Research paper
art ic l e i nf o a b s t r a c t
Article history: Background: Risk and protective factors for postpartum depression have been extensively studied, and in
Received 11 December 2015 recent studies an association between breastfeeding and maternal mood has been reported. The present
Accepted 11 April 2016 retrospective, cross-sectional study was conducted to evaluate the association between breastfeeding-
Available online 19 April 2016
related variables and postpartum depression (based on Edinburgh Postnatal Depression Scale threshold
Keywords: criteria) within the context of other known risk factors.
Breastfeeding Method: Breastfeeding information, demographic information, and scores on the Edinburgh Postnatal
Household income Depression Scale were examined from the Canadian Maternity Experience Survey. This survey contains
Postpartum depression data collected from 6421 Canadian mothers between October 2006 and January 2007, and 2848 women
Risk factors
between ve and seven months postpartum were included in the current analyses.
Social support
Results: In contrast to previous research, logistic regression analyses revealed that when considered
Stress
within the context of other risk factors, breastfeeding attempt and duration were not associated with
postpartum depression at ve to seven months postpartum. Although a relationship between the pre-
natal intention to combination feed and postpartum depression was observed, these variables were no
longer related once other potential risk factors were controlled for. Factors that were associated with
postpartum depression included lower income, higher perceived stress, lower perceived social support,
no history of depression, or no recent history of abuse.
Limitations: A clinical diagnostic instrument was not used and variable selection was restricted to data
collected as part of this survey.
Conclusion: These ndings suggest that the association between breastfeeding and postpartum de-
pression reported by previous researchers may in fact be due to alternative risk factors.
& 2016 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jad.2016.04.014
0165-0327/& 2016 Elsevier B.V. All rights reserved.
46 C.J. Pope et al. / Journal of Affective Disorders 200 (2016) 4550
Table 3
Logistic Regression of breastfeeding intent, attempt, and postpartum depression risk factors with depressive symptoms (EPDS 412).
95% condence
Interval for OR
Note. OR Odds Ratio. Operational denition for signicant risk variables: Level of stress past year [The amount of stress in their life during the 12 months before the index
baby was born (1 not stressful, 2 somewhat stressful, 3 very stressful)]; History depression/antidepressants [Ever been prescribed anti-depressants or been diagnosed
with depression prior to pregnancy (0 no, 1 yes)]; History of abuse (past 2 years) [Experienced verbal, sexual, or physical abuse in the past two years (0 no, 1 yes)];
Level of social support [availability of support when she needed it since the birth of the index baby (1 none of the time, 2 a little of the time, 3 some of the time, 4 most
of the time, 5 all of the time)].
a
Analysis 1, 2, and 3 Ns were 2600, 2608, and 2347, respectively.
b
prior to pregnancy.
c
When entered as either continuous or categorical variables, breastfeeding variables were not signicant in analysis 3, continuous values are presented.
*
p o .05.
**
po .001.
depression. However, the effect size of this association is negli- depression (Thome et al., 2006; Ystrom, 2012). In this study, we
gible. No association between any breastfeeding engagement and examined the dose response effect in two ways: rst, by examin-
EPDS scores was found (r(2812) 0.02, ns). ing duration of exclusive breastfeeding, and second, by examining
In order to assess if there was an association between the duration of any breastfeeding. Correlational analysis revealed that
duration of exclusive or any breastfeeding between women who duration of exclusive breastfeeding was negatively associated with
were depressed (EPDS 412) compared to women who were not scores on the EPDS. However, the effect size of this association was
depressed (EPDS score o10) a binary logistic regression was negligible, suggesting that this association is unlikely to translate
conducted. Neither duration of exclusive breastfeeding (2 (1) into clinically relevant observations. This assertion was further
0.53, ns), nor any breastfeeding (2 (1) 0.12, ns), signicantly supported by the binary logistic regression that did not show a
predicted meeting EPDS threshold criteria for depression. signicant association between duration of exclusive breastfeed-
ing, comparing women who were not depressed to women who
were depressed using EPDS threshold criteria. Further, in contrast
5. Discussion with other research, the results of the present study did not reveal
an association between duration of any breastfeeding and risk of
The primary purpose of this study was to determine if women postpartum depression when examined by correlational analysis
who did not attempt to breastfeed were at an increased risk for or binary logistic analysis. Thus, the results of the current study
postpartum depression. As well, we sought to determine if pre- align with that of the prospective study by Bogen et al. (2001) who
natal intention to breastfeed inuenced the association between also failed to nd an association between breastfeeding intention,
breastfeeding and postpartum depression. The results of this initiation, or duration and postpartum depressive symptoms
study, derived from a large population-based sample, failed to measured at 2 and 12 weeks postpartum.
