Meet the Other Phone. Flexible and made to last.

Meet the Other Phone.
Flexible and made to last.

Buy now

Please or to access all these features

General health

Mumsnet doesn't verify the qualifications of users. If you have medical concerns, please consult a healthcare professional.

Can the MMR or other vac ever cause autism?

334 replies

StarlightMcKenzie · 18/08/2014 22:04

www.ncbi.nlm.nih.gov/m/pubmed/25114790/

OP posts:
MrsWhiskersonTheFirst · 30/08/2014 21:11

noble, can you see from the above, that based on the data he used in his study, it is unlikely that the sample sizes are going to be 'tiny' as you have suggested? Unless you have reason to believe that he excluded large numbers of children from the study without mentioning them?

I'm just trying to understand why someone would choose to jump to the 'small sample size' conclusion. Why are you assuming that they are small? Why not assume that the sample sizes are probably fine but that you'd rather know what they are and that you think further investigation is needed anyway (as Hooker suggested himself)?

noblegiraffe · 30/08/2014 21:15

Hmm because as I already said he clearly had issues with too few children in some of his subgroups.

If he wanted to mitigate those obvious concerns, he'd have told us how many kids there were in other groups.

He didn't. Which causes this sort of mess. Ergo incompetent or dishonest. FFS he has got you scrabbling around figures from google images with your calculator doing guesswork. That is not a sign of good science, btw.

MrsWhiskersonTheFirst · 30/08/2014 21:25

noble, he had problems with females in one subgroup because there were so few of them to begin with. 124 in the entire cohort and 68 in the birth cert cohort. (Table 3) That's compared to 500/243 boys.

noblegiraffe · 30/08/2014 21:38

Look again, mrs, specifically table 4. The analysis for African American males was cut off at 31 months because of fewer than 5 in the cell up to 36 months. There were even fewer African American females, so even the cut off at 31 months was inadequate, so females weren't included at all in the upper age bracket.

There was an issue with sample size for male African American children of normal birthweight in that particular analysis.

MrsWhiskersonTheFirst · 30/08/2014 23:08

Yes, I didn't say there wasn't. You said that he had too few children in some of his subgroups and I was pointing out in one of those subgroups, it was to do with not having enough females to begin with.

Seeing as he does exclude subgroups that are too small, according to the minimum standards required by the Pearson Chi squared test, what are you saying your problem is with the samples? What would you consider to be 'too small a sample'?

noblegiraffe · 30/08/2014 23:32

Do you even understand what the problem with small sample size is? A sample isn't considered 'not small' just because you can run a Pearson Chi squared test over it. The smaller the sample size, the lower the confidence you can have in any results you get from it. Who knows what his sample sizes he had? You are scrabbling about with the numbers from an analysis ten years ago gleaned from google. Fingers crossed they are right. Are they the same if one was a case control and the other a cohort study? Does that affect things? Has anything else been changed? I don't know. And the thing is, you don't know either, because Hooker hasn't given us his numbers.

If he had hundreds of black African American males of normal birth weight, then great. If he didn't, then not so great, getting less great as the numbers fall. But speculation is pointless.

I am wondering why you have completely failed to acknowledge his shitty report-writing that has completely wasted both your and my time today.

Lweji · 31/08/2014 00:43

Every single bloody decent paper that has any statistics on it shows the corresponding sample size. It's pretty basic.
That this article failed to do it, is highly suspicious.

The journal where it was originally published is not a quality recognised one, and it seems that the reviewers also probably had conflict of interests.
And it has been recalled.

Sample size is extremely important to make sure statistical flukes are less likely. Although small sample sizes can be used and published, results should be taken with extreme caution.

MrsWhiskersonTheFirst · 31/08/2014 08:30

noble, don't be patronising. My point is that you don't know that the sample sizes are small. You only know that some were to small for Pearson's chi square and were excluded because of that. You can not extrapolate from that that all the sample sizes were too small. When you say a sample is too small, you clearly aren't referring to those that were excluded so I'm asking you what size you think is acceptable.

There is no 'scrabbling about' with figures. All the tables that I have referred to were given in Catherina's link. You did not accuse her of scrabbling about to find that figure of '22' (that was very poorly calculated by the way). I notice that she hasn't come back to comment on that.

The reason I'm even bothering to look at the figures is because I'm also interested in what kind of figures Hooker was using but instead of assuming that they must be tiny, I looked at the original data. Given that he states that he is reanalysing that data, it does make sense to look at it. Of course I agree that it would be easier if he had included it in his own paper but he didn't so if you're genuinely interested in what he found, you can go looking. My point throughout has been, you don't know, so why are you assuming the worst? In my opinion, it's because you don't want him to have found something.

