Careers in Statistics

1
Careers in Statistics

• The Role of a Biostatistician
• Presentation at Central College
• Pella, Iowa

2
What comes to mind when you hear the word
statistics?

• Sports statistics?
• Vital health statistics?
• Just numbers?

3
Leading causes of death in the U.S.
4
Cancer Death Rates, for Men, US, 1930-2000
Rate Per 100,000
Lung
Stomach
Prostate
Colon rectum
Pancreas
Liver
Leukemia
5
Cancer Death Rates, for Women, US, 1930-2000
Uterus
Lung
Breast
Colon rectum
Stomach
Ovary
Pancreas
6
Tobacco Use in the US, 1900-2000
Per capita cigarette consumption
Male lung cancer death rate
Female lung cancer death rate
7
Trends in Overweight Prevalence (), Adults 18
and Older, US, 1992-2002
1992
1995
2002
1998
Body mass index of 25.0 kg/m2or greater Source
Behavioral Risk Factor Surveillance System,
CD-ROM (1984-1995, 1998) and Public Use Data Tape
(2002), National Center for Chronic Disease
Prevention and Health Promotion, Centers for
Disease Control and Prevention, 1997, 2000, 2003.
8
Discipline of Statistics

• The statistics presented here involve massive
  efforts to collect the data, followed by simple
  calculations.
• We see differences, but are they meaningful?
  Could they have happened by chance?
• If we see trends that we do not like, what can we
  do to effect change?

9
Discipline of Statistics

• Numbers carry information, but we need numerical
  reasoning to make their meaning clear.
• The discipline of statistics uses mathematical
  tools to quantify uncertainty and assign a
  likelihood that what we see could have happened
  by chance.
• Biostatistics is just statistics applied in the
  health sciences.
• Mathematics and math logic play critical roles.

10
Typical Research Questions

• We rely heavily on medications to treat diseases
  and medical conditions.
• How do we know that they work (are efficacious)
  and are safe?
• Heart disease is the 1 cause of death.
• How can we assess if an intervention is effective
  in reducing a persons risk?

11
Competing Possible Errors

• Any research study has the possibility of making
  one of two types of errors
• Type I Declaring a significant effect when in
  fact there is none.
• Type II Failing to find a significant
  difference when there is one.
• In statistics, we use mathematical theory and
  modeling to find optimal decision rules and
  choose sample sizes which will reduce the
  likelihood of these errors to acceptable levels.

12
Example Clinical Trials

• One type of research design to address such
  questions is known as a clinical trial.
• In a clinical trial, patients are randomly
  assigned to either the new drug/treatment or
  standard care.
• When possible, both patients and doctors are
  blinded to the assignment.

13
Clinical Trials

• IHAST2 Trial
• Intraoperative
• Hypothermia for
• Aneurysm Surgery Trial

Michael Todd MD, Bradley Hindman MD, William
Clarke PhD, James Torner PhD University of Iowa
and the IHAST2 Investigators. Supported by NIH
RO1 NS38554
14
IHAST1

• Hypothermia was first used in neurosurgery in
  1955.
• Based on research experience with animals, it was
  believed that cooling the body before
  neurosurgery would be protective against central
  nervous system (CNS) insults.
• At present, at least 50 of current open
  neurovascular cases are cooled (Pemberton, 2003).
• However, this was being done without a rigorous
  assessment of outcomes.

15
IHAST1

• In 1999, U. of Iowa researchers published the
  results of a small pilot clinical trial on this.
  
• They randomly assigned 114 patients with and
  without acute aneurysmal subarachnoid hemorrhage
  (SAH) to either cooling or normal body
  temperature before surgery
• Cooling target body temperature of 33.5? C
• Normal body temperature ??37?C

Hindman, Todd, et al. Neurosurgery. 1999
Jan44(1)23-32 discussion 32-3.
16
IHAST1
Outcome Glasgow Outcome Score at Three Months
17
IHAST1

• Is it beyond reasonable doubt that hypothermia
  is working?
• SAH patients (n52) 71 vs. 57 good
• Looks like hypothermia is helping.
• No SAH (n62) 77 vs. 76 good
• Looks like little effect of hypothermia.

18
IHAST1

• Answer
• No - there is reasonable doubt.
• Statistical methods were used to determine
• (1) These differences, given this sample size,
  could have reasonably occurred by chance .
• (2) It would need a randomized clinical trial of
  at least 900 patients to assure that we could
  detect a 10 improvement (from 65 to 75) in the
  Glasgow outcome score at three months with 90
  power.

