Minutes from the '96 Meeting in Chicago
Hyatt Regency Hotel Downtown, Chicago, IL
August 3 - August 4, 1996
This was the sixth annual meeting of the Association of GCRC
Statisticians.
Notes on the Agenda
- There was a lunch on Saturday and a breakfast Sunday morning that
was included with the registration fee.
- Dinner on Saturday night was not included with registration.
David, the president, presented the problem that dinner was not
included in the registration fee. Many did not know this as it has
been the case in the past. Next year it will be included.
Gary Sexton, the treasurer, deferred the treasurer's report until
later. On April 11, 1997, the following report was attained from Gary:
ASSOCIATION OF GCRC STATISTICIANS TREASURERS REPORT 1996
A. Balance forward (transferred from J. Janowski) as of July 1, 1996
$1921.40
B. Income:
Registration 34 @ $150.00 $5100.00
Dinner 18 @ $45.00 $ 810.00
Over collected fees $ 120.00
Total Income $6030.00
C. Expenses:
Hyatt Regency (2 Lunches, 1 breakfast, 3 breaks) $2652.51
Tucci Benucch restaurant $ 806.45
Baur Audio Visual $ 352.20
Postage and supplies $ 50.49
Room Reimbursement (1 night for Dr. Heitjan) $ 120.22
Total Expenses $3981.87
D. Net Income 1996 (B minus C) $2048.13
E. Account Balance as of April 1, 1997 (A + D) $3969.53
PROPOSED BUDGET 1997
Expenses
Hotel Deposits
(2 lunches @ $22.00, 1 breakfast @ $20.00, 3 breaks @ $7.00,
1 working dinner @ $45.00) per registrant x 40 registrants
$5200.00
Audio Visual Equipment $ 400.00
Supplies for meeting (markers, wall boards, etc.) $ 125.00
Meeting Contingencies $ 250.00
Postage $ 100.00
Total estimated expenses $6075.00
Balance forward $3969.53
Proposed registration fee $125/person x 40 registrants $5000.00
Total projected resources $8969.53
Projected reserves (Reserves - expenses) $2894.53
the DSL manager, the secretary, did not have a report except related
issues in his Computing Committee report. The minutes from
last year's meeting were not distributed to everyone by last year's
secretary. They will eventually be approved via the email list and put
on the web. It was suggested that the Contact/Mailing list be sent
with the minutes in the future.
The only committee report was from the Computing Committee chair,
the DSL manager.
- The web site was announced as being online at
http://www.gcrc.ufl.edu/ags and the home page design was shown. The
members and contact information pages were explained. They are
available only for AGS member use.
- There have been complaints about the "From:" line in
email that comes from the UNC listserver. It shows the message as
being from an individual and not from the list. This is confusing to
some members and solutions to this problem will be investigated.
- In general there were no problems with our web site being public
except regarding the email list archive and the minutes. It was stated:
- The email list archives are no longer available via the web
although they are still available for being searched by sending a
message to the listserver "listserv@gcrc.med.unc.edu". This was agrred
upon as acceptable.
- The minutes will only be public (on the web) after approval by
the AGS via posting to the list.
David Schoenfeld reported that he gave a "data management and
quality" talk at the PDA meeting last March (1996). It seemed
well received, but there was little reaction. Some discussion came up
and Keith Muller suggested this be discussed another time due to the
importance of the topic.
NIH funding of the Biostatistician postions was discussed. The PDA
met with an NIH Comittee to recommend changes to the GCRC Program
Guidelines. Part of those recommendations were to increase the
biostatistician funding to a maximum of 1.0 and make it possible to
have support staff. Related to this, it was suggested that an old copy
of the Biostatistician position rewrite, from the 1992 meeting, should
be resurrected.
We ended the business session with stories from recent site
visits. Here were some things that were said:
- Minority and gender issues - make sure each protocol has table for
gender and minority and numbers do not contradict other figures in
grant with details about how active recruitment was done.
