1. Dr. Jennifer Peck University of Oklahoma Health Sciences Center
Systematic review of the potential adverse effects of caffeine consumption in healthy adults, pregnant women, adolescents, and children: Reproduction and Development Outcomes
Dr. Jennifer Peck
University of Oklahoma Health Sciences Center

2. Disclosure
Manuscript was published in 2010 in Food and Chemical Toxicology.
Member of the ILSI North America Caffeine Systematic Review Team 
ILSI North America provided an honorarium and travel funds to attend this meeting 
Scientific Advisor for the ILSI North America Caffeine Working Group
Prior consultant to ILSI North American Caffeine Working Group for objective review of the evidence for reproductive health effects of caffeine consumption

3. PECO Questions for Reproduction and Development
For [population], is caffeine intake above [dose], compared to intakes [dose] or less, associated with adverse effects on reproduction and development outcomes?
For the evaluation of reproductive and developmental outcomes, our research question followed the PECO format, as it did for all other outcomes assessed, where we specified the population of interest, the caffeine exposure of interest, the comparator as defined by the Nawrot publication and the outcome or endpoint under consideration.
For reproduction and development, the components of the PECO question differ somewhat from that of other outcomes in that we consider 2 populations for comparison to the Nawrot conclusions.
While all studies involved adult populations, the majority of studies involved the assessment of caffeine intake in pregnant women, where we applied the corresponding Nawrot comparator of < 300 mg caffeine/day.
For the few studies evaluating non-pregnant women or men, such as those evaluating fecundity outcomes including male semen parameters, we applied the Nawrot comparator of <400 mg/day for healthy adults.
So, our review of the evidence for reproduction and development consisted of 9 different PECO questions that specifically addressed each endpoint of interest and the specific population and comparator for this question.
As one example, our PECO question for fetal growth endpoints was stated as…
For healthy pregnant women, is caffeine intake above 300 mg/day, compared to intakes of 300 mg/day or less, associated with adverse effects on fetal growth?
Population
Healthy Adults
Healthy Pregnant Women
Exposure
> 400 mg/day
> 300 mg/day*
Comparator
≤ 400 mg/day
≤ 300 mg/day*
Outcome
Adverse reproductive or developmental effects
Example: For healthy pregnant women, is caffeine intake above 300 mg/day, compared to intakes of 300 mg/day or less, associated with adverse effects on fetal growth?

4. Characterization of Data: Reproduction and Development Outcomes
Data Type
Characterization
Number of studies
58 included; 36 excluded
Study types
Mostly observational studies (cohort and case-control)
Only three randomized controlled trials
Populations
Adults; Pregnant Women
Exposure
Primarily self-reported intake (e.g., food frequency questionnaires) of coffee, tea, soda, and chocolate
Categorical groupings (e.g., <1, 1-3, >3 cups/day caffeinated beverage)
Most exposures studied were below the comparator
Outcome (Endpoints)
Fecundability and infertility, spontaneous abortion, recurrent miscarriage, stillbirth, preterm birth, fetal growth, birth defects, childhood behavior, childhood cancers, and “other” endpoints
A total of 94 full text papers were evaluated for reproductive and developmental outcomes.
Of these, 36 did not meet the criteria for inclusion in the Systematic Review. Exclusions were primarily due to the absence of quantitative measurements of caffeine exposure (such as binary assessment of any vs none), the assessment of caffeine metabolites or presentation of data that did not allow for assessment of caffeine intake values relative to the PECO comparator.
Most studies for this outcome were observational designs, either cohort or case-control studies. However, 3 studies were randomized controlled trials.
Most studies were specifically designed to evaluate the effects of caffeine, typically using data collection instruments such as food frequency questionnaires to collect self-reported servings of caffeine-containing beverages or foods, primarily coffee, tea, soda. Chocolate intake was included in a number of studies and caffeine-containing medications and energy drinks were evaluated in only a few studies.
Number of reported servings were either converted by the authors to daily milligrams of caffeine consumed per day or the reported as cups or servings per day, were converted by the research team using standardized methods.
Caffeine intake was commonly grouped into categories, so most studies reported findings for defined ranges of daily intake.
Most of the caffeine exposures assessed in these studies were below the 300 mg/day comparator for pregnant women. In fact, the highest intake level assessed was at or near 300 mg/day for approximately half of the studies evaluated. Under 10% evaluated upper intake levels below the 300 mg/day comparator for pregnant women.
The remaining 1/3 evaluated intake levels that exceeded the 300 mg/day comparator, with the highest level evaluated in a single study reported as >1000 mg/day.
The assessment of reproductive and developmental outcomes addressed 9 endpoints in total. These included measures of fecundability and infertility, spontaneous abortion, recurrent miscarriage, stillbirth, preterm birth, fetal growth, birth defects, childhood behavior, and childhood cancers.
A group of other endpoints were combined for consideration as an “other category” when only one study was identified for specific health endpoints. This other group included only 2 studies that addressed pregnancy-induced hypertension and preeclampsia, and markers of maternal stress.

