is more diffi cult to control, appears to negatively aff ect
`survival. Pituitary surgery alone achieves a mortality
`outcome that is not diff erent from the normal population,
`and should be performed in a centre of excellence.
`The
`long-term outcome of successfully
`treated
`Cushing’s disease has been uncertain. Short-term
`studies10–18 and meta-analyses have produced confl icting
`results,13,19 not least because these studies were hetero-
`geneous with respect to timing after initial treatment,
`how cure or remission was defi ned, and methods of
`treatment, and they were of short duration. Findings
`from our fi rst meta-analysis13 suggested that SMR was
`not signifi cantly increased compared with that of the
`general population, provided that patients achieved
`initial remission. But we were cautious in our conclusion
`
`www.thelancet.com/diabetes-endocrinology Vol 4 July 2016
`
`573
`
`Teva Pharmaceuticals USA, Inc. v. Corcept Therapeutics, Inc.
`PGR2019-00048
`Corcept Ex. 2067, Page 5
`
`

`

`was not signifi cant and should be interpreted with
`caution because only a quarter of deaths occurred in
`men and the proportion of men that died was the same
`as the proportion of women who died. This fi nding
`suggests that more data are required for men. Women
`and men did not diff er with respect to demographic
`characteristics,
`treatment methods or number of
`treatments, substitution therapy with gluco corticoids,
`or comorbidities of diabetes or hypertension. Because
`the sexes were well-matched, we combined the data for
`survival analysis and to determine predictors of survival.
`Compared with fi ndings from our meta-analysis13 and
`those of van Haalan and colleagues,19 the upper CI for
`the overall SMR in our present study was reduced to
`2·12, and in absolute terms the excess mortality is
`small, amounting to fi ve of 1000 patient-years of follow-
`up. Moreover, median duration of survival after cure in
`this highly selected cohort was long at about 40 years
`(time from study entry plus 10 years). Causes of death
`were predominantly circulatory, including deaths from
`ischaemic heart disease, stroke, ruptured aortic
`aneurysms, and pulmonary emboli. Too few deaths
`occurred to be able to analyse these separately or to
`compare them meaningfully between the sexes. These
`fi ndings substantiate those from previous studies13,19 in
`which cardio vascular disease was the main cause of
`excess mortality. Cancer was not a contributor to excess
`mortality in our study, which is in keeping with data
`from studies of patients with other forms of pituitary
`tumours, in which cancer deaths were not increased.22
`With regard to the cancer deaths in women in our study,
`these were from various primary sources and none
`predominated.
`Perhaps surprisingly, given that two-thirds of our
`cohort was
`receiving glucocorticoid
`replacement
`therapy, only two deaths occurred that were directly
`attributed to disseminated sepsis. When considering
`what drives excess overall and circulatory mortality, the
`total
`duration
`and
`degree
`of
`exposure
`to
`hypercortisolism might be important. Such exposure is
`determined by several factors such as, but not
`exclusively: the time taken to reach a diagnosis, which
`might be several years for a rare and insidious
`condition, the time taken from treatment to cure, which
`will depend on the number and methods of treatment,
`and post-treatment requirement for glucocorticoid
`replacement (a higher dose of gluco corticoid treatment
`has been shown to be an independent predictor of
`mortality in patients with acromegaly).23 In our study,
`data were not available on type, dose, frequency, or
`duration of glucocorticoid treatment. Were these data
`available, their variability between centres and over an
`extended time period would probably not have created a
`large enough subgroup for meaningful analysis. And,
`although we show an increase in SMR in patients
`treated with glucocorticoids, this increase is not an
`independent risk factor in Cox regression analysis.
`
`One treatment
`Two treatments
`Three treatments
`
`(n=71)
`
`(n=42)
`
`(n=207)
`
`Articles
`
`100
`
`75
`
`50
`
`25
`
`Survival (%)
`
`50
`
`000
`
`40
`
`100
`
`30
`20
`Time after 10 years
`
`821
`
`34
`15
`4
`
`10
`
`124
`47
`21
`
`0
`
`0
`
`Number at risk
`One treatment
`Two treatments
`Three treatments
`
`207
`71
`42
`
`Figure 2: Kaplan-Meier survival curves stratifi ed by number of treatments
`
`that “patients in remission do not appear to have worse
`mortality but this should be regarded as preliminary.”13
`This caution was justifi ed because fi ndings from two
`later studies—one from New Zealand,14 and one from
`Oxford17—showed signifi cantly
`increased SMRs for
`patients with Cushing’s disease in remission, although a
`large study from Bulgaria18 did not. Why such a
`discrepancy exists between these later reports is unclear.
`Results from the latest meta-analysis with these three
`studies included19 shows that the overall SMR for
`Cushing’s disease initially in remission was 2·5 (95% CI
`1·4–4·2), still with evidence of signifi cant heterogeneity.
`In an attempt to clarify this issue, we deliberately
`selected patients in the present study whose hyper-
`cortisolism had been cured for at least 10 years before
`study entry and remained cured for the duration of
`follow-up until census date or death. With this selected
`cohort of more than 300 participants and more than
`3500 person-years of follow-up since 10 years after cure,
`we showed that overall excess mortality
`is 61%
`(SMR=1·61) higher than that of the general population
`from which they derive. However, we also show that
`within cohorts of patients with Cushing’s disease given
`d