Surgery outcomes for pancreatic cancer: linkage of the APDC with the NSW CCR Efty Stavrou, Deborah Baker, Nicola Creighton, James Bishop

1
Surgery outcomes for pancreatic cancer linkage
of the APDC with the NSW CCREfty Stavrou,
Deborah Baker, Nicola Creighton, James Bishop
2
Background

• Incidence and mortality rates pancreatic cancer
  stable over past 10 years
• Risk factors include
• Age
• Smoking
• Increased BMI
• Occupational exposure to chemicals
• Long standing diabetes mellitus
• Genetics/family history
• 90 ductal adenocarcinomas
• Five-year survival of disease is poor (6 males
  8 females)
• Fifth most common cause of cancer deaths in
  Australia

3
Background

• Patients present with symptoms late in course of
  disease
• Symptoms include
• Upper abdominal and back pain
• Jaundice
• Unexplained weight loss
• Nausea / vomiting
• Onset of diabetes, attack of pancreatitis
• Surgery is only potential cure for pancreatic
  cancer
• Only 20-30 undergo pancreatic resection due to
  late presentation of disease
• In past, high post-operative mortality and
  morbidity
• High volume hospitals have low surgical mortality
• Surgery improves five-year survival

4
Aims

• To investigate
• (i) what patient and cancer characteristics are
  associated with a pancreatectomy being undertaken
• (ii) what difference exists in survival between
  those undergoing pancreatic resection and those
  who do not
• What factors are associated with overall survival
• All primary invasive cases of pancreatic cancer
  diagnosed in NSW residents between 2000 and 2005,
  inclusive, and recorded in the NSW Central Cancer
  Registry (CCR) will be linked to episodes of care
  in NSW public and private hospitals recorded in
  the NSW Admitted Patient Data Collection (APDC)
  databases.

5
Methods

• Data sources
• The Central Cancer Registry (CCR) receives
  notifications of cancer in NSW. It is managed by
  the Cancer Institute NSW for the NSW Health
  Department, and operates under the authority of
  the Public Health Act 1991. The Registry
  maintains a record of all cases of cancer
  diagnosed in NSW residents since 1972. The
  registry is run according to the International
  Association of Cancer Registries guidelines.
• The Admitted Patients Data Collection (APDC)
  includes records for all hospital separations
  from all NSW public and private hospitals and day
  procedure centres and is maintained by NSW Health
  Department.

6
Methods

• Master Linkage Key
• Identifying information (eg name, address, date
  of birth, gender) from the CCR and the APDC were
  included in the Master Linkage Key (MLK)
  maintained by the Centre for Health Record
  Linkage (CHeReL) for record linkage. Records
  were matched using probabilistic record linkage
  methods and ChoiceMaker software. The CCR and
  the APDC data were linked as follows
• NSW CCR cases for pancreatic cancer (ICD-O3
  codes C25.0-C25.9) diagnosed between 1 July 2000-
  31 December 2005 (inclusive).
• APDC records for the period 1 July 2000 30
  June 2006 (to allow patients diagnosed in the
  latter half of 2005 to receive treatment).

7
CCR (CI NSW)
APDC
CCR ID (encrypted) Identifying information (e.g.
name, sex, DOB, )
Patient ID (encrypted) Identifying information
(e.g. name, sex, DOB, )
LINKAGE UNIT (CheReL) Matching CCR Identifying
information with APDC identifying
information Allocate a person identifier for each
person in the research project (Project Person
ID) Send person ID and source ID to relevant data
custodians
APDC Receives file from linkage unit Adds
ASLWH data removes ALSWH ID
CCR (CI NSW) Receives file from linkage
unit Adds CCR data and removes CCR ID
Patient ID (encrypted) Project Person ID
CCR ID (encrypted) Project Person ID
ANALYSIS GROUP Merge the CCR and APDC
de-identified files to create the analysis
dataset
Project Person ID Patient data (e.g. hospital
code, procedure type, area health service of
hospital etc)
Project Person ID CCR data (e.g. type of cancer,
date of diagnosis)
Project Person ID CCR data (e.g. type of cancer,
date of diagnosis) APDC data (e.g. procedure
type, procedure date)
8
Methods

• Master Linkage Key
• Two project keys were created from the Master
  Linkage Key
• A CCR Key containing the encrypted record number
  from CCR and the Project Person Number for that
  record which was returned to CCR data custodian.
• An APDC Key containing the encrypted record
  number from APDC and the Project Person Number
  for that record which was returned to APDC
  custodian.
• The custodians of the APDC and the NSW CCR
  decrypted the encrypted source record numbers
  from the data-sets and attached the relevant
  information to the data-sets which were released
  to the study investigators. The investigators
  matched the datasets using the Project Person
  Numbers.
• A total of 28,249 APDC records for the specified
  time period were matched to 3,834 pancreatic
  cancer cases in the CCR. The parameters for the
  extract from the MLK were such that the false
  positive and false negative rate were each lt0.1.