5. Fecundability and Fertility
I’d like to highlight results for 5 selected endpoints, beginning with data for the endpoints of fecundability and fertility, which were evaluated in healthy non-pregnant adults using the Nawrot comparator of 400 mg/day.
These plots provide a visual display of selected results from each study reflecting the lowest observed effect levels (i.e., LOELs) in orange, or no effect levels (i.e., NOELs) in blue, as compared to the Nawrot comparator reflected by the vertical line. As noted previously, ALL results from each study are not displayed in these figures, but when multiple results were presented, the findings from the most refined analysis performed by the authors was selected for display.
Only 2 studies meeting the inclusion criteria for this Systematic Review assessed fecundability – by examining time to pregnancy.
4 studies evaluated male reproductive parameters by assessing semen quality, sperm DNA damage or salivary testosterone levels in athletes.
2 additional studies of fertility endpoints addressed infertility due to ovulation disorders and age at menopause.
Overall, the findings were generally consistent indicating lack of effects below or above the 400 mg/day comparator. This consistency along with an evaluation of low risk of bias (and low level of indirectness) increased confidence in the body of evidence for these endpoints.
Taken together, we concluded that there is a moderate to high level of confidence that the comparator of 400 mg caffeine/day in healthy adults is acceptable for fertility, fecundability, and male reproductive endpoints.
Conclusion: Comparator of 400 mg caffeine/day in healthy adults is an acceptable intake for fertility, fecundability and male reproductive endpoints

6. Spontaneous Abortion
Among studies of spontaneous abortion, there were a few data points that indicated observed effects at levels below the 300 mg/day comparator, although these were limited two studies where one reported observed effects only among subpopulations with certain phenotypes or genotypes for caffeine metabolism.
There were also other studies reporting lack of effects following consumption as high as >500 mg/day, so Confidence in the overall body of evidence for spontaneous abortion is decreased by the lack of consistency in the direction of findings both above and below the comparator.
Confidence in this evidence is further limited by the inability to fully account for confounding by the pregnancy signal, which is a critical concern that I will address further as I discuss evaluation of study quality in more detail
So considering the totality of the evidence, there is a moderate level of confidence in the body of evidence that the comparator of 300 mg/day is acceptable as an intake that would not be associated with significant concern of spontaneous abortion.
__________________________________________________
NOTES:
CYP1A2 phenotype for metabolic activity; NAT2 genotype for acetylation as markers of caffeine clearance),
An overall low risk of bias (Figure 10) and a low level of indirectness increase the overall confidence of the studies included on spontaneous abortion. Consideration of the magnitude of effect neither strengthened nor lessened confidence in the body of evidence given that in the two studies reporting effects below the comparator, the magnitude was low (OR or RR between 2.0 and 2.4), and for one study effects were only found in subgroups. The overall level of confidence in the body of evidence which supports the conclusion remained moderate,
primarily due to the strong potential confounding effects of the pregnancy signal and inconsistency of findings.
Conclusion: Comparator of 300mg caffeine/day in pregnant women is acceptable as an intake that would not be associated with significant concern for spontaneous abortion

