Smell perception during early pregnancy: no evidence of an adaptive mechanism

Authors


Dr B. Swallow, Applied Health and Clinical Research Centre, Department of Psychology, University of Lincoln, Brayford Pool, Lincoln, UK.

Abstract

Objective  It has been suggested that nausea and vomiting in pregnancy is an evolutionary adaptive mechanism to avoid the ingestion of potentially harmful foods. It has also been suggested that the mechanism that triggers nausea and vomiting in pregnancy may be olfaction and that olfactory senses are invoked to provide this protection. This study aimed to test this theory in a systematic design.

Design  Cross sectional study.

Setting  The antenatal department of a maternity hospital in the north of England.

Sample  Three groups of participants: pregnant women (n= 55), non-pregnant women (n= 42) and men (n= 48).

Methods  Sensitivity was tested towards the odours of six standard stimuli (half safe and half associated with potentially harmful compounds).

Main outcome measures  Odour rating of likeness, strength and pleasantness.

Results  Pregnant women rated safe and odours with potentially harmful compounds differently but not more so than men or non-pregnant women. There was no evidence that pregnancy changed the olfactory processes from the non-pregnant state and only slight differences between pregnant women and men were recorded.

Conclusions  There was no evidence that olfactory processes had undergone any adaptation during pregnancy. The ability to differentiate safe from potentially harmful compounds was common to all three groups studied.

INTRODUCTION

Nausea and vomiting in pregnancy, together with its more serious form hyperemesis gravidarum, is a major health problem responsible for the loss of an estimated 8.5 million working days per annum in the UK.1

Estimates suggest that between 50%2 and 80%3 of pregnancies are characterised by nausea and vomiting in pregnancy. Estimates of hyperemesis gravidarum range from 0.63%4 to 2%.5 Research suggests cultural differences in the severity and incidence, with low incidence among Eskimos and native African tribes.6 A meta-analysis7 showed that of the 31 societies sampled, 73% experienced nausea and vomiting in pregnancy. The societies that did not suffer ate predominately (although not exclusively) a staple diet of maize.

Research into the aetiology of nausea and vomiting in pregnancy has focussed on biological, social and psychological factors. Increases in human chorionic gonadotrophin (hCG) and hyperthyroidism have been implicated8–10 as well as an association with the bacteria Helicobacter pylori11 and abnormal liver function.12

Psychological factors have focussed on psychodynamic explanations such as repression or unconscious rejection/ambivalence towards the fetus.13–16

More recently, a hypothesis has been suggested which argues that nausea and vomiting may be an evolutionary adaptive mechanism to avoid the ingestion of teratogens and other potentially harmful toxins during pregnancy.17,18

It is argued that pregnancy sickness lowers the woman's threshold for toxin detection.18 Studies have indicated that there is a relationship between the reporting of aversive foods by women and the degree of toxicity in the food, lending support to the sickness as an adaptive response hypothesis (Swallow, unpublished data). Eating foods containing harmful toxins is hypothesised to trigger a response that induces vomiting.

In a comprehensive and rigorous review of the research literature, Flaxman and Shearman19 proposed an expanded ‘maternal and embryo-protection hypothesis’; that nausea and vomiting in pregnancy is a mechanism to enable women to avoid foods that might be dangerous to either themselves or their embryos, especially foods that prior to refrigeration were likely to be laden with micro-organisms or toxins.

Although it has been suggested that taste may be implicated in nausea and vomiting,20 it is proposed that the main mechanism for the detection of toxins may be olfactory.18,21 The adaptive mechanism could have evolved to avoid eating potentially harmful food, and the method of avoiding these may primarily be through smell. It follows that variations in the severity of nausea and vomiting in pregnancy may be related to the woman's sensitivity of smell, with those women suffering from hyperemesis gravidarum being particularly sensitive to smells. Some research has indeed reported that pregnant women find the smell, rather than taste, of certain foodstuffs to be offensive and this may invoke nausea (Swallow, unpublished data). In this study, of those women reporting nausea or vomiting, 68% stated that smell was the trigger, compared with 1% reporting taste. Furthermore, there was a positive relationship between the severity of nausea and vomiting and aversive odours.

