Hacker News Reader

Comment by ekianjo | Hacker News Reader

The way the data was collected looks like junk in the first place:

> The subject also had to answer a question “Are you infected with Toxoplasma, the cat parasite that is dangerous especially to pregnant women?” by ticking one of the three suggested answers: “I do not know or am not sure, I was not laboratory tested,” “No (I was tested by a medical doctor and the blood test gave a negative result,” and “Yes (I was tested by a medical doctor and the blood test gave a positive result.” Implicitly, the first answer (“I do not know …”) was ticked.

Ideally you would medically test the patients and make sure they are unaware for what they are tested. Asking them to answer such a question is going to massively decrease whatever you can say about what's claimed.

View on HN

Read further; the subjects of this study were also examined and had labs drawn which included a toxo test, and were issued unique codes which they provided with their study responses:

> At the end of the questionnaire, the participants could provide their unique “guinea pig code.” Because of the rather sensitive sex- and health-related contents of this questionnaire, it was “signed” only by 8% of the responders. However, other questionnaires distributed in this community were signed by most of responders (Flegr & Hodny, 2016). These results showed that most of participants who were aware of their toxoplasmosis status (and nearly all male responders) had been tested for toxoplasmosis in our lab.

Unless I'm badly misreading the paper, they were therefore able to correlate study responses with confirmed presence or absence of latent toxo, with a high degree of specificity. It's odd that they would also include a question on the same subject, and I'm not competent to evaluate their methods of statistical analysis, but even the attempt to correlate study responses with laboratory data puts them head and shoulders above pretty much anyone else I've seen publish on this topic.

Let me quote that again:

> These results showed that most of participants who were aware of their toxoplasmosis status (and nearly all male responders) had been tested for toxoplasmosis in our lab.

So they only had a confirmation for the positive/negative cases, but they have no way to confirm that the ones who say "not sure" were accurately reporting their answers. That's a pretty big gap they use to draw conclusions.

Fair. But also this, the first sentence of the "Limitations" section:

> The major limitation of the present study is that it is based on self-reported information on subjects’ sexual life and self-reported Toxoplasma infection status.

They go on to caution against overgeneralization based on these extremely specific and inherently uncertain results, which is entirely reasonable, but I think it's rather hasty to write off the study as "junk". At the very least, it's better and more careful science even despite its limitations than one ordinarily sees performed in this realm.

I'm not sure you've correctly interpreted that quote.

  most of participants who were aware [...] 
  had been tested [...] in our lab.

This does not preclude any blood draws, in which the subject is tested, but not informed of results and infection status.

They can test samples drawn expressly for the experiment, and not inform the individual, and then still use the lab-verified determination of infection status in the results. They can take blood, find out, and not tell the people if they are infected, even if they didn't know, and let them stay infected without knowing.

In addition, unclear as to how this is causative rather than correlative. Perhaps people who keep pets, or cats in particular, are more likely to have the described personality traits.

yes, the author is known for a lack of rigor.

Know from other publications?

known by the leading toxoplasma researchers at Stanford.

Reply on news.ycombinator.com