Is This Study Flawed?

Factual Questions
1

This article reports on a study showing that infants who grow up in homes with multiple pets are less likely to develop allergies than infants who grow up in homes without pets, or with fewer pets. The conclusion is that the infants are desensitized by the pets’ presence and are then less likely to develop allergies because of the pets’ presence.

The flaw in this logic seems to me that allergies are inherited, and therefore a child with allergies is more likely to have parents who have allergies. Parents who have allergies are less likely to have pets in the home.

Is my reasoning flawed, is the reasoning of those who did the study flawed, or is the article flawed by not reporting that the study someone factored heredity into their findings?

2

There may be more to allergies than the genetic pre-disposition.

However, I REALLY LOVE your analytical view of the causal relationship. I wish everyone, including so-called experts knew what the difference was between correlation and causal effects.

3

Here’s a link to the article in question:

http://jama.ama-assn.org/issues/v288n8/rfull/joc02287.html

It’s written in fairly plain English, and, while I haven’t gone through their methodology in detail, they did take parental allergies into account.

Aside from this article, there has been a slew of recent research indicating that the right kind of challenges to the immune system as a child tends to expand the immune repetoire and reduce hyperreactivity, leading to fewer cases of allergies and asthma. Oddly, one of the “right kinds” of challenges appears to be intestinal worms.

Pinworm is apparently good for you.

mischievous

4

I haven’t read the original article but I doubt that it is flawed. However, I am certain that it is an inherently weak design that cannot give a definitive answer. The only definitive answer can come from a double blind randomized controlled trial (e.g., get a bunch of volunteer families and randomly give one set of families real pets and the other fake pets that cannot be distinguished from real pets and see what happens). This study is an observational study that is subject to what is called residual confounding. For example, say that allergies are inherited so that parents with allergies are more likely to both have children with allergies and not have pets. One can then say that the relationship between pets and allergies is confounded by heritability of allergies. The way to control this is to adjust, statistically, for allergies among the parents. (For example, look at the relationship between pets and child allergies in families with allergic parents separately from same relationship in families without allergic parents). I have to believe that the people who did the study must have done this (“controlled” for allergies among the parents) and that the news report simply failed to report it. The problem is that you cannot always completely control for confounding. For example, maybe there are other factors you fail to control for. For example, maybe parents without pets are fussier and more likely to take their kids to the pediatrician every time they sneeze and thus more likely to be diagnosed as allergic.

Residual confounding is likely to be the reason that the results of a recent, large scale double blind randomized controlled trial of estrogen replacement therapy in postmenopausal women differed from the results of numerous well done observational studies.

5

Uh…, but you haven’t read the article, yet you’re certain that it is an inherently weak design that cannot give a definitive answer, but you doubt that it is flawed?

6

Well then … based on my childhood medical “challenges” as a Foreign Service brat on the Ivory Coast I’ll live forever!

7

Here is a link to some older research. My understanding of it is that children benefit from exposure to livestock and other natural settings because their developing immune systems are exposed to more microbes, etc. and therefore build a defense.

The link refers to a study of East and West germany right after the wall came down. The researcher theorised that the kids growing up in the crouded, dirty east where everyone had livestock in the yard would be very prone to allergies. They found quite the reverse. It turns out that the kids living in the clean suburbs who have never even seen livestock while it was, you know, alive had a much higher rate of allergies.

The current theory is that exposure when young helps your body set itself up for dealing with the allergens when grown

The link:

Hygiene Hypothesis

8

I wonder if the researchers took into account the reason some people with allergies did not have pets was in fact because they had allergies.

The design of this experiment is flawed not counting what I just asked. The sample size was too small, there werent sufficient controls in place. Without the above post in front of me, whomever said that a double blind should have been put in place was correct. Or at least on the right track. But you cant correct flawed design.

9

“Uh…, but you haven’t read the article, yet you’re certain that it is an inherently weak design that cannot give a definitive answer, but you doubt that it is flawed?”

Ringo, although it I have not read the original article, I know from news reports that it was an observational study, not a randomized controlled trial (it helps that an RCT would have been impossible) and that an observational study is inherently weak. I doubt that it is flawed because it was, AFAIK, done by competent people and published in a respectable journal.

When a study is well done but not conclusive because of its design, scientists don’t speak of the study’s flaws but of its “limitations.” Yes, the study has limitations.

I suppose you could say that the differnce between a limitation and a flaw is that investigators are aware of and acknowlege limitations whereas they screw up by missing flaws.

10

Here’s what the original article says:

“In addition, we could not detect differences in the relationship between dog and cat ownership and parental history of asthma, allergy, or hay fever among those examined”

That is, the same number of parents with allergies (which may not be pet related) buy pets as those who have no allergies. However, they took it into account anyway:

“Logistic regression analysis was used to adjust for the effects of possible confounding variables (cord serum IgE concentration, levels of house dust mite allergen in the child’s bedroom at age 2 years, child’s sex, an older sibling, passive exposure to parental tobacco smoke, and parental history of asthma) on relationships between dog and cat exposure and risks of atopy and seroatopy.”

Yeah, ALL studies have limitations. Including double-blind ones, which could not have been done in this case anyway. A good scientist is aware of the limitations in his/her particular study. Again from the article:

“A second caveat is the limited racial, socio-economic, and geographic diversity of our study population, suggesting that our conclusions can only be applied to similar populations of white children. Since our follow-up was limited to an average age of 6.7 years, we do not know if the associations we found will persist as the children grow older, but others have found that the association between dog and cat exposure and a lower risk of allergy-related symptoms persisted to age 12 to 13 years”

There are several other caveats, feel free to read the article.

In my opinion, mangoldm, the study definitely does not have the specific flaw you were asking about, i.e. the confounding of parental allergies and of parental pet-owning tendencies. No study is perfect, this one included, but I think it has done all of the correct things in order to reach the conclusion that it did reach. YMMV.

Let me echo K364 and say that it is nice to see people putting thought into what they see in the news.

11

“In my opinion, mangoldm, the study definitely does not have the specific flaw you were asking about, i.e. the confounding of parental allergies and of parental pet-owning tendencies.”

mischievous, I beg to disagree. I don’t understand how you could possibly conclude that “the study definitlely does not have” residual “confounding of parental allergies and of parental pet-owning tendencies.” All allergies are not equal and all pets are not equal so there is ample opportunity for residual confounding. However, this is a limitation of the study, not a flaw.

12

Yeah, I think we’re quibbling about minor details. Again, in my opinion, with which you are free to disagree, I think they did the correct controls, and I think the study is valid within its range. You can argue that maybe some allergies affect certain pet ownership, and that those allergies, but not others, are inheirited, which would skew the results of this study. I find this unlikely, but as you say, not impossible. However, the authors did everything in their power to control for parental allergies. The fact that they are not omnipotent nor omniscient should not prevent them from being acknowledged for putting out solid research. No one can conduct the perfect study, and yet somehow science advances.

13

mischievous, I don’t claim that there is anything wrong with their controls, or that the study is invalid, or that they shouldn’t be acknowledged for putting out solid research. All I’m saying is that observational studies like this have real limitations. As the estrogen replacement debacle demonstrates, one puts one’s faith in observationa studies at one’s own risk.