O'Dowd-Crawley early intervention study

Number of parents in paid work?

 

Doesn't appear to be. The protocol lists 4 primary outcomes:

1. Feasibility (no data but conclusion is no)
2. Suitability and acceptability of health economic measures (no data)
3. Qualitative information about patients' views (no data)
4. Information on the # of patients engaged and retained (?)

The spreadsheet only contains the secondary outcomes of the usual set of psychometric questionnaires. Aim was for 100 participants, only provided incomplete data for 44. It's also a comparison with "standard care" but we have no idea what that was, as usual not controlled for.

The "intervention" being a hodge-podge of the standard behavioral approach seems to indicate it serves no benefits to whatever this patient population is:

None of the data actually relate to any of this so basically the "intervention" and what is measured have about nothing to do with one another.

Looking at the protocol and noticed this interesting bit of... uh... "special" logic:

Another framing of the primary outcomes:

There is no data on adherence of follow-up rates beyond missing data but there are no details about why data are missing. There is supposed to be a 6 months qualitative study including GP interviews, no data on this.

Unclear whether the interviews even happened and, if not, why they did not happen.

There are various uses of generic fatigue, chronic fatigue and CFS. The data key defines this trial as being a CFS cohort:

But the criteria are clearly not a valid CFS cohort:

The health economics analysis seems to have been dropped. Seems similar in thinking to the one done for PACE.

I think most of the relevant details would be in the TMG notes, just like with PACE:

 

My guess would be part-time and full-time. Reference only says "Number in paid work at [baseline/6 months]".

 

One notable thing here is looking at the description, the intervention is basically the current behavioral paradigm, meaning it found the current paradigm to not be feasible. It's also very similar to the PACE model, which is a bit awkward.

Good reason not to publish. Not sure what they did different, however, to get people to tell them off. Maybe too loose of a cohort, not disabled enough to swallow the BS? Then again it seems that in most of those studies they carefully select their participants to maximize the outcomes they are seeking.

 

The raw results of scientific research of this nature is rarely clean, simply because participants can just answer whatever they want. All you can hope for is useful numbers in the aggregate...

 

Oh, I forgot this is on adults! That would make sense.

Are there really no zeros (for unemployed)?

 

Guiding source of AfMEs pacing leaflet and other info/ literature.

 

I tried to find references from the protocol and the only thing that could fit looking at the WPAI is the question of whether you are currently employed. It's the only binary question so I guess the options 1-2 stand for 0-1. All the 2's at baseline have empty data for # of hours missed from work so it looks like 1=employed and 2=unemployed.

http://www.reillyassociates.net/WPAI_GH.html

 

Ah, so it's not part-time/full-time. It's unemployed/employed. Which is ridiculous, because someone doing 5 hours a week will be classed the same as someone doing 40.

 

Sorry - had no phone/internet for the past few days - fallen tree took out our line.

I've had a quick look. Better this time, but still missing WPAI qs 3 and 4. They may say that they only gave 6-month results for qs 2 and 5 in table 3, but they have provided data for qs 1 & 6 (and bl summary data for qs 3&4), which weren't reported, so I don't think that's a valid excuse (if they give it as one).

In particular, it is useful to know the answers to q4 (How many hours did you actually work in the past 7 days?), because at 6 months it would give an indication of whether pts were reducing their hours to better manage their condition - which wouldn't necessarily show up in q2 (Hours missed).

I presume that the answers to q1 (Are you in paid work?) are 1=Yes, 2=No - which is the opposite way round to some versions of the form.

The txt file explaining the data fields is very poor. It really should give min and max possible values for each variable (where relevant) - but hey, can't have everything, and most of them are guessable.

Analysis-wise - the main thing would be to confirm the data they have already published, which should be fairly straightforward.

eta: I can confirm that data does seem to match the summary data they provided. However, important caveat is that they have summarised baseline medians (IQRs) and then mean (95% CI) outcomes, which given the data characteristics, I would take with a hefty pinch of salt (makes it look much worse than it is): outcome data for hours missed (WPAI Q2) is based on a few pts with large values - medians would have been more representative.

 

The individual data would be preferable to just the results, if poss (don't mention "medians" unless you just want the summary data).

Q3 should be: "During the past seven days, how many hours did you miss from work because of any other reason, such as vacation, holidays, time off to participate in this study?"
Q4 should be: "During the past 7 days, how many hours did you actually work?"

So, best to ask for "Hours missed from work due to other reasons in the past seven days" and "Hours worked in the past seven days" to be safe. [eta]

It's been a fun few days! Several neighbours had their phones and routers fried by a lightning strike a few hours before the tree fell, so Openreach had their work cut out!

 

Slight edit to my post above to clarify (see post itself ^^^^).

It would be useful to have the reasons why a third of the intervention group dropped out ("Did not receive allocated intervention"), but I guess that's probably TMI for us mere mortals.

 

IQRs is usually a sign of heavily skewed data. medians would indeed be helpful, or indeed some way of visualising the distribution of results.

Still disappointing that the data set is still missing some of the most interesting questions.

 

Which happen to be, as usual, the primary outcomes. Which makes it at least the 5th or so time this happened with a PACE researcher? And still they continue to get funding where they submit primary outcomes and a methodology, essentially a contract, and bury them when they can't massage the data into something mildly convincing for the usual outcome they have pre-determined for 3 decades by now.

Because there was a conclusion, which is that PACE-style trials (and the current guidelines) are not feasible, based on non-acceptance by the participants, which was the primary outcome. Surely if they concluded it is not feasible they did do their interviews and ask those primary outcome questions, which were described in protocol as being produced as anonymized data.

So where are the data, Lebowski? Because they sure are somewhere, paid for by public money, having found that the current paradigm is so absurdly bad people who aren't desperate enough, whoever this cohort may be, find it useless. Or maybe? Who knows what they said? Those are just the primary outcomes submitted as part of a publicly-funded feasibility study so it's not as if there was any way of knowing what people think of the current, highly controversial, "medical" paradigm, which somehow does not include much that is actually relevant to medicine and rather is closer to The Secret and life coaching fluff where alternative medicine practicioners also report that their clients feel better but somehow this is different because reasons.

 

Here's the WPAI Q2 data:

WPAI Q2: Hours missed due to illness in past 7 days
Controls
Baseline: 5 of 13 missed any hours (3, 4, 12, 16, 19). [median 0, IQR 0-4]
6 months: 3 of 11 missed any hours (2, 9, 9). [median 0, IQR 0-0]

EI group
Baseline: 7 of 18 missed hours (0.5, 1*, 2, 4, 8*, 10*, 44). [*subsequently dropped out] [median 0, IQR 0-2]
6 months: 4 of 15 missed hours (2, 5, 16, 56). [median 0, IQR 0-0]

Looking for a difference on the basis of means at 6 months is extremely unsafe, and would give a false impression that the EI group have done statistically worse, when there isn't enough information to say that.

However, Q4 would give useful info on the hours actually worked, and whether pts had reduced their hours during the 6 months to compensate.
The outlier (44 and 56 hours in one pt) needs investigation (typo?), particularly as the EQ and SF36 data for this pt were low.

tbh, with data like these, I'm not surprised they stopped the study. I suspect there will be something in qualitative data that confirms that. But if there was any indication that early intervention might pose harm to pts, they really should be writing it up. Unfortunately, there is probably not enough data from controls to enable a proper comparison.