A systematic review on the impact of gestational Lyme disease in humans on the fetus and newborn

Lyme disease (LD), caused by bacteria of the Borrelia burgdorferi sensu lato species complex, is the most common vector-borne disease in North America and Europe. A systematic review (SR) was conducted to summarize the global literature on adverse birth outcomes associated with gestational LD in humans. The SR followed an a priori protocol of pretested screening, risk of bias, and data extraction forms. Data were summarized descriptively and random effects meta-analysis (MA) was used where appropriate. The SR identified 45 relevant studies, 29 describing 59 cases reported as gestational LD in the United States, Europe, and Asia (1969–2017). Adverse birth outcomes included spontaneous miscarriage or fetal death (n = 12), newborn death (n = 8), and newborns with an abnormal outcome (e.g. hyperbilirubinemia, respiratory distress and syndactyly) at birth (n = 16). Only one report provided a full case description (clinical manifestations in the mother, negative outcome for the child, and laboratory detection of B. burgdorferi in the child) that provides some evidence for vertical transmission of B. burgdorferi that has negative consequences for the fetus. The results of 17 epidemiological studies are included in this SR. Prevalence of adverse birth outcomes in an exposed population (defined by the authors as: gestational LD, history of LD, tick bites or residence in an endemic area) was compared to that in an unexposed population in eight studies and no difference was reported. A meta-analysis of nine studies showed significantly fewer adverse birth outcomes in women reported to have been treated for gestational LD (11%, 95%CI 7–16) compared to those who were not treated during pregnancy (50%, 95%CI 30–70) providing indirect evidence of an association between gestational LD and adverse birth outcomes. Other risk factors investigated; trimester of exposure, length of LD during pregnancy, acute vs. disseminated LD at diagnosis, and symptomatic LD vs. seropositive women with no LD symptoms during pregnancy were not significantly associated with adverse birth outcomes. This SR summarizes evidence from case studies that provide some limited evidence for transplacental transmission of B. burgdorferi. There was inconsistent evidence for adverse birth outcomes of gestational LD in the epidemiological research, and uncommon adverse outcomes for the fetus may occur as a consequence of gestational LD. The global evidence does not fully characterize the potential impact of gestational LD, and future research that addresses the knowledge gaps may change the findings in this SR. Given the current evidence; prompt diagnosis and treatment of LD during pregnancy is recommended.


Objectives of the Systematic Review
Conduct a systematic review to identify, critically appraise, extract data on relevant outcomes and synthesize the literature on the impact of Lyme disease during pregnancy. Methods following the best practices for synthesis research prescribed by the Cochrane Collaboration will be utilized to undertake this project.

Study Question
What is the evidence that gestational Lyme disease in humans causes adverse birth outcomes including congenital abnormalities?
Inclusion criteria: is all evidence examining the impact of maternal Lyme disease in humans for any outcome conducted anywhere in the world.
Exclusion criteria: research on non-human hosts including animal models of the impact of Borrelia burgdorferi infection on pregnancy, fetal and newborn outcomes.

Search Strategy
The search algorithm below will be executed in 3 bibliographic databases Pubmed, Scopus and Embase on October 16, 2017 via the Public Health Agency of Canada library. A search verification strategy was employed to identify any literature that was omitted from the bibliographic database search. Search results will be downloaded, deduplicated and managed in reference management software, Endnote.

Algorithms
((lyme or borrelia or borreliosis) and (pregnancy or pregnant or maternal or fetus or foetus or newborn or congenital))

Search Verification
Reference lists of a minimum of four relevant publications (book chapters, literature reviews and/or primary research articles) will be scanned for relevant citations missed by the electronic search. If we are still finding missed publications after four papers, additional papers will be selected and the reference lists will be scanned until we no longer identify potentially relevant research that has not been captured already.

Grey Literature Search
The following websites were searched by using simple combinations of the keywords lyme or borrelia and pregnancy or fetus or newborn to identify potentially relevant pages and each page was screened for primary data related to the topic. Potentially relevant grey literature would be added to the citation list for relevance screening. However, no additional citations were found. Pubmed: n=392 (no limits, mapping on) (((lyme or borrelia or borreliosis) and (pregnancy or pregnant or maternal or fetus or foetus or newborn or congenital))) Scopus: n=403, limited to articles, conference proceedings and journal articles (excluded books, reviews, patents etc.) TITLE-ABS-KEY ( ( ( ( lyme OR borrelia OR borreliosis ) AND ( pregnancy OR pregnant OR maternal OR fetus OR foetus OR newborn OR congenital ) ) ) ) AND ( LIMIT-TO ( DOCTYPE , "ar" ) OR LIMIT-TO ( DOCTYPE , "cp" ) ) AND ( LIMIT-TO ( SRCTYPE , "j" ) ) Embase: n=534 limited to research or work about humans using Embase' filter.

