Jump to comment:

• Caution advised when comparing or pooling seropositivity proportions
  Max J. Hassenstein, Manuela Harries and Stefanie Castell
  Published on: 19 October 2022
• Response to Dong et al: Global seroprevalence and sociodemographic characteristics of Borrelia burgdorferi sensu lato in human populations: a systematic review and meta-analysis
  Amanda Semper, Eilish Hart, Jennifer Warner, Matthew Dryden, Tim Brooks, Zoë Gibney, Wendi Shepherd, Katherine Russell, Sally Mavin and Dominic Mellor
  Published on: 12 August 2022

• Published on: 19 October 2022

  Caution advised when comparing or pooling seropositivity proportions

  • Max J. Hassenstein, Research fellow Helmholtz Centre for Infection Research
  • Other Contributors:
    • Manuela Harries, Research fellow
    • Stefanie Castell, Deputy Head of Department, Department for Epidemiology; Research fellow

  We read the systematic review by Dong et al. [1] with great interest. The authors aimed to describe global seroprevalence estimates for B. burgdorferi s.l., the causal agent of Lyme disease.

  First, estimating seropositivity for a target population (here, the global population) has two challenges we would like to address:
  1) The age and sex distribution of the population providing sample(s) and the target population should correspond. The simple reason is that advancing age and male sex are well-established risk factors for a positive IgG antibody serostatus [2]. Therefore, one may not conclude the general population seropositivity from an aged sample with a large share of males if not corrected accordingly, e.g., by applying weights; otherwise, seropositivity may be overestimated. Unfortunately, the age and sex profiles of the individual studies were seemingly not considered or discussed for their final seropositivity estimates.
  2) Then, as already stated in the first reply to this manuscript by A. Semper et al., the studies containing subjects with medical conditions or even patients with suspected or confirmed Lyme disease symptomatology are of little use for general population estimates of seropositivity (e.g., [3, 4], included by Dong et al.), as these populations do not correspond to the global population. Also, pooling seropositivity proportions for high-risk populations to obtain global estimates potentially introduces bias and, hence, should be a...

  Show More

  We read the systematic review by Dong et al. [1] with great interest. The authors aimed to describe global seroprevalence estimates for B. burgdorferi s.l., the causal agent of Lyme disease.

  First, estimating seropositivity for a target population (here, the global population) has two challenges we would like to address:
  1) The age and sex distribution of the population providing sample(s) and the target population should correspond. The simple reason is that advancing age and male sex are well-established risk factors for a positive IgG antibody serostatus [2]. Therefore, one may not conclude the general population seropositivity from an aged sample with a large share of males if not corrected accordingly, e.g., by applying weights; otherwise, seropositivity may be overestimated. Unfortunately, the age and sex profiles of the individual studies were seemingly not considered or discussed for their final seropositivity estimates.
  2) Then, as already stated in the first reply to this manuscript by A. Semper et al., the studies containing subjects with medical conditions or even patients with suspected or confirmed Lyme disease symptomatology are of little use for general population estimates of seropositivity (e.g., [3, 4], included by Dong et al.), as these populations do not correspond to the global population. Also, pooling seropositivity proportions for high-risk populations to obtain global estimates potentially introduces bias and, hence, should be avoided [5].

  Furthermore, one study [6] included by Dong et al. [1] considered US, UK, Australian, and German serosamples seems to match the exclusion criteria, as it does not provide participant categories corresponding to the samples. Also, this study aimed to compare assays and test results between laboratories, not to assess seroprevalence. According to [6], the samples considered partially were previously tested for Lyme disease or originate from endemic areas, which indicates that these samples may not be representative of the general population and, therefore, may not be suited for general population estimates.
  Then, we perceive that Dong et al. [1] did not differentiate between IgG and IgM class antibodies for their global estimates, despite reported differences in seropositivity proportions for each antibody class (Fig 2). Ideally, IgG and IgM are considered separately and not pooled for global estimates, as IgM serostatus may underestimate long-term serostatus due to its short-term detectability compared to IgG [7].
  We want to add the limitation that different serological tests used over time result in differing seropositivity due to varying sensitivity and specificity [8], i.e., hamper comparability, which may contribute to the heterogeneity found by Dong et al.

  Finally, some minor remarks: it is not clear to us whether the authors only considered peer-reviewed papers, as indicated by the abstract, or also grey literature, as indicated in the methods section. Then, the search strategies presented in the abstract differ substantially from the strategies presented in Supplementary appendix 2. Further unclear aspects from the literature search are: 1) how the search could have started back in 1984, 2) how the last date of the literature lies past the article submission date, with no indication of an updated search during the review procedure, 3) why the authors considered literature with PCR testing methods, as these are not of use for antibody detection.