7. Preterm Birth
Studies of gestational age and preterm birth were considered together for the purposes of this Systematic Review.
Of the 5 studies were identified for this endpoint – 1 was a RCT among coffee drinkers who were allowed to consume caffeine from desired sources (Bech 2007) and 1 was a systematic review (Maslova 2010).
No associations between preterm birth and maternal caffeine consumptions were reported for any intake levels at or above the comparator.
The consistent lack of effects leads to moderate to high confidence in the body of evidence that the comparator of 300mg caffeine/day in pregnant women is an acceptable intake level that would not be associated with significant concern for preterm delivery or adverse effects on gestational age.
______________________________________________
NOTE: 3 were prospective cohort studies
Conclusion: Comparator of 300mg caffeine/day in pregnant women is acceptable as an intake that would not be associated with significant concern for preterm delivery

8. Fetal Growth
The body of evidence for fetal growth was more difficult to integrate based on the inconsistent findings.
Of the 14 studies evaluating endpoints including birth weight, small for gestational age and intrauterine growth restriction, 9 reported no observed effects of maternal caffeine consumption below the comparator of 300 mg/day.
While the remaining 4 studies and 1 meta-analysis reported adverse effects on fetal growth at intake levels below the comparator.
Refined conclusions were more challenging for fetal growth given that all comparison points below the comparator indicated observed effects (suggesting the comparator is too high),
whereas the majority of comparison points at or above the comparator reported a lack of effects (suggesting the comparator was acceptable or too low).
Confidence was reduced in the data suggesting the comparator was too high, however, when considering the low magnitude of the observed effects (RR between 1 and 2 ) —as well as the observation that, in many cases, effects were limited to measures at a single time point or did not reflect clinically relevant changes
Ultimately, the biological significance of changes in birth weight were considered to be more robustly evaluated in studies evaluating SGA or IUGR, which as a whole, did not provide support for effects below the comparator.
Across all studies, confidence in the body of evidence is moderate to high that effects below the comparator of 300mg caffeine/day is acceptable as an intake that would not be associated with significant concern for adverse effects on fetal growth including endpoints of small for gestational age or intrauterine growth restriction..
Conclusion: Comparator of 300mg caffeine/day in pregnant women is acceptable as an intake that would not be associated with significant concern for fetal growth

9. Birth Defects
For birth defects, the evidence base is broad with many types of birth defects evaluated across the 11 studies with few studies reported for any single type of malformation
The majority of data for birth defects were relatively consistent in demonstrating a lack of effects following consumption of caffeine at intakes up to 300 mg/day in healthy pregnant women.
Two studies from the same case-control study population reported weak to moderate effects below the comparator for limb defects and ano-rectal atresia.
While another study reported an association with neural tube defects with intake below the comparator, but
none of these observed a dose response and the study designs were susceptible to recall bias.
Thus, there is a moderate level of confidence in the conclusion that the 300mg/day comparator is acceptable as an intake that would not be associated with significant concern for birth defects.
Conclusion: Comparator of 300mg caffeine/day in pregnant women is acceptable as an intake that would not be associated with significant concern for birth defects

