Assignment 2 : Tickborne Diseases
Orloski, K.A. et al. Emergence of Lyme Disease in Hunterdon County, New Jersey, 1993: A Case-Control Study of Risk Factors and Evaluation of Reporting Patterns. Am J Epidemiol 1998; 147: 391-397.
1. What was the main purpose/hypothesis of the study?
To evaluate potential risk factors for Lyme disease in Hunterdon County, New Jersey and to determine the cause of an observed increase in disease incidence from 1992 to 1993.
2. What was the study design? What are its strengths and weaknesses?
STUDY DESIGN:
Two studies:
a) A descriptive characterization (person, time) of patients with acute Lyme disease and a case-control study of risk factors for acquiring Lyme disease.
b) A cross-sectional or descriptive study of reporting practices for Lyme disease in 8 physicians offices in Hunterdon County in 1993.
STRENGTHS:
· Lyme disease is a rare disease
· The magnitude of exposure-disease relationships can be estimated
WEAKNESES:
· Confounding is possible
· Information on exposure is collected following disease occurrence (retrospective)
· Recall bias can be problematic
· Comparison of cases and noncases involves two separate populations of persons. Controls were recruited from either the physicians’ offices or the local hospital outpatient laboratory, which may have introduced selection bias. Controls may have been more likely than the general population to have a chronic illness or disability that limited their physical abilities or affected their choice of residence. This bias would be expected to accentuate the differences between cases and controls, particularly with respect to risk factors involving physical and outdoor activities.
3. What was the study population? Was it representative?
STUDY POPULATION:
A total of 62 case reports of Lyme disease from Hunterdon County, with ECM lesions and occurrence time of 1/93-11/93; systematic sampling (1:3 of alphabetized list) from among all reports. Controls recruited prospectively during 11/93, age-matched (5 year intervals), persons presenting to private physicians reporting Lyme disease cases or from area hospital lab. Present or previous case of Lyme disease was an exclusion criterion for controls.
REPRESENTATIVE:
Cases:
Not indicated why only 62/216 (29%) of all cases were sampled. The sample size may be too small to demonstrate statistically significant differences for the risk factors examined.
Controls:
Controls were recruited from either the physicians’ offices or the local hospital outpatient laboratory, which may have introduced selection bias. Controls may have been more likely than the general population to have a chronic illness or disability that limited their physical abilities or affected their choice of residence. This bias would be expected to accentuate the differences between cases and controls, particularly with respect to risk factors involving physical and outdoor activities.
This strategy is likely to provide a meaningful sample of cases and age-matched controls (but not otherwise matched) within the geographic and time-based criteria of interest.
COMMENT:
Only the data from the sample of cases and all cases was provided and a comparison of the selected cases (n=51) to the eligible but non-selected cases (n=154) was not performed.
Sample (51) Total (216)
Age (median-yrs) 46 (5-82)* 41 (1-89)
Male (%) 53 56
*Range
All cases were younger and had a higher proportion of males. There may be some variability between selected and non-selected cases in the relationships between the factors and disease. The sample of cases may not be fully representative for all risk factors of Lyme disease. However, we cannot determine that since no information is provided on the non-selected cases. This comparison should always be made to ensure sample is representative of all cases.
4. What exposures or risk factors were measured? Were there any biases or limitations in their measurement?
RISK FACTORS:
Factors evaluated included aspects of their disease syndrome (cases only), characteristics of residential property, frequency of observing deer on property, outdoor activities, tick avoidance practices, and cat ownership. Relevant time period for outdoor activities and tick avoidance was 6/93-8/93, antecedent to the end of the time window for accrual of cases.
BIASES/LIMITATIONS:
· Exposures were assessed by questionnaire; in person for cases, by telephone for controls. Since questionnaires for cases and controls were completed by different methods this could result in differential misclassification of exposure measures.
· Most cases were interviewed 3-6 months after their illness, hence may have diminished recall. However, cases may remember exposures more accurately than controls (recall bias). Only 5/51 cases recalled a tick bite.
· In assessing tick exposure: true habits/activities which plausibly lead to tick habitat exposure are under-represented. No validation efforts are mentioned including the approximate time or frequency for activity types. Not all measured exposures are shown in Table 1 (e.g., fishing and hunting are not shown, but are mentioned in text as having no predictive power). Presume by their exposure assessments that tick exposure was of primary interest (by proxy by other measured variables), the objective being to assess behavioral predictors of infected tick exposure rather than to validate tick bite as the principal mechanism of inoculation of the organism.
5. What was the principal outcome of interest (infection or disease) and how was it measured? Identify advantages and disadvantages with this measure.
OUTCOME:
Reported cases of Lyme disease with documented erythema migrans
MEASURED:
Case reports submitted by physicians to the Department of Health
ADVANTAGES:
EM is relatively characteristic of Lyme Disease
While serology is usually employed to confirm the clinical diagnosis of Lyme disease, it is problematic, frequently resulting in false negative results.