For me, I think an interesting issue has been raised re burying data. This paper may not be the best but I think the issue warrants further investigation. I get the impression that you do not and think this should be buried again because you don't like any suggestion that the MMR may be responsible for autistic regression, even if it is in a small subgroup.

Lweji, so are we dismissing every paper that has ever been published in Translational Neurodegeneration because it's not a high impact one? You seem to be assuming there are conflicts of interest too.

MrsWhiskersonTheFirst · 31/08/2014 08:40

too small

OneInEight · 31/08/2014 09:07

To try and help settle the argument here are some of the numbers from the original DeStefano article: there were 45 cases vaccinated between 24 months and 36 months and 144 controls vaccinated in the same period. The problem comes when you then start dividing the groups by sex (about 80% in each group were males) and by ethnicity (about one third were African American). Doing the sums roughly in my head you are then down to a group size of 12 in some categories which is frankly too small to be reliable.

Whether published in a low impact journal or not you should read critically any scientific paper rather than believing automatically the authors views are correct. As grants and academic positions are awarded on the basis of publications in high impact journals no researcher is going to go to a low impact journal by choice. Papers that end up in this sort of journal tend to be either not very original (rehashing of someone else's work as here would fit) or poor quality (again fit's here as numbers not given).

MrsWhiskersonTheFirst · 31/08/2014 09:23

Why are you focussing on the cases between 24 and 36 months? Hooker said African American males receiving the MMR vaccine prior to 24 months of age or 36 months of age are more likely to receive an autism diagnosis.

Also, where did you get the figure that ? were AA? The figures in Table 5 suggest that over 60% of the birth vert group were AA. ( IIRC I don't have the tables in front of me right now)

noblegiraffe · 31/08/2014 09:45

I suspected the sample sizes were small because there were stated issues with sample sizes in the paper and the actual sizes weren't given which raised suspicions that they hadn't been given precisely because they would give low confidence in the results and an easy dismissal.

mrs seems to be arguing a different line - that the sample sizes are possibly fine, and that it was simply complete incompetence at writing statistical analyses that caused such a vital piece of information to be left out of the report, these rendering it useless to those who want to assess its importance

Neither options are confidence-building.

Given that there are other concerns about Hooker's analysis - that he put case-control data through a cohort study analysis, that he said he was doing one type of statistical analysis but the title of his tables say a different type of test, that he's not a statistician and you would expect a statistician to be involved in the calculations (no co-authors), and that he has a massive conflict of interest, and that the study has been withdrawn from the journal over concerns over the validity of its conclusion; now concerns about sample size are but a drop in the ocean of reasons to think that this study is not good evidence of a link between the MMR and autism.

OneInEight · 31/08/2014 09:50

Table 2 in original paper 35% of controls and 37% of cases were black. The percentage black is black / (sum of black and white) not black / white which is how you have got your 60% from table 5.

I focused on looking at the numbers in that group because it had the strongest P values in the Hooker study & wanted to see why.

Perhaps looking at Table 5 we should be more worried about maternal age and education as these seem to have larger associations than age of vaccination or ethnicity.

MrsWhiskersonTheFirst · 31/08/2014 09:58

noble, I'm sorry but you can't extrapolate that all sample sizes were 'small' (you haven't said what you think too small would be) simply because some were. In one case, this was because there were so few females in the group to begin with.

"that he put case-control data through a cohort study analysis"
Have you read somewhere that you can't do this? You've mentioned it a few times but you don't seem to be able to explain why you think it causes problems - aside from talking about different samples sizes.(Which we don't know)

As for having to have a statistician on board for every study; that just doesn't happen. Most people use software packages to do their statistical analyses. What is this massive conflict of interest?

We don't have to say it's 'good evidence'. We could simply say 'that warrants further investigation' (As Hooker as suggested) and get someone to do it well.

MrsWhiskersonTheFirst · 31/08/2014 10:03

OneInEight, do you have the original paper?

noblegiraffe · 31/08/2014 10:32

Most people use software packages to do their statistical analyses

Yes, and perhaps they are better at statistics than Hooker appears to be and can get away with it. A monkey could press a button and do a statistical analysis these days, that doesn't mean the results will be meaningful. If Hooker had had a statistician on board they could have at least fixed his more obvious mistakes.
As for problems with using case control data for a cohort study - any secondary school pupil knows that using data you have not collected yourself for your purposes can cause problems. If they collected data for a case control study, they would have various inclusion criteria, dealt with missing data in a certain way etc in order to enable their matching. They matched each autistic child with 3 who weren't, which is why you have 3 times as many controls as cases. Hooker then comes in and treats these two thousand kids who have been selected and matched as a general data set of kids, some of whom have autism.

