Emotional distress in infertile women and failure of assisted reproductive technologies: meta-analysis of prospective psychosocial studiesBMJ 2011; 342 doi: https://doi.org/10.1136/bmj.d223 (Published 23 February 2011) Cite this as: BMJ 2011;342:d223
All rapid responses
We thank Lynch and Domar for their interest in our meta-analysis and
address their concerns herewith.
Lynch and Domar questioned the use of anxiety versus depression
measures, the use of state versus trait measures and the timing of
psychological assessments. We included only one construct of emotional
distress per study to avoid correlated effect sizes. We prioritised
anxiety based on theory (1) indicating that state anxiety would be more
closely aligned to the anticipatory reactions associated with initiating
treatment (e.g., tension, worry) than depression, which is more indicative
of outcome emotions (e.g., sadness, feelings of loss). In the review
process we carried out a sensitivity analysis to examine results if we
prioritised depression instead. The pooled effect size was similar and
also not significant: -0.06 [95% CI -.13, 0.01, I2=40%]. The correlation
between state and trait anxiety effect sizes was high (r= .93) suggesting
a redundancy of information if both were included, and this was born out
by the similarity in meta-analysis results if we prioritised trait anxiety
instead of state anxiety (i.e., -0.04 [95% CI -.11, 0.04, I2=0%]).
Finally, and as reported in our article, sensitivity analysis showed that
meta-analysis results did not differ by the timing of assessment, that is,
whether anxiety was measured less or more than one month before the start
of treatment. Additional meta-analysis of only those five studies that
assessed anxiety within two weeks of the start of treatment produces
similar non-significant results 0.01 [95% CI -.11, 0.12, I2=43%]. These
results are consistent with other research showing that although state
anxiety is more sensitive to context than trait anxiety, both demonstrate
high stability over a two-year span (3). The difference across these
sensitivity analyses is in the heterogeneity index, which we believe is
due to the greater number of self-report inventories used to measure
depression compared to anxiety (and state anxiety compared to trait
anxiety). Overall the sensitivity analyses demonstrate that our meta-
analysis results are robust.
Lynch and Domar also express concern about pooling across outcomes (a
positive pregnancy test, an ultrasound confirming fetal viability, or a
live birth). In fact we reported on a sub-group analysis comparing each
outcome that showed higher effect sizes in studies using the pregnancy
test as an outcome measure. However, owing to the lack of a plausible
theory to account for a stress effect at implantation that is not followed
through to later pregnancy, and the fact that these four studies were the
smallest of those included in our meta-analysis, we proposed small study
bias as the most likely cause of this subgroup difference.
We agree that answering this research question is a challenging task.
The method they propose and the problems they identify with this
methodology concur with our discussion of methodological issues, indeed we
were the first to document the impact of ART monitoring on patient
distress (4). However, we disagree that 'keeping patients blinded as to
their cycle progress' is the only way to address the issue for obvious
ethical reasons (withholding information) but also because continuous
stress measurement during treatment is not necessary when there is
considerable stability in anxiety (3). We discuss alternative approaches
in the paper.
Finally, Lynch and Domar express concern about the inclusion of two
large studies. We followed the PRISMA guidelines in our meta-analysis and
included all evidence that we could obtain. The data produced by Lee et
al. 2006 was available and therefore needed to be included in the meta-
analysis. Ignoring studies fulfilling the inclusion criteria based on
publication status would introduce publication bias in our review.
Therefore, as per recommendations we reported the study, assessed its
quality (judged low), and investigated whether low quality studies
demonstrated different effect sizes (not significant). Our colleagues also
queried whether the third largest study included in the meta-analysis
(Lintsen et al., 2009) had a floor effect, in that the mean anxiety score
was low, which would limit the ability to establish any relationship
between distress and conception. However, the lower mean score in the
Lintsen paper is because they used the abridged state anxiety measure (10
of 20 items only) with a range of 10 to 40. The reported mean value of
17.7 would correspond to a mean of 35.4 on the full inventory, which is
consistent with general findings in this population. Further, the large
sample size (N=690) gave sufficient power to detect a difference in
anxiety of 1.25 points between subsequently pregnant and nonpregnant
groups, which shows a sensitive experimental design. Therefore our
inclusion of these studies was based on best practice in meta-analysis.
We disagree that our conclusions were too broad in light of the study
results. Indeed the consistency in results and lack of significant
deviation after thorough exploration with sensitivity and sub-group
analyses adds to our confidence. We make clear in our Discussion that we
are referring to women undergoing a single cycle of ART, and that more
research would be required to broaden the applicability of the findings
(i.e., untreated or differently treated couples, men, etc). As noted, our
preliminary analyses with depression yielded the same conclusion but in
retrospective we might have added a footnote in the article to indicate
that prioritizing depression produced the same meta-analysis result.
We agree that our meta-analysis is probably not the final word on
emotional distress and ART. However, we do believe it presents the latest
word on this topic.