10. Evaluation of Individual Study Quality (Risk of Bias)
As the spectrum of study results were recorded and considered in relation to the comparator, each individual study was also evaluated for study quality according to the 10 domains specified in the OHAT risk of bias tool [developed by the National Toxicology Program, Office of Health Assessment and Translation (OHAT)].
The domains addressed specific elements of study design such as accounting for confounding, missing outcome data and quality of exposure and outcome measurements.
This graphic provides a snapshot of the risk of bias heat map that was developed to display the results of this assessment.
As described in earlier presentations, each component was rated on a scale ranging from “definitely or probably low” risk of bias to “definitely or probably high” risk of bias. Low risk of bias is indicated by shades of green and high risk of bias is marked by shades of pink and red.
The white boxes reflect no assessment where the domain was not applicable to the study design – for instance, where domains addressing experimental components of randomization and allocation concealment were not be applied to observational study designs.
Across all domains, studies of reproduction and development were determined to have an overall low risk of bias as demonstrated by the majority of green shading in this section of the risk of bias heat map, with few individual studies falling at either end of the spectrum.
______________________________________________
In general, when studies were downgraded into the pink and red categories of probable or definite high risk of bias, it was commonly the result of two primary study design limitations :
The first relates to methods used to address confounding by the pregnancy signal
This has primarily been recognized as a concern for studies of spontaneous abortion, but it also applies to studies of other endpoints including stillbirth, preterm delivery and fetal growth.
The phenomenon of the “pregnancy signal” relates to research that has shown that pregnancy symptoms such as nausea, vomiting and aversions to tastes and smells are associated with favorable pregnancy outcomes and reduced risk for pregnancy loss. The specific mechanism for this association between symptoms and pregnancy viability is not understood but is proposed to be due to increasing levels of pregnancy hormones or human chorionic gonadotropins (hCG) which are produced at increasing levels by a well-developing pregnancy.
Women who experience pregnancy symptoms have also been shown to reduce their caffeine intake in response to aversions to strong smells and tastes such as coffee. So, women who do not have healthy developing pregnancies may continue consuming caffeine at higher levels because they are not experiencing the same symptoms and aversions. When these groups are compared without accounting for confounding by the pregnancy signal, then it could appear that those with higher caffeine consumption had an increased risk of adverse pregnancy outcome when, in fact, intake levels were actually a consequence of an unhealthy pregnancy.
Although most studies that investigated fetal loss did evaluate nausea and/or vomiting as a potential confounder, the measures used to account for the presence or frequency of nausea may not provide a valid assessment of the relevant aspects of the pregnancy signal such as aversions to tastes and smells that may be more directly influencing caffeine consumption patterns as a consequence of the health status of the pregnancy.
Because of the known complexities and limitations of methods that attempt to account for pregnancy signal, these studies were assigned a rating of probably high risk of bias with regard to the confounding domain. But, if authors did not attempt to evaluate pregnancy signal, the study was scored as high risk of bias for this element (Q4).
2) In addition, study ratings were also uniformly impacted by the known difficulties of accurate caffeine exposure assessment, which is reflected by the column of red on the right side of the figure.
It is well-appreciated that exposure misclassification is considered to be a major source of potential bias in observational studies of caffeine intake. This is due not only to the issue of potential reporting errors that may arise from self-reported intake, but it’s also due to the wide variety of caffeine sources and specifically the variability in the amount of caffeine contained even within the same type of product. In studies of pregnancy outcomes, there are also concerns regarding the timing of exposure assessment which may not correspond to the etiologically relevant window of exposure for the pregnancy outcome under investigation or capture the changing patterns of intake that occur during pregnancy.
Related to this concern was the issue of potential recall bias, which can occur when there are differences in reporting accuracy between cases and controls. This was a limitation of particular concern for case-control studies of rare outcomes such as childhood cancers and birth defects.
Overall, these challenges lead to uncertainties in exposure measurement that can limit the evidence base.
Overall low risk of bias
Common study limitations:
Difficult to fully account for pregnancy signal (Q4 Confounding)
Caffeine exposure assessment (Q8 and Q11)
+2: Definitely low
+1: Probably low
-1: Probably high
-2: Definitely high

11. Factors that Increased or Decreased Confidence in the Body of Evidence by Endpoint
No. of Studies
Initial Confidence Rating
Based on study type and study features
(OHAT, 2015)
Overall RoB
Domain-based evaluation of risk of bias per OHAT RoB tool
Indirectness
Was the study designed to evaluate the PECO?
Magnitude
Strength of effect (when effect observed below the comparator)
Confounding
Were plausible confounders that would change the observed effect accounted for?
Consistency Were findings consistent in demonstrating effects or lack of effects at or below the comparator
Final Confidence Rating
What is the overall rating when factors that increase or decrease confidence were considered
Fertility, fecundability, and male reproductive measures 7
Moderate ↑ —
Moderate to high
Spontaneous abortion 8
↓/- ↓
Recurrent miscarriage 4
↑/-
Stillbirth
Preterm birth and gestational age 5
Fetal growth 14
Birth defects 11
For each endpoint assessed within the outcome of reproduction and development, we went through a systematic process to qualitatively evaluate our confidence in the body of evidence using the framework established by the IOM, complimented by the OHAT guidelines.
These confidence ratings were used to characterize the data and to provide a structure to develop conclusions by endpoint based on the strength of the underlying evidence.
So Endpoints where the number of studies were too limited to develop conclusion ( such as those with < 5 studies) were also not included in the summary of confidence ratings – so this excluded studies of childhood cancers, child behaviors and single studies of other reproductive endpoints.
For the remaining 7 endpoints, we developed initial confidence ratings based on elements of the study design . The initial confidence rating was moderate for all endpoints based on the criteria that the studies used comparison groups, collected individual-level data and assessed exposure prior to the outcome.
Then, using expert judgement, a number of additional factors were considered, which ultimately either increased, decreased or did not change confidence in the evidence for each endpoint.
As reflected by the upward arrows in the 4th and 5th columns, the general evaluation of low risk of bias and low indirectness (which reflects external validity) increased confidence in the body of evidence for most endpoints.
When considering the remaining factors, evaluation of the magnitude of observed effects - which focused on studies reporting effects below the comparator – decreased confidence in studies of spontaneous abortion and fetal growth due to moderate effects (OR>2.0)reported in some studies for caffeine levels below the comparator.
Confidence in the body of evidence for spontaneous abortion and fetal growth was also decreased due to the difficulties accounting for confounding by the pregnancy signal.
Inconsistency of results also decreased confidence in the evidence for spontaneous abortion but patterns of consistent results increased confidence in the body of evidence for fertility and fecundability, recurrent miscarriage, preterm birth and gestational age and birth defects.
In total, overall confidence in the body of evidence remained moderate for the endpoints of spontaneous abortion, stillbirth and birth defects and increased to moderate to high for fertility and fecundability, recurrent miscarriage, preterm birth and gestational age and fetal growth.
↑ increased confidence no change to confidence ↓ decreased confidence

