1. Moving from SAS to Stata: Making customized tables in RTF using -rtfutil- and other packages
Thank you for your kind introduction, and the opportunity to give this talk.
The title of the talk is Clinical database management: From raw data through study tabulations to analysis datasets
Si litt om bakgrunn, CRO, akademia, SAS, Stata
Inge Christoffer Olsen, Phd
Diakonhjemmet Hospital, Norway

2. Background Used SAS when working for a norwegian CRO
Forced into SPSS when moving to Diakonhjemmet
Managed to force Stata onto the researchers at Diakonhjemmet
I will begin with a quote from the famous physicist Max Planck: ” An experiment is a question which science poses to Nature and a measurement is the recording of Nature's answer”
Meaning we cannot understand Nature without measurements. We need to take care of our measurements!

3. Background
I love Stata!
Programming is so efficient, clear and easy.

4. Background
But…
Programming is so efficient, clear and easy.

5. Stata output
How do you come from here:

6. Final table …to here? Diagnosis Male Female Total Rheumatoid arthritis21 57 78
Spondyloarthritis 76 15 91
Psoriatric arthritis 30
Ulcerative colitis 61 32 93
Crohn’s disease 92 63
155
Psoriasis 5 35
295
187
482

7. Cut and paste 1: Text
tabulate diagnosis sex if visit_no== | Sex Diagnosis | Male Female | Total Rheumatoid arthritis | | Spondyloarthritis | | Psoriatric arthritis | | Ulcerative colitis | | Crohn's disease | | Psoriasis | | Total | |
*Possible to remedy by using fixed width fonts, but still only text

8. Cut and paste 2: HTML * Sometimes OK, but unreliable Randomised and
Sex
Yes ----
Diagnosis
Male Female
Rheumatoid arthritis
21 57
Spondyloarthritis
76 15
Psoriatric arthritis
15 15
Ulcerative colitis
61 32
Crohn's disease
92 63
Psoriasis
30 5
* Sometimes OK, but unreliable

9. Cut and paste 3: Picture
Nice, but useless

10. Dissapointed Only major feature where Stata is inferior to SAS or SPSS
Forced to enter results manually
Error prone!

11. Reports
In SAS it was relatively easy to report tables from statistical analyses in RTF format using PROC REPORT
Characteristic
Statistic
Abatacept N = 6
Adalimumab N = 133
Anakinra N = 2
Certol. Pegol N = 190
Etanercept N = 203
Golimumab N = 327
Infliximab N = 46
Rituximab N = 42
Tocilizumab N = 30
Age (Years) (a) n 6
133 2
190
203
327 46 42 30
Mean
47.77
42.82
46.40
51.48
45.47
44.12
46.05
55.27
53.73
Std. Dev.
15.47
13.57
27.44
14.72
13.62
13.14
13.34
12.21
12.77
Median
51.60
41.20
53.75
45.00
42.90
43.60
54.95
54.30
Min/Max
26.5/64.1
19.0/74.7
27.0/65.8
18.7/82.5
19.3/76.3
17.7/80.5
25.5/79.0
30.7/78.0
29.8/73.7
Sex, n(%)
Male
1 ( 16.7)
64 ( 48.1)
0 ( 0.0)
44 ( 23.2)
80 ( 39.4)
151 ( 46.2)
20 ( 43.5)
6 ( 14.3)
4 ( 13.3)
Female
5 ( 83.3)
69 ( 51.9)
2 ( 100)
146 ( 76.8)
123 ( 60.6)
176 ( 53.8)
26 ( 56.5)
36 ( 85.7)
26 ( 86.7)
No. of prev. Biologics (b) 77 85 62
114 29 32 27
3.7
1.5
4.0
2.0
1.4
2.1
2.2
2.3
0.8
0.7
1.2
1.0
1.3
1.1
3.5
3/5
1/4
1/6
1/8
1/7
1/5
No. of prev. DMARDs
129
172
181
280 40 36 28
2.7
3.0
1.9
2.5
1.6
0.9
2.8
0/4
0/6
0/5
0/3
Biologics Naive, n(%)
Yes
52 ( 40.3)
87 ( 50.6)
119 ( 65.7)
166 ( 59.3)
11 ( 27.5)
4 ( 11.1)
1 ( 3.6) No
6 ( 100)
77 ( 59.7)
85 ( 49.4)
62 ( 34.3)
114 ( 40.7)
29 ( 72.5)
32 ( 88.9)
27 ( 96.4)

12. Reports
Most work a statistician or researcher do in Stata usually ends up in a report or article
Most reports or articles are presented in a document, usually Word (or pdf using Latex)
Unsatisfying to rely on manually entering tables from Stata output
Natively it is possible to export raw results to Excel, but this still mandates a lot of manual work
I hate manual work!
Must be a way to produce Word/RTF tables

13. Solution
Fortunately there is a lot of user-written programs for Stata (available through ssc)
I will present how RTF-tables can be produced using the package –rtfutil- in addition to results compiling packages such as (x)contract-, -(x)collapse-, and -parmest-

