Materials and Methods

Animals—The source population in this study was identified and described through an earlier study.⁵ Briefly, the population consisted of 13,976 accessioned birds at least 5 months of age that were present (and released from quarantine) for at least 30 days between 1991 and 2005 at the ZSSD, which includes the San Diego Zoo in San Diego and the SDZWAP in Escondido, Calif. Health screenings of this population included close monitoring of all birds on a daily basis and veterinary care when needed. Birds with suspected mycobacterial infections, determined on the basis of diagnostic testing performed as a part of clinical evaluations or routine physical examinations, generally were humanely euthanized to protect the healthy population from the risk of infectious disease spread.

During the study period, all birds from the source population that died or were euthanized (because of mycobacteriosis or other health concerns) underwent thorough necropsies and histologic evaluations of complete tissue sets (unless advanced autolysis precluded histologic analysis) to identify potential health problems and confirm clinical diagnoses. Health and postmortem data from the source population were reviewed to identify all birds with acid-fast organisms in evaluated tissues; these birds were considered positive for mycobacteriosis. Specific methods and criteria used for diagnosing mycobacterial infection in the source population, characterization of affected systems, and evaluation of disease severity have been described.⁵

For the present study, a subset of birds from the source population (n = 2,413) was further evaluated. This subset consisted of birds that had ever shared an enclosure with another bird that was diagnosed with intestinal mycobacteriosis. To refine the study population even more, all study birds were required to have been housed with the intestinally infected enclosure mate within 1 year of the enclosure mate's diagnosis (ie, the date that the original tissue samples were collected from which the diagnosis was made, usually corresponding to the date of death). Birds in the source population became eligible for the study once they met initial criteria and remained in the study until they were removed from the ZSSD bird collection or the study ended.

For the purposes of the present investigation, birds from the study population are referred to as exposed if they had previous direct contact with an intestinally infected bird within the specified period. Exposure was not a 1-time event but rather an intermittent and sometimes ongoing occurrence in a dynamic population. Birds in the study population were exposed in a variety of distinct enclosures (n = 88), and the specific characteristics of exposure events (eg, timing and total duration of contact or characteristics of the infected enclosure mate) also varied for study subjects.

Study design—A matched case-control study was used to investigate potential factors associated with avian mycobacterial disease. Birds in the study population with histologic evidence of avian mycobacterial infection (ie, acid-fast organisms detected in any tissue) were defined as cases. Each case was matched to 4 randomly selected controls (birds without a diagnosis of mycobacteriosis) that were approximately the same age (as their matched case's age at diagnosis) and from the same taxonomic group in a manner analogous to that described elsewhere.⁵ Age matching was performed to account for the age-related effects of exposure and disease development. For cases < 1 year of age, matched controls were between 5 months and 1 year of age. For cases ≥ 1 year of age, matched controls were within 1 year of the same age of their matched case but were at least 1 year old. Taxonomic order was used as the matching factor for all groups of birds except Columbiformes, for which pronounced species diversity prompted use of genus as the matching factor.

Evaluation of risk factors for mycobacteriosis—Most risk factors evaluated were those previously examined for the source population⁵ and consisted of biologically plausible hypotheses about potential differences in the odds of disease based on individual bird characteristics, exposure characteristics, potential stressors, and management practices. These factors included sex, hatch type (zoo or wild-hatched), import status (hatched at the ZSSD vs imported into the ZSSD from another institution or the wild), combined hatch-type and import status (comparing wild-hatched, imported birds and zoo-hatched, nonimported birds separately with the reference group of zoo-hatched, imported birds), years in the ZSSD collection, presence in the collection before and after 2001 (to evaluate the effects of a rigorous mycobacteriosis screening program that began in 2001), number and rate of translocations between ZSSD avian enclosures, and whether birds ever moved between ZSSD campuses. Further details on these variables are described elsewhere.⁵

Additional variables pertaining to exposure characteristics were also evaluated. Disease severity of each infected enclosure mate was characterized according to the previously described definitions⁵ and used to further assess disease risk among cases and controls. The odds of having mycobacteriosis were compared between birds that ever shared an enclosure with infected birds with intestinal lesions classified as significant (ie, some disruption of organ function likely and mycobacterial disease considered to have contributed to or caused death) and those exposed only to birds with incidental lesions (ie, organ function unlikely to have been adversely affected). Similarly, birds ever sharing an enclosure with infected birds that had moderate to abundant numbers of acid-fast organisms present in intestinal lesions (ie, acid-fast organisms easy to detect and observed in at least half of the lesions) were compared with those only exposed to other birds with few acid-fast organisms present (ie, acid-fast organisms only present in some affected areas of the intestines and present in low numbers within cells).