12. Developing Conclusions for the Outcome: Integration of Weight of Evidence Considerations
This table provides a summary of the process that the review team used to integrate the weight of evidence for each endpoint and establish conclusions for this systematic review.
The purpose of displaying this table is not to review the many details, but to highlight the various considerations that informed our conclusions.
Ultimately, the conclusion for the reproductive and developmental outcome was informed by a synthesis of the body of evidence that considered the spectrum and volume of data above and below the comparator as well as the evaluation of the quality of the evidence and consideration of type of effect (clinical or physiological) and level of adversity.
For reproductive and developmental outcomes, all endpoints were classified as clinical effects with high adverseness or importance for decision making so there was little variation in this regard.
When considering the range of effects, there was substantial overlap within the body of evidence for the reported ranges of No Observed Effect Levels and Low Observed Effect Levels. This was the case for all endpoints examined, except preterm birth/gestational age and child behavior, where effects were not observed at any level examined.
If we focus on the area with the green shading in the middle of the table…
The X’s indicate that for most endpoints, findings were reported across the entire spectrum of results.
That is, there was at least one more data points that reported no observed effects at intakes below the comparator, which indicates the comparator of 300 mg caffine/day in healthy pregnant women (or 400 mg/day in healthy adults) is acceptable. This is reflected by the X in the middle column.
Similarly, the X’s on the right side column indicate that no observed effects were reported at intakes above the comparator, suggesting the comparator may be too low.
The X’s in the left column indicate that effects were observed at intakes below the comparator, suggesting the comparator may be too high. Most endpoints, however, had few studies with such findings. Although effects below the comparator cannot be ruled out with the currently available data, the effects observed below the comparator were primarily reported for spontaneous abortion and fetal growth.
Considering the totality of the evidence,
The green shading represents the conclusion for each endpoint relative to the comparator. The overlapping green down the middle corresponds with the conclusion that caffeine intake at levels below the comparator is acceptable for all endpoints examined.
Note, however, that for certain endpoints uncertainties are acknowledged by the shift in both directions. For preterm delivery we acknowledge the possibility that the comparator may be too high given the consistent reports of no observed effects at levels at or above the 300 mg/day comparator.
In contrast, for spontaneous abortion we acknowledge that the comparator may be too low for certain subpopulations where effects were observed below the comparator for some metabolic phenotypes. Although other studies reported lack of effect at intakes greater than 500 ng/day.
For fetal growth, effects on birth weight were also reported at intakes below the comparator suggesting that the comparator may be too low; but our conclusions were based on the biological significance of such an effect which is more robustly evaluated in studies evaluating SGA or IUGR, which as a whole did not show effects below the comparator.