19
IHAST2

• The follow-up International, multi-center
  randomized clinical trial was funded by NIH
  (IHAST2).
• 1000 SAH patients undergoing craniotomies for
  aneurysm clipping were randomized to either
  hypothermia or normal body temperature.
• Primary outcome measure Glasgow Outcome Score
  (GOS) at 3 months

20
IHAST2 Glasgow outcome score
The previous 71-57 difference has now shrunk to
a 66-63 difference
21
IHAST2

• Now, is it beyond reasonable doubt that
  hypothermia is working?
• This study had a large enough sample size to find
  a meaningful difference in good outcomes due to
  hypothermia (should it work) with high
  probability.
• Beyond reasonable doubt, it simply does not work
  as believed.

Many other outcomes were also tested. All
consistently showed no benefit.
22
Missouri Turkey Hunting Survey

• The Missouri Department of Conservation is
  interested in turkey hunting success rates.
• Can we obtain an estimate of the hunting success
  rate in every county on every day of the hunting
  season even if we dont have a sample from that
  county?
• (small area estimation)

23
Missouri Turkey Hunting Survey
24
Missouri Turkey Hunting Survey
25
Missouri Turkey Hunting Survey
26
Missouri Turkey Hunting Survey
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
What do statisticians do?

• Surveys
• political surveys
• National Health and Nutrition Examination Survey
• Government Operations
• Census Bureau
• FDA
• Scientific Research
• Cancer trends
• Survival probability
• Genetics
• Business and Industry
• Pharmaceuticals

Source http//www.amstat.org
32
Business Economics, Engineering, Marketing, Comput
er Science
Physical Sciences Astronomy, Chemistry, Physics
Areas where STATISTICS are used
Health Medicine Genetics, Clinical
Trials, Epidemiology, Pharmacology
Environment Agriculture, Ecology,
Forestry, Animal Populations
Government Census, Law, National Defense
Source http//www.amstat.org
33
Roles of a Biostatistician

• Guardian of the science.
• Help write the proposal to get this study funded.
• Maintain an unbiased perspective.
• Determine which data needed to be collected.
• Determine what methods would be used.
• Anticipate how proposed changes in the program
  would affect ability to make inferences and argue
  against them when needed.
• Develop new methods when needed.

34
Roles of a Biostatistician

• Keeper of the Data.
• Work with different state and federal agencies to
  obtain data.
• Check data for recognizable errors.
• Merge data from different sources.
• Maintain and secure the project database.
• Work with others (e.g., national evaluators,
  statistical coordinating centers) to blend your
  data with other data.

35
Roles of a Biostatistician

• Analytic Support
• Investigate the data from many angles.
• Assess assumptions of proposed models.
• Fit statistical models.
• Interpret results of the models for the research
  team.
• Make sure that the investigators do not draw
  conclusions beyond what can be supported by the
  data.
• Author (Write up results / provide graphics.)

36
What About Salaries?

• Annual (median) salary with 0-1.9 years of
  experience for
• Masters PhD
• - Statistician 60,000 83,000
• - Male Statistician 56,500 83,000
• - Female Statistician 62,000 83,000
• - Assistant Professor (4 year college) 53,000
• - Assistant Professor (Research) 63,000
• - Pharmaceutical statistician 73,000 83,000

Source http//www.amstat.org
37
Life as a Graduate Student

• What do you do as a graduate student?
• Classwork
• Teaching assistants
• Research assistants
• Go to football games!

38
Variety of scientific projects

• As you can tell, people work on various
  different projects. The following are brief
  excerpts from some recent projects.

39
Driving Simulator (NADS)
A shared-use research facility operated by The
University of Iowa, The NADS was created to
conduct research which will help save lives,
reduce the costs of vehicle crashes, and lead to
better and safer vehicles, roadways, and drivers.
http//www.nads-sc.uiowa.edu/
40
Iowa Cochlear Implant Center

• While normal hearing has increased significantly,
  implants still do not reproduce a high quality of
  music.
• Is it beneficial to have two cochlear implants
  for those who are profoundly deaf?

41
Iowa Bridge to Employment Study

• Purpose To determine if special services
  provided to Iowans with disabilities would
  increase the likelihood of employment.

Mental Illness Group
Mental Retardation Group
42
Wise Woman
43
Total Physicians
http//www.public-health.uiowa.edu/FACTBOOK
44
Holden Comprehensive Cancer Center
Tumor growth curve estimates from a mixed linear
models analysis.
Group comparisons indicate differences between
specific groups at the 5 level of significance.
45
Spatial statistics
Radon concentrations measured in the basements of
homes in Iowa.
46
Thank you!

• Jake Oleson
• Central College
• class of 1997