Metan question

[Attila Nagy]

Hi,

I have a database containing data from pubmed articles (OAD, species, strains, initial number, start glucose, SD, end number, end glucose mean, end SD).
I would like to check which species/strain is closest to human change in case of a given drug?

Is metan which should be used? "mean1 SD1 n1 mean2 SD2 n2" and interpret effect size stratified by drugs and species?

I attached a sample excel.

Thanks,
Attila

[Nick Cox]

The FAQ Advice spells out in detail why spreadsheet attachments are deprecated here.

https://www.statalist.org/forums/help#stata

[Attila Nagy]

Thanks, It is in csv now

[Nick Cox]

Better, I guess, but still not what we ask for at the link specified.

I move on now. Others know much, much more about meta-analysis than I do.

[Carlo Lazzaro]

Attila:
why not using -dataex- to provide an excerpt/example of your data?

[Attila Nagy]

Thanks:
----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. To install: ssc install dataex
clear
input str20 a29_1_intervention_drug str7 a32_species str20 a33_strain float(a34_initial_number_of_animals a35_end_number_of_animals a42_start_glucose_mgdl_jav_2 a43_sg_sd_jav_2 a44_end_glucose_mgdl_jav_2 a45_eg_sd_jav_2)
"saxagliptin" "rat"   "wi"                    6   6      .         .    300         4
"saxagliptin" "mice"  "enos -/-"              8   8      .         .    390         .
"saxagliptin" "rat"   "SD"                   15  15      .         .  142.3 12.780845
"saxagliptin" "rat"   "Zucker"                7   7      .         .    115         6
"saxagliptin" "rat"   "GK"                    6   6      .         .    379 31.843367
"saxagliptin" "rat"   "Dahl salt-sensitive"  10  10      .         .    158  3.162278
"saxagliptin" "rat"   "Da"                   10  10      .         .    180 18.973665
"saxagliptin" "rat"   "Fisher"               10  10      .         .    549         .
"saxagliptin" "rat"   "SH"                    .   .    105         4    102         5
"saxagliptin" "mice"  "dbdb"                 25   8    550        10    267  5.656854
"saxagliptin" "mice"  "apoE-/-"              10   9  145.8         .    125         .
"linagliptin" "human" "hu"                   65  64  182.3        42    176        32
"linagliptin" "human" "hu"                   58  57  185.2        34    191 30.199337
"linagliptin" "human" "hu"                  157 113      .         .      .         .
"linagliptin" "human" "hu"                  106  93      .         .      .         .
"linagliptin" "human" "hu"                  631 618      .         .      .         .
"linagliptin" "human" "hu"                  523 513 169.62     43.51    158 22.649504
"linagliptin" "human" "hu"                  162 160  152.7      28.7    142  37.94733
"no"          "rat"   "wistar"                6   6      .         .    370        30
"no"          "mice"  "enos -/-"             12  12      .         .    376 3.1176915
"vehicle"     "rat"   "SD"                   15  15      .         .  147.3 15.879231
"vehicle"     "rat"   "Zucker "               7   7      .         .    136         9
"no"          "rat"   "GK"                    6   6      .         .    389  17.14643
"vehicle"     "rat"   "Dahl-S"               10  10      .         .    155  37.94733
"vehicle"     "rat"   "Dahl-S"               10  10      .         .  151.9 14.546477
"vehicle"     "rat"   "Fischer"              13  13      .         .    582 2.1633308
"vehicle"     "rat"   "SH"                    .   .    105         4    109         4
"vehicle"     "mice"  "dbdb"                 16   8    550         2    475 294.15643
"vehicle"     "mice"  "apoE-/-"              10  10  133.5  .5375872  132.1  .7905694
"placebo"     "human" "human"                71  70  185.5      38.8 199.13  35.97638
"placebo"     "human" "human"                67  63    183      34.9 187.24 34.923916
"linagliptin" "human" "human"               159 132      .         .      .         .
"placebo"     "human" "human"               120 105      .         .      .         .
"placebo"     "human" "human"               630 617      .         .      .         .
"placebo"     "human" "human"               177 175 166.42     41.89 176.88  37.04052
"placebo"     "human" "human"                79  78  144.1      29.6  154.2  37.97657
"saxagliptin" "rat"   "wistar"                6   6      .         .    105        10
"saxagliptin" "mice"  "enos -/-"              8   8      .         .    178 .56568545
""            ""      ""                      .   .      .         .      .         .
"saxagliptin" "rat"   "Zucker "               7   7      .         .    124         7
"saxagliptin" "rat"   "GK"                    6   6      .         .    366 14.696939
"saxagliptin" "rat"   "Dahl-S"               10  10      .         .    176 31.622776
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
"saxagliptin" "mice"  "apoE-/-"              10   9  180.2  .7589467    127       .96
"linagliptin" "human" "human"                66  62  189.3      42.4  168.1 35.275555
"linagliptin" "human" "human"                57  55  191.9      37.8  176.7 34.856133
"no"          "rat"   "wistar"                6   6      .         .    100        11
"no"          "mice"  "enos -/-"             12  12      .         .    178  .6928203
"no"          "rat"   "SD"                   15  15      .         .    145 16.266531
"vehicle"     "rat"   "Zucker "               7   7      .         .    129         7
"no"          "rat"   "wistar"                6   6      .         .    374  9.797959
"no"          "rat"   "Dahl-S"                6   6      .         .    140 31.843367
"no"          "rat"   "Dahl-S"                5   5      .         .  184.4  33.98823
"vehicle"     "mice"  "apoE-/-"              10  10  161.1 1.0119288  133.9  .6008328
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
"vehicle"     "rat"   "Zucker "               7   7      .         .    116         4
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
""            ""      ""                      .   .      .         .      .         .
"linagliptin" "human" "human"                66  66  188.7      42.4 172.83 35.177086
"linagliptin" "human" "human"                55  54  188.7      40.1 179.61  35.27265
"vehicle"     "rat"   "Zucker "               7   7      .         .    104         4
end

