On 2014 Jul 14, David Reardon commented:

Given the excellence of this data set, it is unfortunate that the researchers failed to consider the methodology employed in prior record based studies of abortion and mental health (for example, Reardon DC, 2003 and Coleman PK, 2002) and instead made choices which tend to confound rather than clarify the issues at hand.

Perhaps the biggest problem is the failure to examine pre-pregnancy mental health. This was done in prior studies which controlled for psychiatric admissions for at least one year prior to the calculated date of conception (Reardon DC, 2003). Instead, Munk-Olsen chose as a base line for mental health a nine month period prior to pregnancy outcome, a period covering the entire period of pregnancy for delivering women and approximately three months of pregnancy and six months pre-pregnancy for women who aborted.

This poorly explained decision unfortunately introduces elements of an apples versus oranges comparison.

This is problematic for two key problems. First, it totally ignores the pre-pregnancy mental health of delivering women. Women who are excited about having a baby may be less likely to seek mental health. With only a nine-month pre-event window, the authors are even excluding those women who may have been experiencing anxiety or depression if they were struggling to become pregnant.

Secondly, for aborting women, it includes two to three months of a period in the women's lives which is likely highly stressful since these women are, it must be presumed, facing discovery of an unplanned pregnancy and potential conflicts over this with partners, parents, and others. This nine month window therefore not only fails to provide a base line for mental health prior to the subjects' pregnancy but mixes, several months of pre-pregnancy mental health with one or more months of post-pregnancy, pre-abortion stress.

These study design criteria are not only unprecedented in similar studies but simply contrary to the stated objective of controlling for prior mental health. This failure in design is even more puzzling given the fact that the available data set included mental health information for the entire life of the women in the study. Given the availability of all prior mental health treatment dates, it seems self evident that the researchers should have created a scale for exposure to prior mental health treatments prior to the estimated conception date of each woman's first pregnancy covering at least one full year, preferably five years, and perhaps for each subject's entire life.

Given the fact that the authors only exclusion criteria was a history of inpatient psychiatric treatment, we must assume that since the mental health of both groups was relatively similar prior to becoming pregnant. The three fold increase in mental health treatments prior to abortion (14.6 per 1000 yrs) compared to women who gave birth (3.9 per 1000 yrs) would therefore appear to be most likely explained by stress these women faced discovering they were faced with an unplanned pregnancy and, in many cases, the concurrent disruption of relationships with male partners, parents, employers and others.

In addition, aborting women were more likely to involved in unstable and possibly abusive relationships prior to their pregnancies. It is not surprising, then, that many of these women facing the stress of abortion decision-making sought psychiatric advise at a much more elevated rate than they had in the past.

The lack of pre-pregnancy mental health measure, in itself, renders it impossible to draw general conclusions regarding from this study. But there are also numerous additional problems with the study design that further confound the results and interpretation:

• The study excluded women who are most likely to have the most severe reactions to an abortion, namely those who already had a prior history of abortion, those with a prior history of inpatient mental health treatment. It also excluded women who died prior to one year after the pregnancy event, thereby excluding women who committed suicide, even though a record linkage study from neighboring Finland found a six fold higher rate of suicide in the year following abortion compared to the year following childbirth (Gissler M, 1996).

• The study failed to keep women in the two groups separate. Women who had both an abortion and delivery were included in both parts of the analysis groups. This is especially problematic given the evidence women with a prior history of abortion have more stress during and after subsequent pregnancies (Coleman PK, 2002). As a result, by putting women with a history of abortion in the group of delivering women, the "control" group may be adulterated with the very post-abortion effects this study professes to be exploring. The authors should have limited the study to first pregnancy outcome, and should also have included first pregnancies ending in miscarriage, still birth, and other natural losses as a third group.