14. Aim Typical Table 2 in an article Variable Treatment 1 Treatment 2
Difference (95% CI)
Contvar 1, mean (SD)
1.5 (5.54)
0.7 (3.94)
1.29 ( )
Contvar 2, mean (SD)
1.6 (5.67)
0.7 (4.41)
1.09 ( )
Contvar 3, mean (SD)
0.3 (1.01)
-0.2 (1.38)
-0.02 ( )
Catvar 1, n(%)
10 (23.3)
7 (16.7)
6.6 ( )
Contvar 4, mean (SD)
0.1 (0.59)
-0.2 (0.67)
-0.01 ( )
Contvar 5 mean (SD)
0.1 (0.23)
0.0 (0.25)
0.08 ( )
Catvar 2, n(%)
29 (87.9)
39 (92.9)
-5 ( )
Contvar 6, mean (SD)
0.1 (1.28)
-0.2 (1.68)
0.14 ( )

15. Step 1 Use –xcollapse- by treatment to get the mean and SD
The resulting dataset 1 has two lines with mean and SD for each treatment
Restore the original dataset and run some regression analysis (e.g. by -mixed-)
Use –margins- to get the treatment difference
Use –parmest- to store the result dataset 2
Append the second dataset to the first
Add some variable indicating endpoint
Store in an endpoint specific temporary dataset

16. Step 2 Repeat step 1 for each continuous endpoint
Combine all endpoint specific datasets
var
varlab
Treatment
estimate
max95
min95
mean sd
contvar1
Contvar 1, mean (SD) 1 2 3
contvar2
Contvar 2, mean (SD)
contvar3
Contvar 3, mean (SD)
contvar4
Contvar 4, mean (SD)

17. Step 3 Create text variable using e.g. –sdecode- to make RTF-ready ,
varlab
Treatment
estimate
max95
min95
mean sd
text
contvar1
Contvar 1, mean (SD) 1
\qr{1.51 (5.544)} 2
\qr{0.67 (3.943)} 3
\qr{1.29 (-0.35 \endash 2.94)}
contvar2
Contvar 2, mean (SD)
\qr{1.56 (5.667)}
\qr{0.69 (4.412)}
\qr{1.09 (-0.68 \endash 2.86)}
contvar3
Contvar 3, mean (SD)
\qr{0.25 (1.009)}
\qr{-0.15 (1.376)}
\qr{-0.02 (-0.50 \endash 0.47)}
contvar4
Contvar 4, mean (SD)
\qr{0.065 (0.5879)}
\qr{ (0.6674)}
\qr{ ( \endash )}
\qr means right align, \endash is a semi-long dash

18. Step 4
Use –xrewide- (or reshape wide) to get to one line: xrewide text, i(var varlab) j(treatment)
var
varlab
text1
text2
text3
contvar1
Contvar 1, mean (SD)
\qr{1.51 (5.544)}
\qr{0.67 (3.943)}
\qr{1.29 (-0.35 \endash 2.94)}
contvar2
Contvar 2, mean (SD)
\qr{1.56 (5.667)}
\qr{0.69 (4.412)}
\qr{1.09 (-0.68 \endash 2.86)}
contvar3
Contvar 3, mean (SD)
\qr{0.065 (0.5879)}
\qr{ (0.6674)}
\qr{ ( \endash )}
contvar4
Contvar 4, mean (SD)
\qr{0.23 (0.988)}
\qr{-0.17 (1.337)}
\qr{-0.03 (-0.50 \endash 0.43)}

19. Step 5
Repeat step 1 to 4 for categorical and/or time-to-event variables and compile to a final results dataset
Sort the dataset according to the sequence you want to present the data

20. Step 6 Use the –rtfutil- package to write the results dataset to RTF
tempname handle2
rtfopen `handle2' using “Output/Table 2.rtf", template(minimal) replace paper(a4land) landscape
use work/total, clear
file write `handle2' " {\pard\b Typical Table 2 in an article \par}" _n
rtfrstyle varlab text1 text2 text3, cwidths( ) local(b d e)
listtab varlab text1 text2 text3, handle(`handle2') begin("`b'") delim("`d'") end("`e'") ///
head("`b’ \ql{\b Variable} `d' \qr{\b Treatment 1 }`d' \qr{\b Treatment 2} `d' \qr{\b Difference (95% CI)} `e'"
rtfclose `handle2'

21. Strengths
Produce ready-to-use tables to be inserted into an article or report
No manual work!
Possible to include .eps figures directly into RTF document
Quick once you have the results datasets

22. Weaknesses Quite a lot of programming (initially)
Can take a lot of tweaking to get the result exactly as you want
Inclusion of figures are not fully supported in RTF, need to open the document in Word and include the figure files

23. Tabulation Datasets (TD)
Final organisation
Raw output from eCRF
Imported into Stata
Tabulation Datasets (TD)
Analysis Datasets
Results dataset
RTF tables

24. Additional tips
The -project- module is fantastic for organizing and maintaining Stata projects
Utilizes checksums to keep overview of unchanged files, only recently changed do-files and dependent do-files will be run

25. Acknowledgements The –rtfutil- module is written by Roger B. Newson
The –project- module is written by Robert Picard

26. The end
Thank you!