support ndings of previous literature (Figueiredo et al., 2013; The large sample size used in the current investigation, per-
Seimyr et al., 2004) that suggested an association between mitted detection of even small differences (Streiner, 2003).
breastfeeding and postpartum depression. Specically, when other Moreover, the use of an extreme group approach, which excluded
risk factors were considered, breastfeeding attempt or intention to individuals endorsing subclinical levels of depression from the
breastfeed did not reduce the risk for postpartum depression. As analysis, upwardly biased estimates of standardized effect size.
well, previous research has suggested a dose response effect of This increased the likelihood of detecting a signicant difference
breastfeeding with postpartum depression (Hahn-Holbrook et al., between groups (Preacher et al., 2005). Thus, while non-signicant
2013). That is, women who breastfeed more or for longer periods ndings do not assure that no association exists, when viewed in
of time were reported to be at a reduced risk for postpartum light of the considerably large sample size and method of
C.J. Pope et al. / Journal of Affective Disorders 200 (2016) 4550 49
categorization that encouraged differences to be detected, the re- following considerations. The measure of postpartum depression
sults of this study strongly suggest breastfeeding status does not taken in this sample was not a clinical diagnostic instrument.
act as a standalone risk factor for the experience of postpartum However, currently there is no gold standard for measuring post-
depression. partum depression (Vliegen et al., 2014) and the EPDS is the most
As part of this investigation the association of other factors commonly used screening instrument in research for assessing
(previously identied in existing literature as risk factors for depression throughout the perinatal period (Hewitt et al., 2009).
postpartum depression) were also examined. Interestingly, not all As well, administering clinical interviews to a population-based
of the risk factors identied in previous research were found to be sample comes with obvious practical restraints, as such interviews
signicant risk factors for postpartum depression in this sample. In require specialized training and much more time to administer. In
particular, when considering all potential risk factors in the ana- addition, some episodes of postpartum depression may have been
lysis, mothers age, education level, marital status, smoking and missed, as depressive symptoms occurring before or after 57
drinking status were not found to be signicantly associated with months postpartum were not considered in the current in-
depression status. Nor were an unwanted pregnancy, urban living, vestigation. The method by which the data was derived also re-
number of previous live births, or the time postpartum the mother sulted in a few limitations. In particular, the questions used to
returned to work. This suggests that when other risk factors are obtain data in this research were developed for the purposes of
considered, these specic factors may not in fact put women at the CMES. Thus, the current study was restricted to the use of data
increased risk for postpartum depression. However, in agreement collected as part of this survey. For instance, the CMES asked if
with the previous literature, lower household income, higher women had a history of antidepressant use or a diagnosis of de-
perceived stress, and lower perceived social support were found to pression before pregnancy but did not ask women about anti-
be signicantly related to depression status. depressant use or diagnosis of depression during pregnancy or the
In contrast with previous research no history of depression or postpartum period. As well, information pertaining to potential
antidepressant use increased the risk of endorsing EPDS score412 serious medical condition for the infant was not available.
and not endorsing the experience of abuse in the past two years It would be ideal if future research could attempt to replicate
increased the risk of endorsing an EPDS score412. These ndings the results of this study with other large population-based sam-
are counterintuitive to what would be hypothesized based on the ples. It would also be interesting to see if similar results are seen
existing research (Meltzer-Brody et al., 2013) and may be a result when considering other time points during the postpartum period
of the broad way these variables were classied. It is also notable and when using a prospective design. While the current study did
that some of the risk factors identied in the current investigation not look at course of treatment, the observation that the risk
are similar to factors that have been found by previous researchers factors identied in this sample may suggest a more chronic
to predict a more chronic course of postpartum depression. Spe- course of depression brings light to the need for more research
cically, lower family income, less perceived social support, and regarding the course of postpartum depression. Further, research
stressful life circumstances (Pope et al., 2013, 2014; Vliegen et al., investigating the factors that increase the risk of earlier versus
2014) are implicated in more chronic postpartum depression. later onset of the disorder as well as factors that may predict a
Thus, the current investigation may have captured women with a longer course is also needed. This information may be valuable in
more persistent form of postpartum depression. Finally, we in- guiding clinical decision making regarding: 1) which women re-
cluded a large number of risk factors identied in the empirical quire more careful monitoring, 2) the length of time women
literature. Interestingly, the results of the current study suggest should be monitored for as informed by research and risk factors,
that even with the inclusion of many empirically supported risk and 3) which women are at risk of a more chronic course as this
factors, there is still a very large proportion of variance un- may be particularly important when considering pharmacological
accounted for (approximately 78%). However, in binary logistic interventions.
