1
00:00:08,641 --> 00:00:10,010
So, I want to

2
00:00:10,010 --> 00:00:13,813
end with an example of a study
that kind of highlights

3
00:00:13,813 --> 00:00:17,951
some of the points
that I've tried to lay out for you.

4
00:00:17,951 --> 00:00:20,353
And this is a study called ENRICHD.

5
00:00:20,353 --> 00:00:23,123
It was done a couple of years ago.

6
00:00:23,123 --> 00:00:24,157
It's over now.

7
00:00:24,157 --> 00:00:26,926
Called Enhancing Recovery
in Coronary Heart Disease Patients.

8
00:00:26,926 --> 00:00:30,697
And the basic premise of ENRICHD
comes from data like these.

9
00:00:31,598 --> 00:00:36,436
This is not the ENRICHD data, these are
other data from an earlier study.

10
00:00:36,436 --> 00:00:39,539
And it's a phenomenon that we know
very well.

11
00:00:39,539 --> 00:00:43,676
And the phenomenon is this,
that after somebody has a heart attack,

12
00:00:43,676 --> 00:00:47,147
if they're depressed,
they are much more likely to die

13
00:00:47,147 --> 00:00:50,583
than if they're not depressed.

14
00:00:51,451 --> 00:00:53,686
So, what you can see,

15
00:00:53,686 --> 00:00:57,223
this is a man's post-MI, post
heart attack.

16
00:00:57,223 --> 00:01:03,396
And within two months post-MI,
as you can see here, we see a divergence

17
00:01:03,396 --> 00:01:08,701
of participants who become depressed
versus those who do not become depressed.

18
00:01:08,701 --> 00:01:13,139
And you can see the mortality rates
just go up.

19
00:01:13,139 --> 00:01:17,977
It's a phenomenon
that we're -- it's a very real phenomenon.

20
00:01:18,311 --> 00:01:23,316
So, the ENRICHD trial
said, "We need to do something about this.

21
00:01:23,316 --> 00:01:27,053
Let's see what happens
if we treat the depression.

22
00:01:27,053 --> 00:01:31,458
Let's try that." So,
that's what the ENRICHD trial was.

23
00:01:31,458 --> 00:01:37,097
And they also knew that depression was
very strongly linked to social support.

24
00:01:37,097 --> 00:01:40,834
So, they actually had an intervention,
a dual intervention,

25
00:01:40,834 --> 00:01:44,170
to target low social
support and high depression.

26
00:01:45,205 --> 00:01:46,706
And so, the

27
00:01:46,706 --> 00:01:51,678
hypothesis was that they wanted to test
whether treating depression

28
00:01:51,678 --> 00:01:56,149
and low social support
right after a heart attack

29
00:01:56,149 --> 00:02:01,654
was going to reduce death rates
and non-fatal recurrent heart attacks.

30
00:02:01,654 --> 00:02:03,123
Simple study, right?

31
00:02:03,123 --> 00:02:05,391
It was not simple.

32
00:02:05,391 --> 00:02:10,096
So, ultimately, they enrolled almost 2,500
post-heart attack participants

33
00:02:10,096 --> 00:02:13,600
that either presented with depression
or low

34
00:02:13,600 --> 00:02:17,070
social support or both,
very commonly both.

35
00:02:17,570 --> 00:02:22,775
And they randomized into a group
that got usual care versus one

36
00:02:22,775 --> 00:02:27,947
that got a psychosocial intervention to
treat the social support and depression.

37
00:02:27,947 --> 00:02:32,719
And their endpoints were cardiac mortality
and a recurrent heart attack.

38
00:02:32,719 --> 00:02:37,490
And they followed them for almost three
and a half years.

39
00:02:37,490 --> 00:02:42,228
And they were totally masked in terms
of their primary outcomes

40
00:02:42,228 --> 00:02:45,698
and they did an intent to treat analysis.

41
00:02:45,732 --> 00:02:51,070
So, here is what I want you -- no,
this is not what I want you to see.

42
00:02:51,070 --> 00:02:52,539
Their inclusionary criteria were that

43
00:02:52,539 --> 00:02:56,709
they were going to enroll people
within 28 days of having a heart attack.

44
00:02:56,709 --> 00:02:58,778
Now, they wanted to enroll them quicker.

45
00:02:58,778 --> 00:03:01,447
They wanted to get them
more quickly than that.