9
Methods

• Other variables
• Cases were allocated to the 2001 local government
  areas (LGA) based on residential address at the
  time of diagnosis.
• Accessibility/ Remoteness Index for Australia
  (ARIA) values were applied to cases via the LGA
  based on the 2001 census information from the
  Australian Bureau of Statistics
• Socioeconomic status was estimated using the
  Index of Relative Socioeconomic Disadvantage
  (IRSD), one of four Socio Economic Indexes for
  Areas (SEIFA) created by the ABS.
• Degree of spread was based on the degree of
  spread at diagnosis, a variable derived from the
  NSW CCR from the maximum extent of the cancer
  across all reports and notifications dated within
  four months of the date of diagnosis.
• Pancreatic resection was based on ICD-10AM
  procedural codes (block 0978) and ICD10AM used as
  proxy co-morbidity measurement (CVD, Diabetes,
  COPD, Renal disease)
• Hospitals grouped into peer-groups bases on NSW
  Health peer hospital groups (eg Principal
  referral, Major Metropolitan)
• Hospitals grouped by number of pancreatectomies
  undertaken in five years into Low volume (10),
  Mid volume (11-20) and High volume (gt20)

10
Methods

• Statistical Analyses
• Examination of the factors associated with a
  patient undergoing pancreatectomy or not was
  performed using logistic regression and reported
  as crude (univariate) and adjusted odds ratios.
  Statistical significance was taken at the plt0.05
  (two-tail) level.
• Survival analysis was undertaken to examine the
  association between undergoing a pancreatectomy
  or not from date of diagnosis to date of death
  from any cause.
• People who were not known to have died by 31st
  December 2004 were censored at this date.
  Survival rates are reported using the
  Kaplan-Meier product limit method and Cox
  proportional hazard regression was used to
  estimate hazard ratios.

11
Results OR of undergoing a resection or not
Adjusted for ARIA, IRSD, AHS
Variable No Pancreatectomy (n27,703 ) Pancreatectomy (n442 ) Crude OR Adjusted OR p-value
Age group lt50 years 50-69 years 70 years 1,608 (5.8) 11,210 (40.5) 14,885 (53.7) 56 (12.7) 234 (52.9) 152 (34.4) 2.82 (1.96-4.08) 1.77 (1.41-2.23) Referent 2.13 (1.43-3.17) 1.30 (1.01-1.67) Referent lt0.001 0.04
Histology Adenocarcinoma Other specified carcinoma Unspecified carcinoma Unspecified types 14,306 (52.6) 739 (2.7) 9,139 (33.6) 3,028 (11.1) 363 (87.1) 18 (4.3) 27 (6.5) 9 (2.2) Referent 0.83 (0.48-1.43) 0.12 (0.08-0.19) 0.08 (0.03-0.19) Referent 0.96 (0.54-1.72) 0.15 (0.10-0.23) 0.12 (0.05-0.29) 0.89 lt0.001 lt0.001
Degree of spread Localised Regional Distant Unknown 6,044 (21.8) 5,468 (19.7) 9,853 (35.6) 6,338 (22.9) 104 (23.5) 248 (56.1) 61 (13.8) 29 (6.6) 2.41 (1.68-3.47) 6.98 (5.08-9.60) Referent 0.60 (0.35-1.03) 2.97 (2.03-4.35) 6.21 (4.44-8.68) Referent 1.11 (0.64-1.92) lt0.001 lt0.001 0.72
Respiratory disease No Yes 26,564 (95.6) 1,139 (4.1) 419 (94.8) 23 (5.2) Referent 1.43 (0.90-2.28) Referent 1.93 (1.17-3.18) 0.01
Cardiovascular disease No Yes 22, 835 (82.4) 4,868 (17.6) 298 (67.4) 144 (32.6) Referent 2.42 (1.93-3.04) Referent 2.93 (2.23-3.83) lt0.001
Glucose/Diabetes No Yes 23,128 (83.5) 4,575 (16.5) 330 (74.7) 112 (25.3) Referent 1.60 (1.25-2.06) Referent 1.21 (0.91-1.60) 0.20
Renal disease No Yes 26,310 (95.0) 1,393 (5.0) 426 (96.4) 16 (3.6) Referent 0.53 (0.28-1.00) Referent 0.98 (0.51-1.90) 0.95
CALD group English NESB 20,106 (72.6) 7,597 (27.4) 320 (72.4) 122 (27.6) 1.11 (0.87-1.42) Referent 1.26 (0.97-1.66) Referent 0.10
Tumour location Head Body Tail NOS Other 12,979 (46.7) 1,042 (3.8) 1,641 (5.9) 10, 818 (38.9) 1,331 (4.8) 284 (64.3) 11 (2.5) 48 (10.9) 58 (13.1) 41 (9.3) 4.77 (3.42-6.65) 2.55 (1.27-5.11) 5.69 (3.52-9.18) Referent 6.22 (3.79-10.21) 2.38 (1.68-3.38) 1.91 (0.94-3.89) 3.04 (1.84-5.03) Referent 2.77 (1.65-4.65) lt0.001 0.07 lt0.001 lt0.001
12
Results

• Survival
• Median (all-cause) survival time for those who
  underwent a resection was 591 days (95CI 510-700
  days)
• Median survival for those who did not undergo a
  resection was only 283 days (95CI 282-288
  days).
• After accounting for potential confounding
  factors, patients receiving a pancreatectomy had
  statistically better all-cause survival (HR 0.61,
  95 CI 0.51-0.73, plt0.001) than those who did
  not.

13
Results

• Other factors
• Better survival if
• Younger
• Local or regional degree of spread
• Diagnosed with adenocarcinoma
• Tumour located in Head of pancreas
• Live in major cities
• No co-morbidity

14
Future plan

• Following surgery, what factors (eg hospital
  peer-group hospital volume) are associated with
• extended length of stay,
• 30 day mortality,
• readmission,
• unplanned return to theatre and
• overall survival
• Dissemination of outcomes to AHS, surgeons
  practise changes?

15
Acknowledgements

• CHeReL
• Heather McElroy