DISADVANTAGES:
Lyme disease with clinical manifestations other than erythema migrans would be excluded from the analysis
Cases may have been misdiagnosed – diagnose hypersensitivity reactions to tick bites as erythema migrans
Only 9/37 (24%) were confirmed by serology, suggesting that ~75% of cases were not Lyme disease, i.e., misdiagnosis. While this could be due to failure of the test to detect antibodies (low sensitivity), sera were tested 3-6 months after onset of symptoms.
COMMENT:
THE RESULTS OF SEROLOGY DID NOT CHANGE THE STATUS OF CASES AND CONTROLS
6. What were the main findings? Do you agree or disagree? Support your position.
MAIN FINDINGS:
Rural residence, presence of woods, rock walls, deer, and participation in periresidential brush clearing were identified risk factors for in this study. Most of these findings were by univariate analysis; multivariate modeling showed only rural residence and brush clearing to be significant predictors, both with wide 95% CIs.
AGREE/DISAGREE:
AGREE
The results are biologically plausible and consistent with other studies on the epidemiology, based on known features of vector tick ecology.
7. Was there any potential confounding in the data analyses? Was it considered in the data analyses? Explain.
CONFOUNDING:
The authors did not mention their strategy to assess confounding in data analysis or conclusions. This is a major flaw since all such issues should be discussed to the extent that the causal pathways are reasonably understood and the relationship among measured exposures and disease described.
CONSIDERED:
· The method of matching should validate comparisons of cases and controls with respect to measured exposures, but only to the degree that age is an important contributor for such (principally behavioral) tendencies.
· Table 1 (univariate analysis) lists sex among cases and controls, but this was not a matching factor and thus it is not surprising that it differed.
· It would be helpful to know more about the topography and other features of this community so that more precise (perhaps more informative) information regarding particular important risk factors for exposure during the study period could be gleaned. If the area is principally split between developed (including paved) and less developed areas that would be important to know. If the entire area, mostly residential, is substantially rural, than it is conceivably difficult to dissect out the most important Lyme disease exposure characteristics in such a community. So if that was the essential objective, a different community should have been selected for study.
COMMENT:
While cases and controls were comparable in age and race, there was a significant difference for sex (chi-square=4.02, p=0.04).
Group Age (median-yrs) Race(%Caucasian) Sex(%F)
Cases 46 100 47
Controls 45 98 69
If exposures differ between males and females this would introduce a difference between the groups that may be due to sex rather than the exposures. For example, if males are more likely to “Clear brush” the significant finding (OR=4.0, 95% CI=1.5-10.7) for this factor may be due to the greater proportion of males in the case than control group and not the exposure itself. For the other factors, e.g., “Rural residence”, it would appear that sex may not be a confounding factor.
8. Were there shortcomings/limitations to the study? If so, were they of sufficient magnitude to invalidate the results?
SHORTCOMINGS/LIMITATIONS:
· Not indicated why only 62/216 (29%) of all cases were sampled. The sample size may be too small to demonstrate statistically significant differences for the risk factors examined. No power calculation.
· Many of the variables considered in the study are surrogate measures of exposure to ticks, and hence may not reflect true exposure to tick bites. Also, many exposures were either common or rare. Thus, the number of discordant pairs available for analysis was small for many of the variables considered. Given the conventional significance level of 0.05, the power of the study may have been insufficient to detect real differences that may have existed between the cases and controls regarding these variables. The odds ratio estimates were thus imprecise with wide confidence intervals.
· The lack of significance for any of the assessed tick-prevention factors evaluated is particularly problematic given that several are known in other studies to be useful. This draws into question the validity of the exposure-disease estimates provided.
· Only 9/37 (24%) cases were positive by serology suggesting that ~75% of cases were not Lyme Disease, i.e., misdiagnosis.
· One control was positive for Ig G antibodies. Misclassification of a case as a control would be expected to bias point estimates toward the null, however effect probably negligible.
SUFFICIENT MAGNITUDE TO INVALIDATE:
No. All conclusions, however mundane, are entirely consistent with what would be expected and predicted.
9. Write a one-sentence summary of the article that could potentially be used in the context of community health promotion campaigns.
Persons who live in rural areas of Hunterdon County, NJ, including residences with rock walls, woods, deer, and possibly bird feeders, are at a somewhat increased risk of acquiring Lyme disease during summer-fall months of the year and should use personal protective measures against tick exposure.
10. Based on this work, what would be the next study you would want to do if you had the necessary resources? Why?
NEXT STUDY:
Perform field collections of ticks from the identified high-risk area (including possibly small mammals and birds) and test for the agent of Lyme disease, to validate that particular ecozones in Hunterdon County present meaningfully higher risk of infective tick bite than other areas within this geographic region.
WHY:
This information might provide more precision regarding the public health campaigns for Hunterdon County residents, particularly if done prospectively over various weeks throughout summer and fall when Lyme disease incident cases occur.