• The outcome variable used for this study was limited to only a single psychiatric treatment. It did not measure or give any weight to repeated treatments or multiple mental health problems in an effort to evaluate the severity or duration of mental health problems associated with pregnancy outcomes. It should also be noted that the authors' conclusions are not consistent with their data. Specifically, the authors conclude: "our study shows that the rates of a first-time psychiatric contact before and after a first-trimester induced abortion are similar." The fact that many of the psychiatric contact rates during the nine months preceding an abortion were similar to the contact rates in the year following an abortion were statistically insignificant does not alter the fact that, even with all of the problems in the methodology identified above, their data also showed several statistically significant higher rates for specific mental disorders.

Specifically, as seen in table 1, contact rates for neurotic, stress-related or somatoform disorders were higher in every two month period following an abortion, and was statistically significant in two of the six periods. Similarly, for personality or behavioral disorders, the relative risk was higher for four of the six periods and significantly higher for two of the six periods. In other words, when one actually examines the table of reported findings, the conclusion that "in some cases the rate of a first-time psychiatric contact is significantly higher following an abortion compared to the nine months preceding an abortion" is at least equally true, and arguably more accurate.

Finally, I would note that at least one of the authors, Laursen, was also a co-author of an excellent record linkage study from Demark examining mental illness among parents of deceased children that was also published in the New England Journal of Medicine.(Li J, 2005) That study design examined long term mental health effects, beyond one year, and controlled for exposure to multiple losses, and had none of the selection bias problems identified in this Munk-Olesn study.

This exact study design could have, and should have, been used simply by plugging the date of exposure to induced abortion into the same fields used for date of exposure to the death of a child. By using a proven study design in this way, Munk-Olsen and Laursen could have avoided the impression (in this case, the very profound impression) that this new study design (with a nine month pre-event control period and intermixing of subjects into both the abortion group and delivery group, et cetera) was artificially constructed precisely to minimize the number of statistically significant findings associating abortion with increased mental health treatments. In conclusion, this study does nothing to impute previous record linkage studies showing elevated rates of psychiatric treatments following induced abortion.

On 2014 Jul 14, David Reardon commented:

Given the excellence of this data set, it is unfortunate that the researchers failed to consider the methodology employed in prior record based studies of abortion and mental health (for example, Reardon DC, 2003 and Coleman PK, 2002) and instead made choices which tend to confound rather than clarify the issues at hand.

Perhaps the biggest problem is the failure to examine pre-pregnancy mental health. This was done in prior studies which controlled for psychiatric admissions for at least one year prior to the calculated date of conception (Reardon DC, 2003). Instead, Munk-Olsen chose as a base line for mental health a nine month period prior to pregnancy outcome, a period covering the entire period of pregnancy for delivering women and approximately three months of pregnancy and six months pre-pregnancy for women who aborted.

This poorly explained decision unfortunately introduces elements of an apples versus oranges comparison.

This is problematic for two key problems. First, it totally ignores the pre-pregnancy mental health of delivering women. Women who are excited about having a baby may be less likely to seek mental health. With only a nine-month pre-event window, the authors are even excluding those women who may have been experiencing anxiety or depression if they were struggling to become pregnant.

Secondly, for aborting women, it includes two to three months of a period in the women's lives which is likely highly stressful since these women are, it must be presumed, facing discovery of an unplanned pregnancy and potential conflicts over this with partners, parents, and others. This nine month window therefore not only fails to provide a base line for mental health prior to the subjects' pregnancy but mixes, several months of pre-pregnancy mental health with one or more months of post-pregnancy, pre-abortion stress.

These study design criteria are not only unprecedented in similar studies but simply contrary to the stated objective of controlling for prior mental health. This failure in design is even more puzzling given the fact that the available data set included mental health information for the entire life of the women in the study. Given the availability of all prior mental health treatment dates, it seems self evident that the researchers should have created a scale for exposure to prior mental health treatments prior to the estimated conception date of each woman's first pregnancy covering at least one full year, preferably five years, and perhaps for each subject's entire life.

