Association of adverse childhood experiences with the development of multiple sclerosis, 2022, Eid et al

[Hutan]

https://jnnp.bmj.com/content/early/2022/03/08/jnnp-2021-328700

Abstract
Objective
To study whether exposure to childhood emotional, sexual or physical abuse is associated with subsequent multiple sclerosis (MS) development.

Methods
A nationwide, prospective cohort study based on participants in the Norwegian Mother, Father and Child cohort study. Enrolment took place 1999–2008, with follow-up until 31 December 2018. Childhood abuse before age 18 years was obtained from self-completed questionnaires. We identified MS diagnoses through data-linkage with national health registries and hospital records. The Cox model was used to estimate HRs for MS with 95% CIs, adjusting for confounders and mediators.

Results
In this prospective cohort study, 14 477 women were exposed to childhood abuse and 63 520 were unexposed. 300 women developed MS during the follow-up period. 71 of these (24%) reported a history of childhood abuse, compared with 14 406 of 77 697 (19%) women that did not develop MS. Sexual abuse (HR 1.65, 95% CI 1.13 to 2.39) and emotional abuse (HR 1.40, 95% CI 1.03 to 1.90) in childhood were both associated with an increased risk of developing MS. The HR of MS after exposure to physical abuse was 1.31 (95% CI 0.83 to 2.06). The risk of MS was further increased if exposed to two (HR 1.66, 95% CI 1.04 to 2.67) or all three abuse categories (HR 1.93, 95% CI 1.02 to 3.67).

Interpretation
Childhood sexual and emotional abuse were associated with an increased risk of developing MS. The risk was higher when exposed to several abuse categories, indicating a dose–response relationship. Further studies are needed to identify underlying mechanisms.

Open Access - data can be obtained by the Norwegian Institute of Public Health

 

[Kalliope]

This has gotten much more attention in Norwegian media than the study proving EBV causes MS.

 

[NelliePledge]

This has gotten much more attention in Norwegian media than the study proving EBV causes MS.

 

[Hutan]

The reported percentage of the women in the sample who reported childhood abuse developing MS was 0.49%, compared to 0.36% in the women who didn't report childhood abuse.

So, if the 0.36% incidence in the women who didn't report childhood abuse had applied to the 14477 women who reported childhood abuse, we would have seen 52 women developing MS versus the 70 who actually did. This could be just a chance result.


There was some researcher discretion around who was defined as having MS - bias may have affected the approach.

If the woman was registered in NPR with an MS diagnosis, but not in the MSR, we used hospital records to further validate the diagnosis using the 2017 diagnostic criteria for MS.16 We were able to refute incorrect MS diagnoses from the NPR based on the information from the hospital records. NPR-identified MS cases for whom we did not have access to the hospital records for validation, were excluded.

There is the issue of confounders - factors that might affect the incidence of MS that are probably associated with childhood abuse. It isn't clear just from the text how they dealt with confounders that they had data for e.g.

The women exposed to childhood abuse more often had a history of smoking, were overweight and had more depression at study baseline.

And they note that there are other potential factors that they did not have data on:

As in all observational studies, residual confounding may be another limitation. We had detailed information on behavioural risk factors in adulthood such as smoking and obesity, but childhood abuse may be associated with other environmental factors such as diet, nutrition, physical exercise, and parental smoking, which could be independent risk factors for MS.

I think this is a major problem. For example, a family that has a lot of financial stress, perhaps unemployed parents, is more likely to have incidents of childhood abuse and is less likely to be able to take annual winter holidays somewhere sunny. Therefore, childhood vitamin D levels of the women reporting childhood abuse are probably highly correlated.

 

[rvallee]

All other things: not equal.

If people want to find something vague, they'll find it.

The big question is why do some people so desperately want this ideology to be true? What really drives this mindless obsession? This is basically a repeat of the bad mother trope of autism.

 

[Hutan]

Yes. It's conceivable that traumatic experiences in childhood do somehow make a person more susceptible to later developing MS. But, there are other more likely explanations for an association between childhood abuse and MS, with poverty likely to be the common factor.