Relevance Screening (RS)
The relevance screening level will be done on the title, abstract and keywords where available. There is 1 question and the answers are based upon the inclusion / exclusion criteria and can be found in the appendix.

Inclusion / Exclusion criteria
Potential inclusion/exclusion criteria 1) Time frame -no time frame 2) Country -All 3) Language -English, French. All other languages will be identified, the paper will be obtained and we will evaluate if there are resources to include papers in other languages. 4) Document Type: Any article, report or thesis containing primary data (data collected by the author/ author's organisation). All literature reviews, letters, commentaries, new reports etc that do not contain primary data will be excluded at relevance screening (based on title/abstract) or the beginning of the second level (based on the full paper). 5) Study design -all 6) Population-studies on humans and the impact of infection on human pregnancies. Animal models and other studies on B. burgdorferi infection in animals will be identified as such at relevance screening and excluded from the review. 7) Pathogen -Any of the B. burgdorferi group of borrelia.

Risk-of-bias assessment and GRADE:
Relevant studies that meet all eligibility criteria will undergo a risk of bias assessment. Most relevant studies are expected to be case reports, cross-sectional or cohort design. A risk of bias assessment tool will be developed to assess the internal validity of the study, ie: whether it answers the research question correctly. In this sense we are assessing systematic error, deviation from the truth, in results or inferences (Balshem et al., 2011, Guyatt et al., 2011b, Higgins & Altman, 2008. These biases may vary in direction and magnitude; however it is impossible to know the extent that the biases have influenced the results of a study (Higgins & Altman, 2008). The risk of bias evaluates selection bias, performance bias, attrition bias, detection bias, reporting bias and confounding bias. The results of this help us to use the GRADE criteria to grade the evidence (Balshem et al., 2011, Higgins & Green, 2011, The, 2013.
For each outcome a Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria was applied (Guyatt et al., 2011a, Higgins & Green, 2011, The, 2013. The risk of bias assessment aims to assess the internal validity of the study which informs one of the GRADE criteria (Higgins & Green, 2011, Higgins & Altman, 2008. The other 6 GRADE components include indirectness of evidence, unexplained heterogeneity, imprecision/high prevision of results, high probability of publication bias, studies are underestimating the measure of effect and detection of a dose-response gradient. GRADE criteria are summarized across groups of like studies to indicate the level of confidence in the current evidence (The, 2013). The one to four star grading system indicates: **** high confidence that the effect estimate is close to the true effect; *** moderate confidence in the effect estimate, but future studies may be substantially different; ** limited confidence in the estimate of effect, the true effect may be substantially different; * very little confidence in the estimate of effect, the true effect is likely to be substantially different (Balshem et al., 2011, Guyatt et al., 2011b.

Data Extraction
A data extraction form was developed to extract general study information and capture specific quantitative and descriptive outcomes from each study captured in this review. See the Data extraction form in the Appendix.

Review management:
To ensure rigour in the review process, all steps will be conducted using pre-tested tools by two independent reviewers. All references identified in the review will be de-duplicated in the reference management program Endnote© (Thomson Reuters, USA) and imported into the systematic review management software DistillerSR (Evidence Partners, Ottawa, Canada) to facilitate review management and progress. All extracted data will be downloaded as Excel spreadsheets for analysis.

Data Analysis:
Extracted data from relevant articles will be descriptively characterized and summarized. Quantitiatve outcomes including prevalence, counts, and measures of association will be converted where necessary into a standard effect size metric, based on the mostly commonly used measure reported in relevant studies (Borenstein et al., 2009). Data will then be stratified into sufficiently comparable subgroups if there are some, and random-effects meta-analysis will be conducted to determine average effect sizes and the extent of heterogeneity across studies in each subgroup [8]. Meta-analysis will be conducted using Stata (StataCorp LP, College Station, USA). We do not anticipate having outcomes with enough data to warrant meta-analysis. Observational study: Assignment of subjects into an exposed group versus a control group is outside the control of the investigator.