  References
  1 Dong Y, Zhou G, Cao W, et al. Global seroprevalence and sociodemographic characteristics of Borrelia burgdorferi sensu lato in human populations: a systematic review and meta-analysis. BMJ Glob Health 2022;7(6). doi: 10.1136/bmjgh-2021-007744
  2 Wilking H, Fingerle V, Klier C, et al. Antibodies against Borrelia burgdorferi sensu lato among Adults, Germany, 2008-2011. Emerg Infect Dis 2015;21(1):107–10. doi:10.3201/eid2101.140009
  3 Karatolios K, Maisch B, Pankuweit S. Suspected inflammatory cardiomyopathy. Prevalence of Borrelia burgdorferi in endomyocardial biopsies with positive serological evidence. Herz 2015;40 Suppl 1:91–95. doi:10.1007/s00059-014-4118-x
  4 Dersch R, Sarnes A, Maul M, et al. Immunoblot reactivity at follow-up in treated patients with Lyme neuroborreliosis and healthy controls. Ticks Tick Borne Dis 2019;10(1):166–69. doi:10.1016/j.ttbdis.2018.09.011 [published Online First: 26 September 2018].
  5 Spronk I, Korevaar JC, Poos R, et al. Calculating incidence rates and prevalence proportions: not as simple as it seems. BMC Public Health 2019;19(1):512. doi:10.1186/s12889-019-6820-3 [published Online First: 6 May 2019].
  6 Best SJ, Tschaepe MI, Wilson KM. Investigation of the performance of serological assays used for Lyme disease testing in Australia. PLoS One 2019;14(4):e0214402. doi:10.1371/journal.pone.0214402 [published Online First: 29 April 2019].
  7 Coors A, Hassenstein MJ, Krause G, et al. Regional seropositivity for Borrelia burgdorferi and associated risk factors: findings from the Rhineland Study, Germany. Parasit Vectors 2022;15(1):241. doi:10.1186/s13071-022-05354-z [published Online First: 4 July 2022].
  8 Cook MJ, Puri BK. Commercial test kits for detection of Lyme borreliosis: a meta-analysis of test accuracy. Int J Gen Med 2016;9:427–40. doi:10.2147/IJGM.S122313 [published Online First: 18 November 2016].

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 12 August 2022

  Response to Dong et al: Global seroprevalence and sociodemographic characteristics of Borrelia burgdorferi sensu lato in human populations: a systematic review and meta-analysis

  • Amanda Semper, Principal Scientist and Scientific Programme Manager UK Health Security Agency - Porton (Rare & Imported Pathogens Laboratory)
  • Other Contributors:
    • Eilish Hart, PhD Student
    • Jennifer Warner, Knowledge Discovery Scientist
    • Matthew Dryden, Consultant in Infection
    • Tim Brooks, Clinical Services Director
    • Zoë Gibney, Principal Scientist (Zoonoses)
    • Wendi Shepherd, Consultant in Emerging Infections and Zoonoses
    • Katherine Russell, Consultant Medical Epidemiologist
    • Sally Mavin, Director
    • Dominic Mellor, Consultant in Veterinary Public Health

  Through a systematic review and meta-analysis, Dong et al (1) have calculated a global B. burgdorferi sensu lato (Bbsl) seroprevalence estimate of 14.5% (95% CI 12.8% to 16.3%). We question the accuracy and appropriateness of such an estimate.

  As the authors demonstrate, seroprevalence estimates based on orthogonal 2-tier serological testing with a confirmatory Western-blot assay decrease the risk of false-positive results and are more reliable than those using single assays. Yet the pooled 14.5% estimate includes studies that used single assays, apparently without adjusting for the decreased reliability of single-tier testing. When studies using single-tier assays were excluded, the pooled estimate was reduced to 11.6% (95% CI 9.5% to 14.0%). The 14.5% estimate is based on studies spanning four population categories general, high-risk, tick-bitten and having Lyme-like symptoms. When these sub-groups were compared, the general population had a pooled seropositivity rate of 5.7% (95% CI 4.3% to 7.3%). We argue that only the general population category is relevant when estimating an unbiased population seroprevalence.

  Irrespective of accuracy, using a headline global seroprevalence estimate may be misleading, implying homogeneity when, as the authors report, there is wide variation in B. burgdorferi seroprevalence between countries and regions. Furthermore, the authors suggest that analysis of seropositivity to anti-Bbsl antibodies enhances understanding of th...

  Show More

  Through a systematic review and meta-analysis, Dong et al (1) have calculated a global B. burgdorferi sensu lato (Bbsl) seroprevalence estimate of 14.5% (95% CI 12.8% to 16.3%). We question the accuracy and appropriateness of such an estimate.