As for his massive conflict of interest - well he would appear to have had a burning desire to put out an internet video accusing the CDC of conducting a massive racist experiment. That should be taken into account when considering his handling of the data and whether it might have any effect on the results he presented.

Beachcomber · 31/08/2014 11:26

CatherinaJTV, in answer to your question of Sat 30-Aug-14 17:58:41

"Beachcomber - do you accept Hooker's claim about the data then, which essentially says that autism is not associated with MMR in any gender, any colour, any age, except for little black boys who were vaccinated between the ages of 24 and 36 months? Do you think that is the truth now?"

I find this a really strange thing to ask. Are you gathering material for a blog post? Your question is ridiculously unscientific. Hooker does not claim to have shown that there is no association between MMR and anyone - that it what you wish to read into his paper, or try to get people like me, who are skeptical as to MMR's safety, to say, in some sort of 'gotcha'. I've been paying attention to autism science now for 10 years now and I've seen a lot of epidemiological studies come and go. I've had a lot of people try to pull a 'gotcha' on me too.

There is a lot I want to say in answer to you, I'll try to be brief. First of all, Hooker doesn't say anything about having found solid evidence that an autism/MMR association does not exist for any specific population. How could he possibly make such a claim in such a small epidemiological study? And how could he prove a negative? It is a very odd twisting indeed that one would have to make of what he says in order to come up with what you have above CatherinaJTV (which I guess is why you use the loaded and unscientific word 'essentially'). I thought you knew how to read scientific papers Confused

What Hooker says is this;

The results show a strong relationship between child age at the administration of the first MMR and autism incidence exclusively for African American boys which could indicate a role of the vaccine in the etiology of autism within this population group. This particular analysis was not completed in the original Destefano et al. [14] (CDC) study. Although the previous study considered MMR timing and African Americans in general, no statistically significant effect was observed. This is in contrast to our result for African Americans in general, because the CDC study limited the total African American cohort to include only those individuals who possessed a valid State of Georgia birth certificate which decreased the statistical power of their analysis. Although a statistically significant relationship between first MMR age and autism incidence was seen in the general (all races) population within the earlier Destefano et al. [14] study, the coauthors interpreted this result as an artifact of “healthcare seeking behavior” citing that autistic children would receive their vaccines earlier in order to enroll in State of Georgia early intervention programs. However, it is highly unlikely that this type of behavior would be seen exclusively in African American males and thus, alternative hypotheses must be explored, including the possibility that the MMR vaccine may be causally linked to autism in African American males.

CatherinaJTV, you appear to have picked up the word 'exclusively' and run with it. Hooker is saying that this particular data set only showed a relationship in African American males. Do you understand? He is doing what scientists are supposed to do and reporting accurately what the data he had available to him showed. The data does not allow him to rule out associations in other subgroups. I would have thought that would be obvious to anyone used to reading scientific papers. I would have thought that it is obvious that this is one, small, epidemiological study with limitations. It is the sort of study that may throw up clues that merit further examination. That's all.

As I have said upthread, people seem to be missing the point of what Hooker has done in this paper. They seem to be (willfully?) failing to understand his motivation.

Hooker did not design a study to examine the timing of MMR/autism in African American children. He did not design a study to examine timing of MMR/autism in Caucasian children, or Hispanic children, or boys or girls or anyone else.

What Hooker did was design a study to investigate if the CDC had buried data and if so to bring that data to light. He explains this very very clearly. He explains that he required the data via a FOIA request. He explains that the original study protocol was not followed and that inclusion criteria was changed after data had been collected and analyzed.

No matter how much you, or Orac or anyone one else would like - what Hooker does not do is rule out the existence of any subgroup which is vulnerable to damage from MMR vaccination. What Hooker does is show that Dr Thompson appears to be telling the truth when he says that the CDC buried data. Data which should have led to further examination and a bigger study. But of course all the MMR defenders don't want to talk about that. They don't want to talk about Thompson's admission or DeStfano's cavalier attitude to scientific method. Nope. They want to talk about Hooker's weak data; which isn't his, it is the CDC's, and which they accepted when it was being used to say what they want to hear (by Thompson and DeStefano, who now say that data was purposefully omitted).