Further on the possibility of the association just being the result of chance
Time trends in the incidence and prevalence of multiple sclerosis in Norway during eight decades
In that study, the prevalence of multiple sclerosis in 2013 in Norway is reported by county, with hundreds of thousands of people being in each county. Despite the paper concluding that latitude is not a factor, the variation in prevalence by county is very large, ranging from 142 to 275 per 100,000 people.

While the authors do mostly only claim an association, they do say

In conclusion, children exposed to adverse experiences had an increased risk of developing MS later in life, independent of known environmental risk factors for MS.

They therefore do slide into extrapolating their finding beyond their sample to people in general, and turning an association into a cause. I suspect they are not working hard to prevent media reporting the association as actually causal.

The big question is why do some people so desperately want this ideology to be true? What really drives this mindless obsession? This is basically a repeat of the bad mother trope of autism.

Yes. I think it's likely to be 'terror management'. That is, I can believe that I and my children are safe from MS, because we have a nice, morally upright family background. Never mind that numerically far more people without a history of childhood abuse experience develop MS than those with a history of childhood abuse.

I also think that the idea is used to bolster the idea that the mind is powerful and can cause disease. And so, if an emotional response to trauma made physical can cause disease, the mind can be trained to cure the disease. The finding provides support for the 'CBT as cure' idea for all sorts of diseases, and strengthens the idea that people are personally responsible for fixing their illness. A potential cure is of course attractive to those with a disease.

What these authors seem to fail to recognise is the harm that this focus on childhood abuse can cause. On the basis of this questionable association, people with MS may blame their family for causing their disease, and may perhaps even reframe fairly normal family life as abusive. This may fracture families at a time when the person with MS most needs support. People with MS may waste time and energy trying to fix their mind and personality instead of doing something more useful or enjoyable. Health systems may waste money providing therapies that do nothing useful.

Importantly, more likely associations and biological hypotheses may not get the research attention they deserve while money is being spent following the early childhood trauma idea.

 

[Hutan]

Exposure to either emotional, sexual, or physical abuse was defined as responding ‘yes, as a child <18 years’ to the respective category. We considered women who answered ‘no, never’ to the abuse items as non-exposed.

One last observation. It looks as though these pregnant women were asked if they had ever suffered from the three kinds of abuse, so as a child, but also as an adult. The findings relating to abuse as an adult are not reported in this study. Perhaps they will be reported in another paper. Or perhaps there was no finding that supported the idea of later trauma causing MS. If it's the latter, then what we are seeing in this paper has been to some extent cherry picked, and the analysis should probably have taken into account the multiple chances of finding an association.

 

[hibiscuswahine]

I don’t find the association strong or very interesting and sad this is then publicised in the manner described. Of their mother and baby cohort only 41% participated. I am afraid I do not know the incidence of MS in women who have not had children in Norway , this would add to the lack of predictive value of it. So when I read this it doesn’t add much to my knowledge. Abuse of any kind is common through all socioeconomic groups.

They do say it is an association only, they are considering the many factors that might underlie brain inflammation and autoimmunity. It is a hot-topic is Neurology and Psychiatry, lots of debate about whether for instance depression and schizophrenia is an immune disorder but they have been debating this for decades with little to show for it. Scientific research comes up with thousands of associations. Unfortunately they can be picked up and then have destructive consequences as Hutan has mentioned.

 

[Snow Leopard]

The fact that almost 20% of women experienced childhood abuse is more alarming than the very small MS association.

 

[Hutan]

The supplementary materials has this about the adjustments. No further detail is given.

Hazard ratios adjusted for adverse socioeconomic factors (≤ 9 years elementary school, single mother or low household income), smoking (ever vs. never) and BMI≥ 25 before study baseline). Birth year was included as a stratification factor in the Cox model. No other covariable violated the proportional hazard assumption.

It's worth noting that these adjustments relate to the woman at the time of the study, when she was pregnant. The fact that the woman is a single mother or has a low household income now is not a perfect measure of the conditions the woman experienced when a child.

It would be very interesting to get the same data, and see if a reasonable approach to analysis can produce a different result.


Edit to add: Here's the paragraph in the text about what was done. The adjustments around birth year look suspicious. If child maltreatment is a risk factor for MS, why should it matter how common the child maltreatment is at a particular time?