Is the citation primary research on pregnant women, fetus or newborns and the impact of
• Cross-sectional: Examines the relationship of a risk factor and outcome (disease) at a point in time on representative samples of the target population. • Cohort study: is a study in which individuals with differing exposures to a suspected risk factor are observed through time for occurrence of an outcome • Case-control study: compares exposure to the risk factor in subjects who have an outcome (the 'cases') with subjects who do not have the outcome, but are otherwise similar (the 'controls') and drawn from the same sampling frame. There may be an occasional experimental design -please include under "other" 9) Was the allocation sequence adequately generated? (GRADE 1-1) "RCT, ChT Selection bias: systematic differences between baseline characteristics of the groups that are compared."
Unclear: they simply stated that it was "randomized" (formerly partial).
No: Sample drawn without a formal process of random selection: judgment, convenience, purposive.
10) Was the allocation sequence adequately concealed from the participants and the researcher? (GRADE 1-2) "RCT, ChT Selection bias: systematic differences between baseline characteristics of the groups that are compared."

Yes: [text]
Unclear: [text] No: [text] NA -not an experiment [RCT, ChT] Yes: concealment was sufficient and allocation was unlikely to be foreseen (in advance of or during enrollment) ___ page #___ Unclear: author only indicated "blinding" or "concealed treatment" was used.
No: no concealment strategy described or was insufficient.

Yes: [text]
Unclear: [text] Yes: Does the sample reflect the proportion of high risk and low risk people in the population the

Low RoB
Unclear RoB

High RoB
Low risk of bias: no biases were indicated in the assessment. Thus plausible bias is unlikely in all key domains (within this study). (Across studies: most studies indicate low risk) Unclear risk of bias, there are plausible bias that raises doubt about the results as some key domains are "unclear (within this study). (Across studies: most information is from low or unclear RoB).
High Risk of bias indicates that in one or more of the domains serious plausible bias was identified (within the study).
(Across studies: The proportion of studies that are at high risk of bias is sufficient to affect the interpretation of results.)

21) Does this study examine the question of interest directly? (GRADE 2-1, new)
Yes, this study directly addresses the question of interest. Please state which outcomes were directly answered [text] No, this study indirectly examines the question of interest. Please state which outcomes were not directly answered [text] A study may indirectly address the question of interest if: e.g. risk factors we wish to compare are measured independently in two separate trials compared to controls. e.g. the population, risk factors , comparisons or outcomes were not exactly what we are trying to draw conclusions for.

* Downgrading occurs if there is reason
to believe that there may be differences in the conclusions due to indirectness.
22) Was this study funded by or was there involvement of individuals employed by or affiliated with industry (drug or chemical) or a special interest / advocacy group? (GRADE 5-1, No, There are no concerns based on the authors, funding and declarations in the paper.

Yes[text]
This criteria for down-grading would be used if all or most of the trials captured are industry funded or declare heavy sponsor involvement (e.g. advocacy groups), in which case there are concerns that studies of null or negative effect may have been suppressed from new) Identify in text box details: -if there was a declaration of involvement.
-if the study was funded by such an organization -if the author's affiliation was for such an organization. publication.

Select yes and provide details if there was industry or advocacy group sponsorship.
23) Is there reason to believe that due to the population studied, the magnitude of effect (association) of the risk factor (outcome) may be underestimated? new) Yes, an underestimation is likely No, there is no reason to believe the estimated effect is underestimated.
You would answer yes ONLY if there was good reason to think that the study underestimated the potential association or effect of a risk factor due to the population that was sampled.
e.g. The magnitude of association was lower than it likely is in the general population because the comparison group has a similar disease which in also more likely to result in having the exposure of interest. • Measure of association (e.g. odds ratio, relative risk) + EITHER a measure of variability (SE, CIs, variance) or an exact P-value e.g. Odds Ratio e.g. OR 2.5 (2.1-2.9), OR 2.5 (SE 0.4) etc.
If the measure of effect is different across confounders, please specify these results as well.
e.g. The odds of detecting abnormalities in newborns were 2.5 times higher in LD seropositive pregnant women. Continuous: Sufficient information includes: • Mean, sample size, + EITHER a measure of variability (e.g. SD, CIs) or exact P-value/t-value or • Sample size and P-value/t-value from t-test or • Difference in means and a measure of variability (SD, SE, CIs, variance) or • Difference in means, sample size, + EITHER a common SD or an exact P-