Given the fact that the authors only exclusion criteria was a history of inpatient psychiatric treatment, we must assume that since the mental health of both groups was relatively similar prior to becoming pregnant. The three fold increase in mental health treatments prior to abortion (14.6 per 1000 yrs) compared to women who gave birth (3.9 per 1000 yrs) would therefore appear to be most likely explained by stress these women faced discovering they were faced with an unplanned pregnancy and, in many cases, the concurrent disruption of relationships with male partners, parents, employers and others.

In addition, aborting women were more likely to involved in unstable and possibly abusive relationships prior to their pregnancies. It is not surprising, then, that many of these women facing the stress of abortion decision-making sought psychiatric advise at a much more elevated rate than they had in the past.

The lack of pre-pregnancy mental health measure, in itself, renders it impossible to draw general conclusions regarding from this study. But there are also numerous additional problems with the study design that further confound the results and interpretation:

• The study excluded women who are most likely to have the most severe reactions to an abortion, namely those who already had a prior history of abortion, those with a prior history of inpatient mental health treatment. It also excluded women who died prior to one year after the pregnancy event, thereby excluding women who committed suicide, even though a record linkage study from neighboring Finland found a six fold higher rate of suicide in the year following abortion compared to the year following childbirth (Gissler M, 1996).

• The study failed to keep women in the two groups separate. Women who had both an abortion and delivery were included in both parts of the analysis groups. This is especially problematic given the evidence women with a prior history of abortion have more stress during and after subsequent pregnancies (Coleman PK, 2002). As a result, by putting women with a history of abortion in the group of delivering women, the "control" group may be adulterated with the very post-abortion effects this study professes to be exploring. The authors should have limited the study to first pregnancy outcome, and should also have included first pregnancies ending in miscarriage, still birth, and other natural losses as a third group.

• The outcome variable used for this study was limited to only a single psychiatric treatment. It did not measure or give any weight to repeated treatments or multiple mental health problems in an effort to evaluate the severity or duration of mental health problems associated with pregnancy outcomes. It should also be noted that the authors' conclusions are not consistent with their data. Specifically, the authors conclude: "our study shows that the rates of a first-time psychiatric contact before and after a first-trimester induced abortion are similar." The fact that many of the psychiatric contact rates during the nine months preceding an abortion were similar to the contact rates in the year following an abortion were statistically insignificant does not alter the fact that, even with all of the problems in the methodology identified above, their data also showed several statistically significant higher rates for specific mental disorders.

Specifically, as seen in table 1, contact rates for neurotic, stress-related or somatoform disorders were higher in every two month period following an abortion, and was statistically significant in two of the six periods. Similarly, for personality or behavioral disorders, the relative risk was higher for four of the six periods and significantly higher for two of the six periods. In other words, when one actually examines the table of reported findings, the conclusion that "in some cases the rate of a first-time psychiatric contact is significantly higher following an abortion compared to the nine months preceding an abortion" is at least equally true, and arguably more accurate.

Finally, I would note that at least one of the authors, Laursen, was also a co-author of an excellent record linkage study from Demark examining mental illness among parents of deceased children that was also published in the New England Journal of Medicine.(Li J, 2005) That study design examined long term mental health effects, beyond one year, and controlled for exposure to multiple losses, and had none of the selection bias problems identified in this Munk-Olesn study.

This exact study design could have, and should have, been used simply by plugging the date of exposure to induced abortion into the same fields used for date of exposure to the death of a child. By using a proven study design in this way, Munk-Olsen and Laursen could have avoided the impression (in this case, the very profound impression) that this new study design (with a nine month pre-event control period and intermixing of subjects into both the abortion group and delivery group, et cetera) was artificially constructed precisely to minimize the number of statistically significant findings associating abortion with increased mental health treatments. In conclusion, this study does nothing to impute previous record linkage studies showing elevated rates of psychiatric treatments following induced abortion.