The models were stratified by the women’s birth year in groups and adjusted in a two-step approach for (1) possible confounders and (2) possible confounders and mediators.

We considered birth year and childhood social status19 as possible confounders and used early drop-out from school as a proxy for the latter. Birth year was taken into account as the incidence of child maltreatment probably has decreased during the last decades prior to inclusion in MoBa. 20. Possible mediators were smoking, high BMI, and adverse socioeconomic status as an adult—factors associated with both childhood abuse4 21 and MS.11 22–24

 

[Hutan]

I think this is worth looking at.

(Just as background - This is from a Cochrane resource
https://s4be.cochrane.org/blog/2016/04/05/tutorial-hazard-ratios/

Confidence Interval (CI): is the range of values that is likely to include the true population value and is used to measure the precision of the study’s estimate (in this case, the precision of the Hazard Ratio). The narrower the confidence interval, the more precise the estimate. (Precision will be affected by the study’s sample size). If the confidence interval includes 1, then the hazard ratio is not significant.

The authors of this MS paper acknowledge this, noting that hazard ratios for confidence intervals that include 1 are regarded as not significant.)

Here are the results reported in the paper:



Look at the right hand column where they included some limited adjustments for confounding. Even with the birth year manipulation, the only trauma type that does not have a 1 in the confidence interval is sexual abuse. Neither Humiliation nor Threat are significant, and it is only when the two categories are amalgamated that the new trauma type 'Emotional Abuse' sneaks into significance with a confidence interval lower bound of 1.03.

Importantly, when 'Any childhood abuse' is considered, the confidence interval includes 1. By their own reporting, 'Any childhood abuse' does not produce a significant hazard ratio. 'Any childhood abuse' doesn't affect the incidence of MS. Also, have a look at the range of the confidence intervals, specifically the one for Sexual Abuse. The range is very large, indicating very poor precision. In other words, the data are all over the place, there is no consistent impact of childhood trauma on MS incidence.

But, the impact of Sexual Abuse does seem significant...?

I refer you to Supplementary Table 2.



I suspect that the authors actually did this analysis first. It presumably includes 97 women who had MS at baseline (306+91-300). The paper talks about a potential reporting bias in these women - if the women knew the theory, then they might have tended to over-report childhood abuse when that information was collected. It turns out that they did not, overall. But, nevertheless, the authors decided to remove these 97 women from the analysis that they reported in the paper itself. The mean after adjustments increases from 1.46 to 1.65, and the confidence interval changes from (1.04 to 2.06) to (1.13 to 2.39). That is, a hazard ratio that is just squeaking over the line appears to be much more solid.

What I can't figure out is that Supplementary Table 2 suggests that 39 women out of 306 women with MS (12.7%) reported sexual abuse, and the paper's Table 2 suggests that only 34 women out of 300 women with MS (11.3%) reported sexual abuse. So, the frequency of sexual abuse is higher in the Supplementary Table 2 sample, but the mean hazard ratio of sexual abuse is a lot lower (1.46 instead of 1.65). What's going on? Did the adjustments for confounders and the approach to mucking around with birth date stratification change? [Edit - I initially thought the total women in the Supplementary analysis was 306, not 306+91]

In summary, I suggest that this study has data that is all over the place as evidenced by the non-significant or barely significant, and very wide, confidence intervals of the hazard ratios. The data had to be tortured into providing the story that the authors wanted, with various poorly documented and poorly justified adjustments. Even with those manipulations, the story is far from convincing if you just look at the numbers.

 

[Mithriel]

I have said before that sexual abuse is not a specific abuse, it can range from name calling to serious physical attack with suffocation. In fact all abuse has a wide range of seriousness.

When talking about preventing abuse in children it is valid to lump it all together but medical consequences will be different in each case.

Excessive stress in childhood may cause medical problems, malnutrition or neglected care may cause developmental problems and physical injury may have a lasting effect. If these things are what they mean they should say so instead of childhood abuse which is a meaningless hangover from Freudian psychotherapy.

 

[Jonathan Edwards]

In summary, I suggest that this study has data that is all over the place as evidenced by the non-significant or barely significant, and very wide, confidence intervals of the hazard ratios. The data had to be tortured into providing the story that the authors wanted, with various poorly documented and poorly justified adjustments. Even with those manipulations, the story is far from convincing if you just look at the numbers.