  As the authors demonstrate, seroprevalence estimates based on orthogonal 2-tier serological testing with a confirmatory Western-blot assay decrease the risk of false-positive results and are more reliable than those using single assays. Yet the pooled 14.5% estimate includes studies that used single assays, apparently without adjusting for the decreased reliability of single-tier testing. When studies using single-tier assays were excluded, the pooled estimate was reduced to 11.6% (95% CI 9.5% to 14.0%). The 14.5% estimate is based on studies spanning four population categories general, high-risk, tick-bitten and having Lyme-like symptoms. When these sub-groups were compared, the general population had a pooled seropositivity rate of 5.7% (95% CI 4.3% to 7.3%). We argue that only the general population category is relevant when estimating an unbiased population seroprevalence.

  Irrespective of accuracy, using a headline global seroprevalence estimate may be misleading, implying homogeneity when, as the authors report, there is wide variation in B. burgdorferi seroprevalence between countries and regions. Furthermore, the authors suggest that analysis of seropositivity to anti-Bbsl antibodies enhances understanding of the global epidemiology of Lyme disease. Caution should be exercised here since Lyme disease as a clinical entity is only reliably described from temperate areas of the northern hemisphere, mirroring the distribution of its tick vectors (2). At least 20 species of bacteria belong to the Bbsl complex (2, 3) and could be transmitted to humans through a tick bite, but only a handful are pathogenic and cause Lyme disease (3). The serology assays used internationally to estimate borrelia exposure may recognise antibodies generated against non-pathogenic Bbsl species. Consequently, the Bbsl global seroprevalence estimate overestimates those who have experienced clinical Lyme disease (or asymptomatic infection) and may explain why Dong et al give Bbsl seroprevalence estimates for countries where no Lyme disease cases have been confirmed.

  Dong et al derived a seroprevalence estimate for the UK in the 11-20% range based on two Scottish studies, with calculated seroprevalence estimates of 4.17% and 36.48% (4, 5). One (4) used Scottish blood donors. The other (5) utilised sera from individuals with suspected Lyme disease; it was not a study on either seroprevalence or incidence, was not representative of the general population and its inclusion introduces bias into the results. The true seroprevalence figure for Scotland is likely closer to the 4.2% estimate but importantly, there is regional variation (0 - 8.6%) within Scotland (4). Dong et al have used the biased Scottish estimate to extrapolate from one UK nation to others (England, Wales and Northern Ireland). This is misleading since published data on lab-confirmed Lyme disease cases show a higher incidence for Scotland than the other UK nations (6), although published seroprevalence data for the general populations in the other UK nations is lacking (6).
  The study by Dong et al has been widely reported by media outlets leading to headlines claiming 1 in 7 people globally have been exposed to B. burgdorferi at some time (7). As described above, the interpretation of the ‘global’ seroprevalence estimate is flawed as are media releases based on it and may mislead and alarm the public unnecessarily. What is most important is that the public understand where high risk areas are locally and when travelling overseas, so that they can take appropriate measures to limit their exposure to ticks and to prevent Lyme and other tick-borne diseases.

  References
  1. Dong Y, Zhou G, Cao W, Xu X, Zhang Y, Ji Z, et al. Global seroprevalence and sociodemographic characteristics of Borrelia burgdorferi sensu lato in human populations: a systematic review and meta-analysis. BMJ Glob Health. 2022;7(6).
  2. Mead PS. Epidemiology of Lyme disease. Infect Dis Clin North Am. 2015;29(2):187-210.
  3. Bobe JR, Jutras BL, Horn EJ, Embers ME, Bailey A, Moritz RL, et al. Recent Progress in Lyme Disease and Remaining Challenges. Front Med (Lausanne). 2021;8:666554.
  4. Munro H, Mavin S, Duffy K, Evans R, Jarvis LM. Seroprevalence of lyme borreliosis in Scottish blood donors. Transfus Med. 2015;25(4):284-6.
  5. Mavin S, Evans R, Milner RM, Chatterton JM, Ho-Yen DO. Local Borrelia burgdorferi sensu stricto and Borrelia afzelii strains in a single mixed antigen improves western blot sensitivity. J Clin Pathol. 2009;62(6):552-4.
  6. Lorenc T, Jones-Diette J, Blanchard L, et al. Incidence and surveillance of Lyme disease: systematic review and policy mapping. London: EPPI-Centre, Social Science Research Unit, UCL Institute of Education, University College London; 2017.
  7. Wetzel C. More than 1 in 7 people worldwide have had Lyme disease. New Scientist. 2022.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top