- Had a lot of computer questions due to having an SM as a
"statistical" reviewer.
- A raw statistician, new Ph.D., was a reviewer.
- MDs got hassled about consent forms.
- Each protocol needs a statistics section.
- One statistician was favorably reviewed for being a PI on a
protocol for training clinical biostatisticians on the GCRC.
- Short courses and training were almost universally well received.
- PD talk about Center needs to include comments about the
importance of Biostatistics at the unit.
- There were questoins among our group about Biostatisticians that
are Co-Investigators reviewing their own protocol. At some Centers
this was fine while other Centers avoid this.
- Within this talk there was dicussion on the setup and
responsibilities at different
sites: some have GRA assistance, some do analyses of data, there are
varying degrees of Biostatistician and SM interaction, some SMs do
data management in addition to computing work, some SMs are doing
Biostatistics.
This talk presented some of the theory, applications, difficulties of
missing data including such terms as missing, completely missing,
censored, rounded, dropouts, longitudinally missing, and ignorability
conditions. Specifically the concepts of missing at random, observed
at random, and missing completely at random were presented along with
the situations in which they are sufficient for ignorability. Also,
how ignorability is distinct for Bayesians and Frequentists was
demonstrated. Examples included missing binary data and right censored
data.
This initial session was a small group brainstorming exercise. The
members were divided into three groups and each group had a
facilitator and a recorder. We were supposed to come up with issues
and opportunities for the AGS. No value judgements at this point, just
idea generation. Below are the issues/opportunities that were
generated by all three groups (note: some are repeats):
- Regular discussions of site visits: format? mechanism? review by
statisticians.
- Optimal GCRC model: 1992 document as a starting point; expanded role.
- Funding data management: database design.
- Training grants for statisticians.
- Define data quality: role of statistician and CDMAS person.
- Educational outreach.
- GCRC PR.
- Is the AGS a subset of the JSM?
- Industry studies.
- Facilitate interaction between Centers: share resources.
- Extent of involvement in data analysis.
- Data quality
- personnel? CDMAS manager, statistician, data manager, etc.
- clinical investigators: do they record the proper data?
- ethical issues in data analysis, e.g., tossing outliers
- should we encourage database creation or just accept spreadsheets?
- Should we analyze the data?
- some centers require R01s, etc for stat work
- other centers provide "seed" money
- some assume GCRC support
- with regard to non-GCRC projects whom should we help?
- What % effort and % funding are appropriate?
- GCRC statisticians
- graduate students (MS-level individuals)
- Should standard statistical procedures be put into simple forms?
- Should we review industrial studies?
- Should some of our effort be involved with education/training?
- training clinical investigators in sound statistical methods
- involving biostat graduate students in clinical studies
- Use of statistics and its misuse, e.g., effect size determination,
making investigators "happy"?
- Dealing with statistical controversy.
- one-sided vs. two-sided
- 0.8 power vs. 0.9 power
- missing value problem
- others
- Should we have input with respect to protocol review at site
visits?
- Should we help find appropriate software for investigators?
- SAS/not for investigators
- Minitab
- BMDP/not for investigators
- Statistica
- JMP
- Stat Xact
- S+/not for investigators
- SPSS
- Hardware choice.
- Can we ever "prove" the null hypothesis? Sample size for
Phase I studies.
- Why are we meeting separately from other GCRC groups?
- What do we want the group to do?
- How can we improve the research?
- How do long term changes at NIH affect us?
- How can AGS help us?
- How can AGS help/affect outside?
- How do we stay funded?
- Job description - what should we be doing?
- Biostat departments supporting CRCs; institution resources
available, support of CDMAS required
- Variety in CRCs
- Should there be statistical protocols?
- How can we reduce redundancy in training?
- Should we review NIH protocols? How? AGS guidelines for reviewing.
- Answer to "help us": education, talks; email help;
sharing of solutions (?)
- Should statisticians be 100%? Yes: focus, commitment. No: not
willing to do it, professionally not good.