13. WoE Conclusion: Reproduction and Development
When the weight of evidence was considered for pregnant women, we concluded that an intake of 300 mg caffeine/day was generally without significant concern regarding overt adverse effects on reproduction and development.
The majority of studies included in this SR do not report reproductive or developmental effects at intake levels below 300 mg caffeine /day for healthy pregnant women
Effects below the comparator, however, cannot be ruled out with the currently available data,
For some endpoints, including fetal growth and spontaneous abortion, evidence indicated that intake at the level of the comparator could possibly be high for some subgroups or outcomes;
However, uncertainties in the evidence base were encountered. These included considerations for the biological relevance of findings regarding modest changes in birth weight in the context of more limited support for effects on small for gestational age at intakes below the comparator.
For both endpoints, there are also uncertainties in the body of evidence due to the limitations of methods used to control for the pregnancy signal
Overall, There is a moderate level of confidence in the evidence base which supports this conclusion.
Pregnant Women:
When the weight of evidence was considered, 300 mg caffeine/day was generally without significant concern for associations with adverse reproductive and developmental effects
For some endpoints (fetal growth and spontaneous abortion) comparator may be high, but evidence base contains uncertainties
Difficulties in evaluating this evidence arose from:
Consideration of biological relevance (birth weight changes)
Inability to fully account for the pregnancy signal

14. WoE Conclusion: Reproduction and Development
When the weight of evidence was considered for healthy adults, an intake of 400 mg/day is generally without significant concern regarding overt adverse effects on fecundability and fertility as measured by time to pregnancy, male reproductive parameters, ovulatory infertility and age at menopause.
There is a moderate to high level of confidence in the evidence base which supports this conclusion.
As a result, the current body of evidence for reproductive and developmental endpoints is generally consistent with what was reported by Nawrot et al. (2003) and other reviews of caffeine and reproductive and developmental outcomes;
As such, the current guidance for pregnant women is supported by the findings of this systematic review
NOTE:
IFIC (2008) – International Food Information Council,
EFSA (2015) – European Food Safety Authority Scientific Opinion).
Healthy Adults:
The comparator of 400 mg/day of caffeine was found to be an acceptable intake level that was not associated with concern for fertility and fecundability endpoints:
Time to Pregnancy
Male Reproductive Parameters
Ovulatory Infertility
Age at Menopause

15. Future Research Considerations
For future research considerations, we conclude that:
Further evaluation of potential effects of maternal caffeine intake on childhood cancer is warranted due to the limited number of existing studies with inconsistent results. Because the data were limited to only 3 studies of childhood cancer, a conclusion for this endpoint was not developed in this systematic review.
However, for the 3 case-control studies of childhood leukemia and Central Nervous System (CNS) tumors that were identified, the use of healthy control groups raised concerns about potential recall bias, reducing confidence in reported associations.
Studies are needed that incorporate techniques to assess or minimize potential recall bias when studying rare events such as childhood cancers. By nesting case-control studies within existing prospective cohort studies or by incorporating control groups that include those with other childhood conditions unrelated to exposure, future studies can strengthen this body of evidence.
Improved assessment of the pregnancy signal
Future efforts to disentangle the complex relationship between the pregnancy symptoms, changes in caffeine intake during pregnancy and pregnancy viability will benefit from improved, prospective measurement of all relevant dimensions of the pregnancy signal including aversions to caffeine-containing products such as coffee and the timing of symptoms in relation to the relevant exposure period. Studies are also needed to clarify the importance of the pregnancy signal for all relevant endpoints including fetal loss as well as fetal growth and congenital malformations.
Identify sensitive sub-populations
It is suggested that some subgroups exhibiting differences in caffeine metabolism may be more susceptible to adverse reproductive and developmental outcomes. This is an area that warrant further research.
Optimize exposure assessment to minimize measurement error (major challenge for observational study designs)
An ongoing challenge for observational studies of caffeine intake is accurate measurement of exposure. Studies should continue to optimize exposure assessment with attention to capturing the changing patterns of intake that occur during pregnancy.
Studies that incorporate ‘quantitative bias analyses” to evaluate the magnitude and direction of bias introduced by exposure measurement error based on sensitivity and specificity obtained from validation studies can also serve to advance the
Further evaluation of prenatal caffeine exposure and childhood cancer
Improve assessment of the pregnancy signal
Identify sensitive subpopulations
Optimize exposure assessment to minimize measurement error (major challenge for observational study designs)