1. Folkman S, Lazarus RS. If it changes it must be a process: Study
of emotion and coping during three stages of a college examination. J Pers
Soc Psychol. 1985;8:150-70.
2. Verhaak CM,.Smeenk JMJ,.Evers AWM, Kremer JAM, Kraaimaat FW, Braat DDM.
Women's emotional adjustment to IVF: a systematic review of 25 years of
research. Hum Reprod Update. 2007;13(1):27-36.
3. Usala PD, Hertzog C. Evidence of Differential Stability of State and
Trait Anxiety in Adults. J Pers Soc Psychol 1991;60(3): 471-79.
4. Boivin J,Takefman JE. Stresslevel across stages of in vitro
fertilization in subsequently pregnant and nonpregnant women. Fertil
Competing interests: As per our original statement in the article
Meta-analysis is not the final word on the effect of emotional distress on fertility treatment outcomes
We read with interest Professor Boivin and colleagues' meta-analysis
examining the effect of "emotional distress" on assisted reproductive
technologies (ART) failure. This has been a controversial area of research
and the authors are to be commended for attempting to synthesize the
relevant work. The implications for a relationship, or a lack of one,
between psychological distress and IVF outcome are substantial and as
such, we feel compelled to point out several limitations that raise
concern regarding the validity of the reported findings.
First, one of the basic tenets of meta-analysis is that the studies
being combined must include exposure and effect measures that are similar
enough that when pooled yield meaningful results. This principle was not
followed in the authors' analysis. With regard to exposure, the authors
used various measures of anxiety and depression "prior to stimulation" as
a proxy for emotional distress. The authors further stated that they
prioritized state anxiety as it is a better measure of anxiety at the
moment. However, in this context, state anxiety is not the appropriate
measure. State anxiety is defined as recording how one feels that very
moment and it is not a good proxy for an underlying tendency towards
anxiety. According to the instructions on the Spielberger State Inventory,
one is to indicate "how you feel right now, that is, at this moment". In
nine of the 14 studies, the psychological assessment was made prior to
cycle initiation, up to three months prior in fact, which depending on the
individual's circumstances might not at all reflect their anxiety levels
at the time treatment occurred. It has been noted that infertility
patients report renewed optimism and reduced psychological distress prior
to their first IVF cycle, as the treatment modality is known to yield the
highest per cycle pregnancy rates.
With regard to their choice of outcomes, the authors included studies
in which the outcomes of interest were a positive pregnancy test, an
ultrasound confirming fetal viability, or a live birth. While the first
two outcomes may appear similar enough, it is well-recognized that many
pregnancies, particularly those resulting from ART are lost prior to
ultrasound confirmation. While the true outcome of interest varies
according to the research question (i.e., effect of distress on pregnancy
versus live birth rates), it is inappropriate to combine these
The fact is, answering this research question is a challenging task.
To truly establish whether or not there is a significant relationship
between psychological distress and treatment outcome, one would have to
assess psychological state just prior to and throughout the treatment
cycle. However, the assessment of cause versus effect is complicated by
the fact that the vast majority of IVF patients receive feedback as to how
their cycle is progressing. Patients learn within days of cycle initiation
how well they are responding to the medications designed to stimulate
their ovaries. Patients who respond less vigorously have poorer prognoses,
and this knowledge is likely to impact their emotional state. Conversely,
patients who are responding well to the medications are more likely to
conceive, and this knowledge is likely to lead to a more positive
emotional state. The only way to truly assess this relationship would be
to keep patients blinded as to their cycle progress.
In addition, there are questions about two of the three largest
studies included in the meta-analysis. The study with the largest sample
size (Lee et al.) was presented in 2006 but has not yet been published.
The third largest study (Lintsen et al., 2009) had a floor effect, in that
the mean anxiety score was low, limiting the ability to establish any
relationship between distress and conception.
Given the limitations of the research, we feel that it is premature
for the authors to assert such definitive conclusions. In the Discussion
section, they state that "these findings should reassure women that
emotional distress caused by fertility problems or other life events is
unlikely to be a cause of failure of fertility treatment". Later on, they
go on to state that their findings..."provide doctors with the evidence to
reassure women that feelings of tension, worry or depression experienced
as a result of their fertility problem, its treatment, or other co-
occurring life events are unlikely to further reduce chances of
pregnancy". In fact, the most commonly used psychological questionnaire
for this meta-analysis as noted above was State Anxiety. Thus,
extrapolating the conclusions to depression and worry are unsupported.
We encourage the authors to rethink their analysis and consider
grouping the studies into like exposures and outcomes that can be
appropriately combined. While the studies examining the effect of
psychological distress on fertility outcomes to date have been
historically underpowered and yielded largely conflicting results, there
is certainly a substantial body of biologically plausible etiologic work
suggesting that the association warrants further study.
Competing interests: No competing interests