------------------ copy up to and including the previous line ------------------

[David Fisher]

Hi Attila,

A few, very brief, initial comments:

* Firstly, thanks for the dataex example. This really does make it easier for Forum members to read and examine your data. However, I would first rename the columns with something shorter, e.g. "intervention, species, strain, startN, endN, startmean, startsd, endmean, endsd" (assuming I've understood correctly!) Also, does each line represent a different study? Basic science is not my field, but I would assume that you'd want to know which study was which? (unless you have purposely redacted the names for some reason?)

* Many of the studies only give you an "end" measurement. As it stands, such studies cannot be analysed. Again, this is not my field: how do you plan to deal with such studies?

* To answer your original question, yes, that form of command would basically work. In particular, you could add " , by(species) " if you were interested in comparing subgroup results by species. However, you should first consider the data in depth: is it comparable? what assumptions could be made? what model to use? E.g. I can see there is a large amount of heterogeneity; in particular, some of the "mice" differences are enormous compared to the "human" differences. Are you sure the units are the same for every study?

* I would ask colleagues and/or do a literature search to see how other researchers in your field have tackled this type of problem. E.g. have any similar meta-analyses (or, more generally, reviews of the published data) been done, with similar-looking data?

Best wishes,

David.

[Attila Nagy]

Hi David!

Thanks a lot for your help!

* Also, does each line represent a different study? Yes.
unless you have purposely redacted the names for some reason? no, I can provide pubmed id etc, so these are aggregated data from pubmed, each line different study

* Many of the studies only give you an "end" measurement. As it stands, such studies cannot be analysed. Again, this is not my field: how do you plan to deal with such studies? We would deal only with those studies in which n1 mean1 sd1 n2 mean2 sd2 are presented


* To answer your original question, yes, that form of command would basically work. In particular, you could add " , by(species) " if you were interested in comparing subgroup results by species. However, you should first consider the data in depth: is it comparable? what assumptions could be made? what model to use? E.g. I can see there is a large amount of heterogeneity; in particular, some of the "mice" differences are enormous compared to the "human" differences. Are you sure the units are the same for every study? Yes units are the same, and the research question would be: Which species is closest to human (related to glucose change (end-start))? (by oral antidiabetic drugs)

Thanks,
Attila