Whether pregnant women are more sensitive to odour detection has not been systematically studied.

This study aims to investigate smell perception in an experimental setting using materials that have been standardised. Three groups of participants will be used: pregnant women in their first trimester of pregnancy, non-pregnant women and men. If the olfaction-based evolutionary adaptive theory is tenable, we can predict three hypotheses. First, it can be predicted that pregnant women will have more sensitive olfaction than the other groups. Second, it could be hypothesised that the pregnant group will rate smells of foodstuffs that have potentially harmful substances more negatively than the other groups. Finally, in the pregnant group, we would anticipate a direct relationship between degree of nausea and vomiting and aversions to the smells.

METHODS

The sample size was determined using a 0.05 significance criterion (α), and assuming a medium population effect size, and a power of 80% (1 −β), it was calculated that a sample size of approximately 52 participants per group would be required.22

Pregnant participants were women attending the Hull Maternity Hospital antenatal clinic. The clinic serves a Northern England urban community of which 98% are Caucasian. Booking appointments to the clinic are made to coincide with 10 to 13 weeks of gestation.

Women attending the antenatal clinic at the Hull Maternity Hospital between October 2001 and January 2002 were invited to participate in the study. A total of 76 pregnant women were approached. Of these, 13 were withdrawn due to problems related to their ultrasound scan and 6 declined to participate (reasons were not solicited). Non-pregnant women and men were chosen during the same period. Partners and friends accompanying pregnant women were invited to participate (37 females, 50 males). In addition, non-pregnant participants were recruited from the University through advertisements and personal contact (25 females, 8 males).

In total, 57 pregnant women, 62 non-pregnant women and 58 men participated in the study.

The Nausea and Vomiting in Pregnancy Instrument (NVPI), hedonic rating scales and odour identification, and standard odours were used in this study.

The NVPI2 consists of three questions:

  • 1How often have you felt like being sick (nauseous) in the past week?
  • 2How often have you retched (but without actually being sick) in the past week?
  • 3How often have you been physically sick during the past week?

Scores ranged from 0 (not at all) to 5 (all the time). The items were summed to form a NVPI score ranging from 0 to 15.

The authors designed a series of hedonic rating scales for each of the six odours presented to the participants. For each smell, participants were asked to rate the following on a series of six 14-cm visual analogue scales: pleasant–unpleasant, strong–weak, dislike extremely–like extremely were utilised, and irritating–calming, warm–cool and familiar–unfamiliar were used as distracters.

For each stimuli, participants were also asked to recognise the odour from a list of eight presented to them (the six odours used in the study plus two distracters—beef and tea).

Five of the standard odours were each developed from five liquid chemicals (Appendix 1). A base liquid for each smell was established by combining 0.1 mL of each of the five chemicals with 100 mL methanol. Working solutions for strawberry, vanilla and melon were then made up by dissolving 5 mL base solution with 100 mL water. Dissolving 2 mL of base solutions with 100 mL water made up working solutions for cabbage and fish. From these solutions, 8 mL was poured into a 10-mL clear jar with a screw top. An attempt to make a standard for coffee was unsuccessful. Following piloting on a non-clinical sample, it was found that identification was unreliable. Coffee was therefore produced by dissolving three granules of coffee (Café Direct) in 8 mL water.

All finished smells were clear in colour except coffee that had a slight brown tint.

Coffee was chosen because in earlier research (Swallow, unpublished data), 19% of pregnant women stated an aversion. Similarly, fish and other fried food was considered aversive by over 28% of the sample. Cabbage was selected as it has known quantities of toxin allyl isothiocyanate.19,23 These odours were also identified in earlier work as containing potentially harmful compounds.21 The ‘safe’ odours (strawberry, melon and vanilla) were selected because there is no literature relating to potential harmful compounds and participants in earlier research did not highlight them as aversive.

Local approval to perform the study was granted by the Hull and East Riding Research Ethics Committee and from the Hospital Trust Research and Development Consortium Quality Group.

Each of the six odours was placed in separate rooms to avoid contamination. Participants were escorted from one room to another and given the following instructions:

Please take a good sniff of the jar. Once you have started rating the smell, please do not go back to it to have another sniff.