regression an equivalent statistic to R-squared does not exist, so
this estimate is based on a pseudo R square value (Tabachnick and
Fidell, 2014). Based on this estimate it appears that even when we 6. Conclusions
consider a number of empirically supported risk factors, the fac-
tors leading to postpartum depression are still largely unknown. The results of this study suggest that breastfeeding status alone
More recent research points at potential underlying or mediating may not be a signicant risk factor for postpartum depression.
biological mechanisms, such as alterations in hormonal levels, Thus, if a relationship does exist, it is more complicated than the
inammation, and genetic predisposition (Skalkidou et al., 2012). way the relationship is currently conceptualized. Moreover, while
The role of these biological variables was not assessed in the breastfeeding status did not predict postpartum depression in this
current study as information pertaining to these particular vari- study, support was found for the role of other previously reported
ables was not available. risk factors. The results of this study suggest that future research
The results of this study should be interpreted within the should seek to better understand the inuence that nancial dif-
connes of its strengths and limitations. The large population- culties, higher perceived stress, and limited social support have
based sample of postpartum women used is a notable strength of on the clinical prognosis of depression in new and expecting
this investigation. A larger sample size promotes more precise mothers.
estimates of the magnitude of an association and the ability to
detect small but statistically signicant associations that may be
Role of funding source
overlooked when smaller sample sizes are used. Larger sample V. Sharma has received grant support from, participated on scientic advisory
sizes have a reduced potential for producing false-positive results boards for, or served on the speakers bureaus of AstraZeneca, Bristol-Myers Squibb,
than do smaller samples and allow for covariates to be assessed in Cephalon, Eli Lilly, Janssen, Lundbeck, the Ontario Mental Health Foundation, Pzer,
a manner that encourages more interpretable results (Hackshaw, Servier, and the Stanley Foundation. C.J. Pope gratefully acknowledges nancial
support from the Canadian Institutes of Health Research.
2008). The sizable number of mothers included in this sample also
allowed for the inclusion of multiple covariates without the risk of
over-tting (models with too many variables for the number of Acknowledgments
observations; Babyak, 2004). Dr. Karen McQueen is thanked for her helpful comments. The authors are
The ndings of this study should be viewed in light of the grateful to the Public Health Agency of Canada and Statistics Canada for compiling
50 C.J. Pope et al. / Journal of Affective Disorders 200 (2016) 4550
the data and making it available for empirical exploration. The Maternity Experi- Morrell, J., Barkham, M., Light, K., Richards, D., 2009. Methods to identify
ences Survey was developed and implemented by the Public Health Agency of postnatal depression in primary care: an integrated evidence synthesis and
Canada in collaboration with Statistics Canada. The survey can be accessed through value of information analysis. Health Technol. Assess. 13 (36), 147230.
Research Data Centres located across Canada. For more information please contact McKee, M.D., Zayas, L.H., Jankowski, K.R.B., 2004. Breastfeeding intention and
the: Maternal and Infant Health Section of the Health Surveillance and Epide- practice in an urban minority population: relationship to maternal depressive
miology Division at the Public Health Agency of Canada cpss@phac-aspc.gc.ca. The symptoms and mother-infant closeness. J. Reprod. Infant Psychol. 22 (3),
work reported here is based on a Masters thesis submitted by the rst author in 167181.
partial fulllment of the requirement for the degree of Master of Arts in Clinical Meltzer-Brody, S., Boschloo, L., Jones, I., Sullivan, P.F., Penninx, B.W., 2013. The
Psychology. This research was presented at the 76th annual conference of the EPDS-Lifetime: assessment of lifetime prevalence and risk factors for perinatal
Canadian Psychological Association in Ottawa, Ontario on 5 June 2015. depression in a large cohort of depressed women. Arch. Womens Ment. Health
16 (6), 465473.
Nielsen Forman, D., Videbech, P., Hedegaard, M., Dalby, J., Secher, N.J., 2000. Post-
partum depression: Identication of women at risk. BJOG: Int. J. Obstet. Gy-
References naecol. 107 (10), 12101217.
OHara, M.W., McCabe, J.E., 2013. Postpartum depression: current status and future
directions. Annu. Rev. Clin. Psychol. 9, 379407.
Babyak, M.A., 2004. What you see may not be what you get: a brief, nontechnical Pope, C.J., Sharma, V., Mazmanian, D., 2014. Recognition, diagnosis, and treatment
introduction to overtting in regression-type models. Psychosom. Med. 66 (3), of postpartum bipolar depression. Expert Rev. Neurother. 14 (1), 1928.
411421. Pope, C.J., Xie, B., Sharma, V., Campbell, M.K., 2013. A prospective study of thoughts
Beck, C.T., 2001. Predictors of postpartum depression: an update. Nurs. Res. 50 (5), of self-harm and suicidal ideation during the postpartum period in women
275285.
with mood disorders. Arch. Womens Ment. Health 16 (6), 483488.