46
00:03:01,447 --> 00:03:05,285
Because you saw on that graph that
the divergence happened at two months,

47
00:03:05,285 --> 00:03:09,656
so they wanted to enroll them
as quickly as possible but they knew that

48
00:03:09,656 --> 00:03:13,726
it would not be feasible to enroll them
much more quickly than this.

49
00:03:13,726 --> 00:03:14,661
Why is that?

50
00:03:14,661 --> 00:03:18,097
Well, you have a patient
who just had a heart attack,

51
00:03:18,097 --> 00:03:20,600
they're in a hospital,
there are families involved,

52
00:03:20,600 --> 00:03:23,870
they have other things to think about
besides your study.

53
00:03:23,870 --> 00:03:28,408
So, they settled on recruiting them
within about a month after a heart attack,

54
00:03:28,741 --> 00:03:32,445
and they had a bunch of criteria

55
00:03:32,445 --> 00:03:37,183
-- they had some inclusionary criteria
that could substantiate

56
00:03:37,183 --> 00:03:41,387
the fact that they had a heart attack.

57
00:03:41,387 --> 00:03:44,023
They looked at enzymatic increases.

58
00:03:44,023 --> 00:03:48,228
And then they identified
whether they were depressed

59
00:03:48,228 --> 00:03:55,335
and/or had low social support
and there are good measures to do that.

60
00:03:55,335 --> 00:03:59,305
Here's what I want to show you.

61
00:03:59,305 --> 00:04:01,941
This is the CONSORT diagram.

62
00:04:01,941 --> 00:04:06,145
They screened almost 34,000
people to get 2,481.

63
00:04:06,145 --> 00:04:08,781
So, they screened this number.

64
00:04:08,881 --> 00:04:12,218
Some didn't meet criteria for having had

65
00:04:12,218 --> 00:04:15,989
a heart attack, right, but not that many.

66
00:04:15,989 --> 00:04:21,661
Well, assume they had a heart attack,
so they met that criteria.

67
00:04:21,661 --> 00:04:26,866
And then some of them, fair number,
7,000 of them almost,

68
00:04:26,866 --> 00:04:32,538
weren't depressed or had social support,
so they didn't meet those criteria.

69
00:04:32,538 --> 00:04:37,277
But look at this, 23,000
didn't meet their medical eligibility.

70
00:04:37,277 --> 00:04:42,015
Given what we've just talked
about, what does that mean?

71
00:04:50,256 --> 00:04:50,990
Did they

72
00:04:50,990 --> 00:04:54,594
do something wrong with their inclusionary
or their exclusionary criteria?

73
00:04:54,594 --> 00:04:56,763
Yeah. You can't say it's wrong.

74
00:04:56,763 --> 00:05:00,033
I mean, these are some of the best
researchers

75
00:05:00,033 --> 00:05:02,201
around who were doing this study.

76
00:05:02,201 --> 00:05:03,269
It's not wrong.

77
00:05:03,269 --> 00:05:05,438
But ultimately, when you step back

78
00:05:05,438 --> 00:05:09,442
and look at it,
this is where they had the problem.

79
00:05:09,442 --> 00:05:11,244
This is where it happened.

80
00:05:11,244 --> 00:05:16,649
So, it wasn't that it was wrong,
but they had to screen so many people

81
00:05:16,649 --> 00:05:21,354
in order to get, you know, 2,500 people
because of these eligibility criteria.

82
00:05:21,788 --> 00:05:25,091
Scientifically,
it was the right thing to do,

83
00:05:25,091 --> 00:05:28,594
but it made for a very difficult study.

84
00:05:28,594 --> 00:05:33,299
And to operationalize
this a little bit more clearly for you,

85
00:05:33,299 --> 00:05:38,237
what this means is that for every 100
participants that they screened

86
00:05:38,237 --> 00:05:42,342
-- these weren't people that called up
and said, "I

87
00:05:42,342 --> 00:05:46,479
want to participate." These are 100
participants that they screened,

88
00:05:46,479 --> 00:05:50,583
only 7 were actually enrolled
and randomized into the trial.

89
00:05:50,583 --> 00:05:52,652
That's a lot of work.

90
00:05:53,619 --> 00:05:55,588
Another way of thinking about

91
00:05:55,588 --> 00:06:00,093
it, in order to get 1,
they had to screen 14.

92
00:06:00,093 --> 00:06:04,964
And of course, they had a lot of sites,
and not all

93
00:06:04,964 --> 00:06:08,501
the sites were really able
to adequately enroll participants.