This was repeated for each odour. Pregnant participants were then presented with the NVPI, and finally all participants were asked to complete a questionnaire with demographic information.

Odours were replaced after every 10 participants by fresh samples.

Using SPSS for Windows, differences between groups were analysed using appropriate analysis of variance. Relationships were established using Pearson's correlation coefficient.

RESULTS

Most pregnant women fell in the age categories below 35. In order to control for age differences in olfaction, analysis between groups were confined to participants aged less than 35, reducing the sample sizes to 55 pregnant women, 42 non-pregnant women and 48 men. χ2 showed no significant age differences between the three groups [χ2(6) = 9.2, ns].

The mean gestational age of the pregnant women was 11.9 weeks (SD = 1.3); 24 (44%) were primigravid; and 24 (44%) reported having children. Fourteen (33%) of the non-pregnant group and 14 (29%) of the male group reported having children.

To test the hypothesis that pregnant women would be more sensitive to odours, multiple analyses of variances were conducted to investigate differences in pleasantness, strength and liking of each of the six odours. Results are indicated inTables 1–3.

Table 1.  Mean pleasantness ratings (and standard deviations) for all smells (low scores = pleasant).
 Pregnant (n= 55)Non-pregnant (n= 42)Men (n= 48)Multiple analysis of variance
  • *

    P < 0.05.

Strawberry6.1 (3.0)5.6 (3.1)6.1 (2.7)0.46
Vanilla3.9 (2.6)3.7 (2.1)3.8 (2.3)0.15
Melon8.6 (4.0)7.4 (3.7)7.8 (3.8)1.88
Coffee6.7 (3.3)5.8 (3.2)5.3 (3.2)2.3
Cabbage11.5 (1.8)11.6 (1.6)10.9 (2.0)1.7
Fish10.5 (3.0)11.0 (2.5)10.1 (2.9)1.2
Total47.4 (7.5)45.1 (6.4)43.6 (7.5)3.5*
Table 2.  Mean strength ratings (and standard deviations) for all smells (low scores = strong).
 Pregnant (n= 55)Non-pregnant (n= 42)Men (n = 48)Multiple analysis of variance
  • *

    P < 0.05.

Strawberry8.5 (3.9)8.1 (3.5)8.1 (2.9)0.22
Vanilla8.1 (3.3)8.0 (3.2)7.3 (3.1)1.06
Melon3.0 (2.4)4.3 (2.8)4.0 (2.6)3.72*
Coffee5.3 (4.0)6.1 (3.9)6.2 (3.5)0.78
Cabbage1.5 (1.2)1.7 (1.4)2.1 (1.6)2.39
Fish2.1 (1.5)2.2 (2.6)2.4 (1.8)0.28
Total28.5 (7.0)30.5 (7.9)30.0 (7.6)0.97
Table 3.  Mean liking ratings (and standard deviations) for all smells (low scores = dislike).
 Pregnant (n= 55)Non-pregnant (n= 42)Men (n= 48)Multiple analysis of variance
  • *

    P < 0.05.

  • **

    P < 0.01.

Strawberry6.9 (3.2)7.2 (2.8)6.9 (2.4)0.15
Vanilla8.7 (2.8)8.9 (2.1)9.2 (2.3)0.57
Melon4.7 (3.5)5.9 (3.7)6.4 (3.3)3.17*
Coffee6.4 (3.3)7.7 (3.1)8.0 (3.3)3.37*
Cabbage2.3 (2.2)2.1 (2.2)3.1 (2.4)2.14
Fish2.8 (2.7)2.0 (1.7)3.5 (2.9)4.21*
Total31.9 (7.9)33.8 (7.1)37.0 (7.4)6.03**

Table 1 shows that overall pregnant women rated the smells as more unpleasant [F(2,142) = 3.5; P < 0.05] but there were no differences in ratings of individual smells. Tukey's Post Hoc test showed this difference was between the pregnant group and the men.Table 2 indicates no overall differences in ratings of strength. However, pregnant women rated melon as stronger than the other two groups [F(2,142) = 3.7; P < 0.05]. Tukey's Post Hoc test showed the difference was between the pregnant and non-pregnant women groups.Table 3 indicates that overall, pregnant women disliked the smells more than the other two groups [F(2,142) = 6.06; P < 0.01]. In particular, pregnant women disliked melon [F(2,142) = 3.17; P < 0.05], coffee [F(2,142) = 3.37; P < 0.05] and fish [F(2,142) = 4.2; P < 0.05]. Tukey's Post Hoc tests showed that all these differences were between the pregnant group and the men, except for the fish smell where the difference lay between the non-pregnant group and the men.