The MMR defenders are always going on about the value of the peer reviewed study and The Science and The Scientists over the testimony and eye witness accounts of parents. Well, I hope they will have some pretty harsh words to say about CDC scientists who admit to manipulating data because it didn't fit in with their personal belief. Who omitted data about a condition that has millions of dollars being spent on it in order to try to understand it and help those who suffer from it. Scientists who buried data because they had decided before they did their study what the results should be.

Beachcomber · 31/08/2014 11:31

Forgot to say, CatherinaJTV, did you pay attention to this part?

Although a statistically significant relationship between first MMR age and autism incidence was seen in the general (all races) population within the earlier Destefano et al. [14] study, the coauthors interpreted this result as an artifact of “healthcare seeking behavior”

Beachcomber · 31/08/2014 11:36

Sorry for triple post - the above post should have been addressed to Noblegiraffe not CatherinaJTV. It may help you with your questions over the statistical significance of the numbers.

MrsWhiskersonTheFirst · 31/08/2014 11:42

"any secondary school pupil knows that using data you have not collected yourself for your purposes can cause problems"

Does that prevent scientists from analysing data from other studies? I think you'll find that it does not. Also, do you automatically reject any paper that does not have a statistician as an author? Or just the ones that suggest things that you don't want to hear?

noblegiraffe · 31/08/2014 11:47

What Hooker did was design a study

Hooker didn't design a study at all. He ran some programmes over some data that some other people had collected in a particular way for their study.

I'm not sure I understand the objections to using birth certificate data either. The original authors ran an analysis on some data and found statistically significant results related to black boys. It doesn't take much intelligence to wonder whether it is a race thing, or a third (or more) variable which is related to race. Like if you see ice cream sales are related to deaths by drowning you might then decide to dig deeper and only look at entries in your data set where ice cream sales and numbers of people going swimming are given. When you look at this subset of data, you realise that it's not ice cream sales that are important, but the numbers of people swimming, and high numbers of people swimming and high ice cream sales are linked. There is a clear reason why deaths by drowning and number of people swimming are linked, so this explains the ice cream data you saw in your original population.
So they look at a subset of data where they not only have information about race, but also about education and other factors (which came from having the full birth certificate). When they look at this data set, the results for race aren't statistically significant any more because they are tied up with other factors that relate to autism such as low birthweight, or to a diagnosis of autism, such as education. That's not wilfully dropping 41% of your data for no reason at all other than to cover up the truth.

noblegiraffe · 31/08/2014 11:53

Also, do you automatically reject any paper that does not have a statistician as an author?

No, but if the (single) author is so bad at statistics that they make mistakes that even I can spot, it would certainly make me concerned about ones I haven't spotted. Extraordinary claims require extraordinary evidence, and this one falls at the first hurdle.

MrsWhiskersonTheFirst · 31/08/2014 11:55

What statistical mistakes has he made that you have spotted?

Beachcomber · 31/08/2014 11:58

Noblegiraffe, I have already said on this thread that I think you have totally missed the point of what Hooker has done.

It's like you can't see the wood for the trees.

Read this bit of his paper carefully and stop concentrating so much on trying to paint Hooker as dishonest or incompetent and turn your focus for a few minutes on DeStefano et al and consider their honesty and or competence.

The results show a strong relationship between child age at the administration of the first MMR and autism incidence exclusively for African American boys which could indicate a role of the vaccine in the etiology of autism within this population group. This particular analysis was not completed in the original Destefano et al. [14] (CDC) study. Although the previous study considered MMR timing and African Americans in general, no statistically significant effect was observed. This is in contrast to our result for African Americans in general, because the CDC study limited the total African American cohort to include only those individuals who possessed a valid State of Georgia birth certificate which decreased the statistical power of their analysis. Although a statistically significant relationship between first MMR age and autism incidence was seen in the general (all races) population within the earlier Destefano et al. [14] study, the coauthors interpreted this result as an artifact of “healthcare seeking behavior” citing that autistic children would receive their vaccines earlier in order to enroll in State of Georgia early intervention programs. However, it is highly unlikely that this type of behavior would be seen exclusively in African American males and thus, alternative hypotheses must be explored, including the possibility that the MMR vaccine may be causally linked to autism in African American males.

noblegiraffe · 31/08/2014 12:06

mrs, for the billionth time: not giving his sample sizes. Assuming that was a mistake and not dishonesty. Out of interest, why are you so keen to defend what appears to at the least be a poorly written, poorly considered paper which has since been withdrawn?

Beachcomber, I just dealt with that in the ice cream sales post.

Swipe left for the next trending thread