- Role of the AGS
- Sharing stories; site visit fiascos
- statistical support; answers to specific questions; trip
funding
- guidelines; statistical duties
- Role of the GCRC Statistician
- Interface with Biostat departments - other nearby stats.
- reviewing NIH protocols; advisory committee duties; executive
committee membership
- training and education services
- accounting and documentation for FTEs
Selected Topics
- Site visits/Staying funded: allocate block of time at each annual
AGS meeting; statistician on site visit.
- Oral reports (about 1 and 1/2 hours) perhaps also posted on the
Web and we could have a schedule of upcoming site visits on the web.
- Role of statisticians on site visit.
- Adjudication if poor review by non-statistician; review each
protocol.
- Qualification of statistical reviewer. GCRC or non-GCRC?
Statistician on review team will understand need - perhaps even be an
advocate.
- Amateur statistician can give inappropriate review.
- Pro forma if no request for expansion.
- Data Quality: no resources for data management.
- Require data archived on GCRC computer systems?
- Should we have a data quality seciton in protocol?
- Audit data? Resources? Personnel? Who's responsible? Need NIH
guidelines.
- What are the quality standard? R01? Lower? Co-op model?
- Statistician often has no involvement with the data, so what
responsibility is implied?
- Require forms with protocol?
- Statistician is not responsible to ensure investigator honesty.
- Review data management plan before protocol submission to SAB.
- Auditing is time consuming and expensive.
- Data falsification may "blow up" on statistician.
- Auditing often leads to hostility.
- Need to educate investigators.
- Quality control guidelines.
- CDMAS energy diverted to system administration.
For official motions and how they passed, please see Sunday's Workshop--priorities from the previous days dialogue
section of the mintues.
Again this dinner was not included in the price of registration and
was, as always, optional. Cost was $45 a plate. Guests welcome.
This talk introduced topics at an introductory level, but tried not to
spend too much time there. There was much discussion in the middle of
the talk about security issues and having your GCRC system broken into. It
was noted that this is a tough issue as you can always have more
security and it's not clear what is enough.
The presented material is available at
http://www.gcrc.ufl.edu/talks/ags96.
Taking the best of the best from yesterday's late afternoon session,
we prioritized the issues that were most important and arrived at this
list:
- Statisticians that have been recently site visited will give
oral presentations reporting on their experience. A motion was
made and formally passed by a unanimous vote to have formal sessions
regarding site visit experiences at all future AGS meetings. The
statisticians that have been site visited within the last year for
their Center's grant renewal will give oral reports to the group based
on their site visit experiences.
Another suggestion was made to have a schedule of site visits on
the web, but no motion was made with respect to this issue.
- Make a recommendation to NIH that statisticians should be
included on all site visit committees. A motion was made and
formally passed by a unanimous vote.
There was concensus as shown above that the work of GCRC
Statisticians should be reviewed by statisticians at site
visit. However, there was much discussion as to how we should make
this recommendation and specifically who is the statistician that
should do the reviewing? What are this person's qualifications? We
agreed that recommended qualified statisticians would include, but not
be limited to, "GCRC statisticians, past or present, or ones who
are qualified to hold such a position, or that works as a statistician
in single-institute core units, such as cancer centers, AIDS centers,
aging centers, etc." We were concerned that this was too vague or
appears too self-serving, but general agreement was reached that this
is acceptable wording.
- Data quality is important especially in the training of new
investigators and an educational document should be drafted to
distribute to new investigators. This motion was made and formally
passed by a unanimous vote. Specific quality guidelines will be
formally discussed at the next AGS
meeting and a working group was formed, chaired by Dick Kryscio, to
work via email.
Most of lunch was spent discussing the issues and opportunities as
described above. We officially voted to have the next meeting in
Anaheim, California which is the next location for the Joint
Statistical Meetings.
Return to the GCRC Home Page
This page is maintained by the DSL manager.
Last modified on August 14, 1998.
the DSL manager, MS / Data Services Mgr / dsl.mgr@gcrc.ufl.edu