Analyses were also conducted to investigate differences in perceived irritation, warmth and familiarity. There were no significant results for any of the odours. Finally, χ2 analyses were conducted to investigate differences in frequency of correctly identifying odours. There were no significant differences between groups.

In order to test the hypothesis that women would show more aversions to foodstuffs containing potentially harmful compounds, mixed design analyses of variance were conducted. For this analysis, ratings for the three odours associated with potentially harmful compounds (coffee, cabbage and fish) were combined and compared with those without known compounds (strawberry, vanilla and melon) for three of the six rating scales (pleasantness, strength and liking).

Figure 1 shows results for ratings of pleasantness. There was a significant difference in the combined pleasantness ratings between the groups [F(2,142) = 3.55; P < 0.05]; the post hoc test showing the difference was between the male and pregnant groups. There was also a significant difference between safe and potentially harmful compounds [F(1,142) = 254.3; P < 0.001]. However, there was no significant interaction between safety and group [F(2,142) = 1.3; ns], indicating that pregnant women did not rate the safe smells as more pleasant then either of the other groups.

Figure 1.

Ratings of pleasantness for odours for safe and potentially harmful compounds. ▪------▪ non-pregnant women; ▴—-—▴ men; □——□ pregnant women.

Figure 2 shows results for ratings of strength of odours. There was no significant difference between the groups in the combined rating of strength [F(2,142) = 0.97; ns], but a significant safety effect [F(1,142) = 244; P < 0.001]. There was no significant interaction between the groups [F (2,142) = 1; ns].

Figure 2.

Ratings of strength of odours for safe and potentially harmful compounds. ▪-----▪ non-pregnant women; ▴—-—▴ men; □——□ pregnant women.

Figure 3 shows strength of liking of odours. There were significant differences in the combined likeness rating between the groups [F(2,142) = 6.02; P < 0.01] and also safety [F(1,142) = 209; P < 0.001]. There was no interaction [F(2,142) = 1.1; ns] between the groups.

Figure 3.

Likeness ratings for safe and potentially harmful compounds. ▪-----▪ non-pregnant women; ▴—-—▴ men; □——□ pregnant women.

Relationship between nausea and vomiting in pregnancy intensity and odour sensitivity was analysed by Pearson's correlation coefficient, which was calculated with intensity of nausea and vomiting in pregnancy, as measured by the NVPI, pleasantness, strength and liking of each of the odours. There were no significant relationships between NVPI scores and odour pleasantness, liking or intensity.

In summary, pregnant women did not demonstrate a consistent difference in their olfactory function compared with men or non-pregnant women. The ability to recognise potentially harmful compounds as unfavourable was not different from men or non-pregnant women.

DISCUSSION

This is the first systematic study of pregnant women's sensitivity to differentiate and categorise odours. Pregnant women rated safe and potentially harmful compounds differently but not more so than men or non-pregnant women. There was no evidence that pregnancy changed the olfactory processes from the non-pregnant state and only slight differences between pregnant women and men were recorded.

In order to lend support to the theory that nausea and vomiting in pregnancy is an evolutionary adaptive mechanism, it was necessary to support three hypotheses. The first, that pregnant women will have more sensitive olfaction than the other groups was not fully supported. Pregnant women rated smells as less pleasant than men, but not non-pregnant women. In terms of strength of odours, the only stimulus rated as stronger by pregnant women was melon, and this was significantly stronger than the non-pregnant group. Pregnant women also significantly disliked the smells more than men, but not non-pregnant women.

The second hypothesis, that the pregnant group will rate smells of foodstuffs that have potentially harmful substances more negatively, was also unsupported. There was no effect seen between the three groups (pregnant, non-pregnant and men) and safety for overall pleasantness, strength or liking of smells.