I think it is interesting to see this sort of attempt to invoke trauma for a condition for which trauma as a cause is inherently very implausible. It casts a light on similar studies in ME. The studies are probably not directly comparable but it shows that people will try very hard to get the result they are expecting.

 

[Snow Leopard]

I have said before that sexual abuse is not a specific abuse, it can range from name calling to serious physical attack with suffocation. In fact all abuse has a wide range of seriousness.

Name calling is considered sexual harassment, rather than abuse in the context of studies like this.

 

[Arnie Pye]

To study whether exposure to childhood emotional, sexual or physical abuse is associated with subsequent multiple sclerosis (MS) development.

Abuse of any kind is very likely to raise cortisol levels. Perhaps that needs to be investigated? High cortisol levels are not harmless.

 

[Joan Crawford]

This research is presented as a "prospective cohort study". That's true for the review of whether or not pt's developed MS, but not so for records of abuse. That data was assessed retrospectively via questionnaire. That is a weakness/problem.

I detailed the multiple issues with retrospective, questionnaire type methodology for collecting this type of data via a paper in Psychological Medicine in 2021. This was in relation to FMS - though I suspect the same/similar applies in MS:
https://www.cambridge.org/core/jour...ycheva-et-al/F8AC978336EDA0E9AAFCEBBC8CA9641E
If you can't access the article let me know.

In summery: This method is likely to find a positive link - usually small. It is unlikely to find a real-world effect. No psychotherapeutic approach for trauma resolves pt's pain and other physical symptoms - as would be expected if causative. Repeating this type of research over and over causes unnecessary distress for pt's and those caring for them.

It can be seductive for media and society to believe in this dubious type of link. Perhaps an effort at keeping them apparently psychologically safe from harm. It doesn't, but that's not the point.

The high level of abuse discovered is much more shocking than some dubious, and possibly artefactual link.

I would have hoped that the link between EBV and MS would be more worthy of media attention.

 

[Hutan]

Abuse of any kind is very likely to raise cortisol levels. Perhaps that needs to be investigated? High cortisol levels are not harmless.

Sure, but this study didn't actually find any correlation between MS and childhood or adolescent
1. humiliation (‘Has anyone over a long period of time systematically tried to subdue, degrade or humiliate you?’),
2. threat (‘Has anyone threatened to hurt you or someone close to you?’),
3. or physical abuse (‘Have you been subjected to physical abuse?’),
4. or indeed 'any type of abuse'.

Maybe the sample wasn't big enough, but this paper is not evidence of an association between childhood abuse and MS.

 

[Hutan]

I detailed the multiple issues with retrospective, questionnaire type methodology for collecting this type of data via a paper in Psychological Medicine in 2021.

Thanks Joan. This paper was a bit different, as they asked the women to recall childhood abuse after childhood but before most of the women who went on to develop MS had done so. (Only 97 out of 397 had MS at the time they answered the survey.)

Actually the women who had already developed MS at baseline and so might have been expected to over-report childhood abuse reported less than those who did not have MS at the time they were answering the survey.

 

[Snow Leopard]

This research is presented as a "prospective cohort study". That's true for the review of whether or not pt's developed MS, but not so for records of abuse. That data was assessed retrospectively via questionnaire. That is a weakness/problem.

The data was collected at baseline for the women who were pregnant (as part of the MoBa Survey) between 1999–2008, whereas the MS data was based on medical records (which can have biases!) up to December 2018.

edit - it seems Hutan beat me to it!

 

[Snow Leopard]

Here are the results reported in the paper:

View attachment 17156

That table seems to have some errors in the quoted percentages.

The authors state that 300 women developed MS during the follow-up period and the table reports 229 no abuse, 71 with abuse (=300 so it adds up).

But 56/300 is not 20%, 48/300 is not 17%, 20/300 is not 8%, 34/300 is not 13%, 22/300 is not 9%. What is going on?

That aside, the effect for the sexual abuse category is certainly significant statistically, at least for the raw data. (My odds ratio calculations give the same numbers)