Talk:Main Page/Archive 1

Logos

I kinda liked the embedded logos... --Timmay

Me too, but I asked Manny to take 'em down until we have explicit permission to use them. Copyrights and all that. --Dave Iberri 16:33, 31 May 2012 (PDT)

My bad. I emailed them, and I will get back to you soon. --mlam 21:06, 31 May 2012 (PDT)

Usable Articles

Can we change the default view of usable articles to the "by category and disease" and get rid of the alphabetical order version as it is (or, alternatively, bury it within the "by category and disease" view)? Tbplante 02:57, 5 June 2012 (PDT)

Done. --Dave Iberri 13:22, 27 June 2012 (PDT)

Listing by specialty

I love the new change with listing by specialty. Kudos! Tbplante 13:42, 18 June 2012 (PDT)

Glad you like it. --Dave Iberri 13:22, 27 June 2012 (PDT)

View by level of training

Can we list articles by level of training (students, interns, residents)? Tbplante 07:40, 21 September 2012 (PDT)

Baseline characteristics

For baseline characteristics, why don't we just average the numbers across the two arms. If the groups are similar, we should average the numbers. When I look at this table to apply to my patient, i want to know what the average values were across the trial, and how it applies to my patient.

Also I left some endpoints or clinical variables out because I felt like it could get cumbersome to include all, especially if the trial was not significantly powered to test certain secondary variables anyways. Therefore, I would vote for being more concise and like cliffnotes, rather than super comprehensive. It's ok to leave things out. If you want more information, I would encourage to read the article. We should emphasize the main points--"what do i need to know?" Otherwise, these summaries can become less helpful? Just my perspective.  :) --mlam 19:12, 9 October 2012 (PDT)

(Moved discussion here from Talk:ALLHAT since it applies to all articles.) I totally agree with the idea of cutting out irrelevant baseline characteristics. I disagree with averaging the characteristics of the two groups though, mostly because 1) averaging the groups should generally yield the same values as either of the study groups (by the definition of randomization) and 2) it'd be tons of work to average the values correctly, with nothing to gain. --Dave Iberri 08:38, 10 October 2012 (PDT)

For the baseline characteristics, I like just populating with data from one arm rather than averaging the groups for the burdensome of the calculations and lack of difference between the groups (usually). I think it is good habit to have baseline characteristics as a default article component -- we can leave it to the group to decide which characteristics don't matter. As for what to include from trials, I tend to want to bring all endpoints into the summary, regardless of how silly they may be (you guys have probably noticed this in my reviewing process). I like the rigor of our collaborations of the week, those reviews are the best we have produced. There isn't an easy way to go back and add to ("modernize") our original articles. Am intending on going back and reviewing them by dropping pieces to add in the discussion tab -- is there a better way to do this? Only 10 more to go until 100!! -- Tim Plante 10:33, 10 October 2012 (PDT)

Agreed. Regarding modernization, I think it's fine to just make changes/additions to the articles directly, rather than posting them to the article discussion pages. --Dave Iberri 09:24, 11 October 2012 (PDT)

Yes to cutting out "minor" baseline characteristics and avoiding clutter. Again, my preference is to average the characteristics of the two groups. If you guys are not strongly against it, then I am much in favor of it, and I can do the work of averaging the values if its cumbersome. Because of the definition of the randomization, it should generally be the same, which means that the mean is the average across both arms. It is the generally the standard across all the journals that when they speak about a baseline characteristic in the text, they take the mean of both arms, and not the mean of one arm. For example in NSABP P-1, under baseline characteristics, the text says "39.3% were 35-49...30.7% were 50-59...30.0% were 60 or older...Almost all were white (96.4%), more than one-third (37.1%) had a hysterectomy." I can give countless other examples. I understand the argument that it is not significantly different; but I think that it looks cleaner in the baseline characteristics section when we do not have to explain in an extra line saying that these values below are from one treatment arm... vs another. And why not state the baseline characteristics for BOTH arms instead of one?

--mlam 22:24, 10 November 2012 (PST)

As I've said before, I'm opposed to averaging because 1) the process is cumbersome, 2) Manny, you won't be able to keep up with all the trials so someone else will have to do this as well to keep things consistent, 3) there is little to gain precisely because the groups are by definition similar. --Dave Iberri 08:12, 11 November 2012 (PST)

Number needed to treat

I agree with Timmy that it would be great to summarize the articles to include this--or calculate it if it's not stated. I also think that we should use it in moderation (i.e. for primary outcome +/- major secondary outcomes/adverse effects only) and avoid cluttering the outcomes section. However, before we apply it for all the trials, what do you guys think is the best format to present this?

All-cause mortality (annual)

12.8% vs. 19.7% (RR 0.65; 95% CI 0.52-0.81; P=0.00013; NNT=14.5 patients for 1 year)

All-cause mortality (annual)

12.8% vs. 19.7% (RR 0.65; 95% CI 0.52-0.81; P=0.00013; NNT=14.5)

All-cause mortality (annual)

12.8% vs. 19.7% (RR 0.65; 95% CI 0.52-0.81; P=0.00013)

The NNT is 14.5 patients for one year.

Or should we mention it in the Major Points also or instead? Thoughts? My preference is the third entry above, and possibly mentioning it in the Major Points as well.

--mlam

I totally agree with putting in NNT and NNH. I think we could place it after the P-value in the parenthetical statistics note, as you've done above. It should be simple; just NNT=3 or whatever. Two thingss: 1) no decimals and 2) let's avoid rates like "patients for 1 year". For #1, there is no such thing as a fraction of a human, hence no decimals. For #2, the NNT is calculated based on incidence of the outcome, which is already in terms of a rate, hence no need for things like "patients for 1 year". --Dave Iberri 08:17, 11 November 2012 (PST)

I think that we should have a timeframe if it's not otherwise obvious. I think we can exclude it from the above analysis because it says "annually" in the description of the outcome but for reviews that have the follow-up buried in the "design" section, it's more useful. The NNT loses context when it isn't near a timeframe. I agree about the splitzies. --Tim Plante 13:55, 11 November 2012 (PST)

I like the simplicity of the second option as well. Good point!--Agree with not hacking humans into fractions, therefore no decimals. Agree with putting in time frame if not otherwise obvious; sometimes outcomes are not stated in the context of the median followup.

One further idea: Rather than recounting the amount of time for each outcome, why not start incorporating the length of time in the defining thinger above the outcomes...

Reported as intervention vs. placebo. All outcomes at 1 year.

Thoughts? --Tim Plante 18:39, 13 November 2012 (PST)

Love it. --Dave Iberri 23:46, 13 November 2012 (PST)

PICO Question of the day

Hi team! Any chance that we can bring these back to the front page? I liked them. --Tim Plante 07:01, 2 December 2012 (PST)

Done. Thanks for pointing that out. I'd just forgotten to add it back in. --Dave Iberri 18:36, 2 December 2012 (PST)