Bloch, M., Schmidt, P.J., Danaceau, M., Murphy, J., Nieman, L., Rubinow, D.R., 2000. Preacher, K.J., Rucker, D.D., MacCallum, R.C., Nicewander, W.A., 2005. Use of the
Effects of gonadal steroids in women with a history of postpartum depression.
extreme groups approach: a critical reexamination and new recommendations.
Am. J. Psychiatry 157 (6), 924930.
Psychol. Methods 10 (2), 178192.
Bogen, D.L., Hanusa, B.H., Moses-Kolko, E., Wisner, K.L., 2010. Are maternal de-
Public Health Agency of Canada, 2009. What Mothers Say: The Canadian Maternity
pression or symptom severity associated with breastfeeding intention or out-
Experiences Survey. Retrieved from http://www.phac-aspc.gc.ca/rhs-ssg/sur
comes? J. Clin. Psychiatry 71 (8), 10691078.
vey-eng.php.
Borra, C., Iacovou, M., Sevilla, A., 2015. New evidence on breastfeeding and post-
Seimyr, L., Edhborg, M., Lundh, W., Sjgren, B.J., 2004. In the shadow of maternal
partum depression: the importance of understanding women's intentions.
depressed mood: experiences of parenthood during the rst year after child-
Matern. Child. Health J. 19, 897907.
Cox, J.L., Holden, J.M., Sagovsky, R., 1987. Detection of postnatal depression: de- birth. J. Psychosom. Obstet. Gynaecol. 25 (1), 2334.
velopment of the 10-item Edinburgh Postnatal Depression Scale. Br. J. Psy- Skalkidou, A., Hellgren, C., Comasco, E., Sylvn, S., Sundstrm Poromaa, I., 2012.
chiatry 150, 782786. Biological aspects of postpartum depression. Womens Health 8 (6), 659672.
Dagher, R.K., Shenassa, E.D., 2012. Prenatal health behaviors and postpartum de- Streiner, D.L., 2002. Breaking up is hard to do: the heartbreak of dichotomizing
pression: is there an association? Arch. Womens Ment. Health 15 (1), 3137. continuous data. Can. J. Psychiatry 47 (3), 262265.
Davey, H.L., Tough, S.C., Adair, C.E., Benzies, K.M., 2011. Risk factors for sub-clinical Streiner, D.L., 2003. Unicorns do exist: a tutorial on proving the null hypothesis.
and major postpartum depression among a community cohort of Canadian Can. J. Psychiatry 48 (11), 756761.
women. Matern. Child. Health J. 15 (7), 866875. Tabachnick, B.G., Fidell, L.S., 2014. Using Multivariate Statistics, 6th ed. Pearson,
Dennis, C., 2004. Can we identify mothers at risk for postpartum depression in the London, pp. 500506.
immediate postpartum period using the Edinburgh Postnatal Depression Scale? Thome, M., Alder, E.M., Ramel, A., 2006. A population-based study of exclusive
J. Affect. Disord. 78 (2), 163169. breastfeeding in Icelandic women: is there a relationship with depressive
Dias, C.C., Figueiredo, B., 2015. Breastfeeding and depression: a systematic review of symptoms and parenting stress? Int. J. Nurs. Stud. 43 (1), 1120.
the literature. J. Affect. Disord. 171, 142154. Tuohy, A., McVey, C., 2008. Subscales measuring symptoms of non-specic de-
Figueiredo, B., Canrio, C., Field, T., 2013. Breastfeeding is negatively affected by pression, anhedonia, and anxiety in the Edinburgh Postnatal Depression Scale.
prenatal depression and reduces postpartum depression. Psychol. Med. 3, 110. Br. J. Clin. Psychol. 47 (2), 153169.
Hackshaw, A., 2008. Small studies: Strengths and limitations. Eur. J. Nutr. 32 (5), Vliegen, N., Casalin, S., Luyten, P., 2014. The course of postpartum depression: a
11411143. review of longitudinal studies. Harv. Rev. Psychiatry 22 (1), 122.
Hahn-Holbrook, J., Haselton, M.G., Dunkel Schetter, C., Glynn, L.M., 2013. Does World Health Organization, 2008. Indicators for Assessing Infant and Young Child
breastfeeding offer protection against maternal depressive symptomatology?: a Feeding Practices: Part1 Denitions. World Health Organization, Geneva,
prospective study from pregnancy to 2 years after birth. Arch. Womens Ment. Switzerland.
Health 16 (5), 411422. Ystrom, E., 2012. Breastfeeding cessation and symptoms of anxiety and depression:
Hewitt, C., Gilbody, S., Brealey, S., Paulden, M., Palmer, S., Mann, R., Green, J., a longitudinal cohort study. BMC Pregnancy Childbirth 12, 3641.