94
00:06:08,501 --> 00:06:12,405
And that's something
that when you have a multi-site trial

95
00:06:12,405 --> 00:06:15,908
you will find that sites recruit
at different rates.

96
00:06:15,908 --> 00:06:20,213
There may be reasons to keep sites
that have low recruitment

97
00:06:20,213 --> 00:06:22,448
because they're recruiting,
you know, particular

98
00:06:22,682 --> 00:06:26,085
characteristics of the patient population
that you're interested in.

99
00:06:26,352 --> 00:06:30,189
But it is a phenomenon
that occurs fairly frequently.

100
00:06:30,189 --> 00:06:35,261
So, you know, there's been a lot of talk
among the investigators

101
00:06:35,261 --> 00:06:39,699
about why there were these differences,
why they have these problems

102
00:06:39,699 --> 00:06:41,334
with recruitment and retention.

103
00:06:41,334 --> 00:06:44,971
And some of them had to do with access.

104
00:06:44,971 --> 00:06:50,243
Some of them had to do with the fact
that they were recruiting

105
00:06:50,243 --> 00:06:55,681
at very research-heavy sites,
so there was a lot of research going on.

106
00:06:55,681 --> 00:07:00,787
And you have to think about that,
is there a lot of stuff

107
00:07:00,787 --> 00:07:05,124
going on in this area
where you're going to be recruiting?

108
00:07:05,124 --> 00:07:07,894
They didn't have that
quite collaborative teams,

109
00:07:07,894 --> 00:07:11,063
and so that reduced their ability
to recruit.

110
00:07:11,063 --> 00:07:15,768
And their burden was high,
so these participants had to come in

111
00:07:15,768 --> 00:07:18,538
and participate
in a psychosocial intervention, which

112
00:07:18,538 --> 00:07:23,276
always is a little bit more burdensome
than taking a pharmacologic pill.

113
00:07:24,644 --> 00:07:27,513
And they had a lot of assessments

114
00:07:27,513 --> 00:07:32,418
and the study went on for three
and a half years, right?

115
00:07:32,418 --> 00:07:35,254
So, this was a very burdensome study.

116
00:07:35,254 --> 00:07:40,593
They -- but the real key is here,
they had really restrictive eligibility

117
00:07:40,593 --> 00:07:43,863
criteria,
as you can see from the exclusion,

118
00:07:43,863 --> 00:07:47,333
the high number of patients
that were excluded.

119
00:07:47,333 --> 00:07:51,604
So, let me walk through
a couple of overall conclusions.

120
00:07:52,004 --> 00:07:53,239
Know your literature.

121
00:07:53,239 --> 00:07:58,144
Know the history of your literature,
know where things sit, be able

122
00:07:58,144 --> 00:07:59,779
to identify contextual factors,

123
00:07:59,779 --> 00:08:03,483
know where your question lies
on that research continuum.

124
00:08:03,483 --> 00:08:04,717
That's really critical.

125
00:08:04,717 --> 00:08:10,456
Make sure you're designing the right study
for the right time for that literature.

126
00:08:10,456 --> 00:08:13,726
Make sure you have clarity
on your question.

127
00:08:13,726 --> 00:08:19,866
And this sounds kind of silly, like, "Of
course I have clarity on my questions.

128
00:08:19,866 --> 00:08:24,170
I want to know if this does this."
But understanding the complexity

129
00:08:24,170 --> 00:08:28,107
of your question
really means that you have to dive deep

130
00:08:28,107 --> 00:08:32,745
into some of the issues that we're talking
about and more, as well,

131
00:08:32,745 --> 00:08:37,750
to understand what your -- you know,
all of the complexities of your question.

132
00:08:37,750 --> 00:08:40,586
Have a really good,
very specific targeted understanding

133
00:08:40,586 --> 00:08:43,823
of the patient characteristics
that you want to target.

134
00:08:44,390 --> 00:08:48,528
Make sure the participants
match the outcomes, your primary outcomes,

135
00:08:48,528 --> 00:08:52,698
make sure it's a good match,
it's a feasible match.

136
00:08:52,698 --> 00:08:58,304
And make sure that you're asking
the right question at the right time.

137
00:08:58,304 --> 00:09:02,241
Going back to internal versus
external validity, figure out

138
00:09:02,241 --> 00:09:07,647
in the context of your question,
what's most important; or in other words,

139
00:09:07,647 --> 00:09:13,152
think about what you worry about the most,
false negatives or false positives?