The third hypothesis, that in the pregnant group there would be a relationship between degree of nausea and vomiting and aversions to the smells, was also unsupported. Analysis was conducted to investigate NVPI scores and pleasantness, strength and liking for all stimuli and no significant correlations emerged.

These results do not support findings from previous research.19 The current authors found 68% of women reported smell as the trigger to their vomiting (Swallow et al., unpublished data). Most past research has focussed on taste, rather than odour and it may be that for the aversion to be significant the pregnant woman must taste and/or see the food as well as smell it. It is possible that odour could be a by-product of the aversion, acquired through conditioning.

Alternatively, it is possible that the actual potentially harmful compound contained in the foodstuff may be the trigger. The strength of this study in precisely controlling the stimuli may also have proved to be a weakness. The stimuli used in this experiment did not, for ethical reasons, include any of the potentially harmful compounds found in the natural substances associated with the stimuli. Perhaps a repetition of this study could be undertaken using a range of actual foodstuffs, some containing known harmful toxins.

It is also possible that it is the smell per se that is responsible, rather than the woman's sensitivity to smell. This study examined sensitivity using pleasantness, strength and liking. However, there may be another olfactory process not linked to sensitivity that affects vomiting.

Finally, it is possible that the range of stimuli presented to the participants lacked sufficient sensitivity to obtain a significant effect. The data show a non-significant tendency for pregnant women to be more sensitive to odours. Perhaps different odours, especially those containing frying fat, would yield different results. In earlier work (Swallow, unpublished data), the smell of fat and frying meat was a significantly reported aversive stimulus.

It is an attractive hypothesis to consider the possibility that nausea and vomiting in pregnancy has a role in the protection of the developing fetus, as a number of foodstuffs contain potentially harmful compounds. For example, celery and parsnips contain the mutagen and carcinogen 8-methoxypsoralen18 and also linear furanocoumarins that may be associated with dermatitis23; cabbage, cauliflower, sprouts, broccoli and mustard contain the toxin allyl isothiocyanate19,23; and cocoa and nutmeg contain safrole.18 Potatoes, tomatoes and aubergines contain teratogens (glycoalkoloids). Cranberries contain a considerable amount of quercetin, and other fruits such as blueberries and thornless blackberries contain small amounts of this mutagen.23 Common mushrooms contain agaritine that has been associated with a risk of tumour development in mice.23 Meat and dairy produce, which is not fresh, contain potentially harmful bacteria. Frying and burning food changes the chemical composition to produce, among other toxins, mutagens.18 Mytocoxins are toxic compounds produced by fungi on grains and other foods and some are potent carcinogens (e.g. aflatoxin B1). Corn, milk and peanuts may contain them.23

Our results concur with those of Humel et al.24 in that nausea and vomiting in pregnant patients was not related to olfactory sensitivity and provide additional evidence that pregnancy is not a state of increased olfactory sensitivity.

In conclusion, this research has failed to support the evolutionary adaptive hypothesis that pregnant women have olfactory processes different from non-pregnant women or men. However, repetition is required using a different range of natural and synthetic stimuli before dismissing odour as an important factor in nausea and vomiting in pregnancy.

Acknowledgements

The authors would like to acknowledge the following people who have made this study possible: midwifery staff from the antenatal clinic of Hull Maternity Hospital, technical support at the University of Lincoln, the obstetricians who allowed their patients to be studied and the patients and volunteers who participated.

Accepted 8 June 2004

Appendix

Appendix 1 Chemical composition of odours

StrawberryVanillaMelonCabbageFish
Methyl propionateAnisyl propionate2,6-Dimethyl-5-heptanalMethyl sulphide (dimethyl sulphide)Trimethylamine
Phenethyl butyrateBenzoinCis-6-nonenal4-(methylthio) butanolPyridine
3-Phenylpropyl isovalerateCinnamaldehydeMethyl 3-nanenoatePropyl mercaptanButylamine
Piperonyl acetateEthyl vanilinCis-6-nonen-l-olDimethyl disulfideFurfuryl mercaptan
Maltyl isobutyrateVeratraldehydeCis-2-nonen-l-olDimethyl trisulphidePhenethylamine

Ancillary