To further evaluate qualitative exposure differences across the study population, 5 attributes representing the location, timing, and characteristics of exposure events were assessed. Many birds had multiple exposures in different settings, so presence or absence of a particular type of exposure among all exposure experiences was classified as follows. To determine the effect of exposure in a small enclosure, birds that were exposed at least once while living in a small enclosure (< 46.5 m² and < 142 m³) were compared with birds that were exposed only when living in larger enclosures, with details on enclosure size assessments described elsewhere.⁵ To determine the effect of exposure at the San Diego Zoo, birds that were ever exposed at the zoo were compared with birds for which exposures occurred only at the SDZWAP. To determine the effect of exposure at an early age, birds that were exposed at < 1 year of age were compared with birds that were not exposed until they were at least 1 year of age. To determine the effect of exposure within 30 days of diagnosis of mycobacteriosis, birds that shared an environment with an intestinally infected bird within 30 days of the infected bird's diagnosis (when an increase in shedding may occur because of progressive disease^7,8) were compared with birds that were exposed only during other intervals. Finally, to determine the effect of exposure to a conspecific, cases or controls that were exposed at least once to a diseased conspecific were compared with those that were only exposed to other bird species.

The duration of exposure was estimated by counting the number of days that case and control birds had shared an environment with an intestinally infected bird within 1 year before or after that infected bird's diagnosis. For birds exposed to > 1 infected enclosure mate, days were summed for all individual exposures to yield a total accounting for multiple exposures. The total exposure duration in days was evaluated as a continuous variable.

Statistical analysis—Incidence of avian mycobacteriosis, determined on the basis of total bird-years at risk among exposed birds, and 95% CIs were calculated, with disease-positive birds contributing to population bird-years at risk until 1 year prior to becoming a case. Lifetime prevalence was determined for all birds in the exposed population and for each major species grouping. Lifetime prevalence was also determined, separately, among birds that had postmortem histologic examinations performed during the study period. Overall incidence rates and prevalence were then compared with source-population data from the previous study⁵ for birds that were present during the same period but that never shared an enclosure with another bird in which mycobacteriosis was diagnosed. The Spearman rank correlation coefficient was used to evaluate the potential for correlation between exposure (proportion of total ZSSD population exposed) and prevalence (proportion of total ZSSD population and exposed population with positive results for mycobacteriosis). Exact 95% CIs were determined for all proportions and rates.

Univariate conditional logistic regression (accounting for matching) was used to screen for factors to include in the multivariable modeling process and to estimate unadjusted ORs, 95% CIs, and Wald statistics. Associations identified with a value of P < 0.25 were further evaluated by use of a multivariable conditional logistic regression model⁹ to calculate the adjusted ORs and 95% CIs for cases relative to controls as a function of the evaluated characteristics and exposure histories. Methods for model fitting, assessment of linearity, assessment of confounding, assessment of influential observations and outliers, and criteria for selecting the final model were performed as described.⁵ Potential inflation of the ORs attributable to sparse data was examined.¹⁰ Analyses were conducted with the aid of statistical software.^a Statistical significance was defined as a value of P < 0.05.

Results

Avian mycobacteriosis was diagnosed in 3.5% (85/2,413; 95% CI, 2.8% to 4.3%) of the 2,413 exposed birds that were identified for the study. Within this exposed population, 985 of the birds died during the study period and underwent postmortem examination; 8.6% (85/985; 95% CI, 7.0% to 10.6%) of them had histologic evidence of mycobacteriosis.

The incidence of mycobacteriosis among the exposed population during the entire study period from 1991 through 2005 was estimated at 7.9 cases/1,000 bird-years at risk (85/10,814 total bird-years at risk; 95% CI, 6.3 to 9.7 cases/1,000 bird-years at risk). During the same period in the full ZSSD source population, lifetime prevalence was 0.7% (62/9,215; 95% CI, 0.5% to 0.9%) and incidence was 2.0 cases/1,000 bird-years at risk (62/30,435; 95% CI, 1.6 to 2.6 cases/1,000 bird-years at risk) for birds that never shared an enclosure with a bird in which mycobacteriosis had been diagnosed.

Exposed birds represented 22 taxonomic groups and 434 different species. Mycobacteriosis was diagnosed in birds from 9 of the 22 taxonomic orders (Table 1) and in 60 different species and subspecies that ranged in age from 6 months to > 19 years (median age, 5.6 years) at the time of diagnosis; 6 birds were diagnosed when < 1 year of age. Prevalence among exposed birds varied among taxonomic groups, with the highest prevalence in exposed Columbiformes (6.6%) and Psittaciformes (5.7%) and no disease detected in several species, despite exposure to infected enclosure mates. Prevalence across taxonomic groups from the entire ZSSD bird population did not correlate with the proportion of exposed birds within each taxonomic order (Spearman ρ = 0.25; P = 0.266) but was similar to the prevalence pattern across taxonomic groups in the exposed subset of the bird population (Spearman ρ = 0.71; P < 0.001).

For the case-control analyses, 79 mycobacteriosis-positive birds were matched to 316 randomly selected, nondiseased controls. Six cases were not included because of a lack of available controls. Ten of the 17 variables in the univariate analysis met the screening criterion (P < 0.25) for further consideration in the multivariable analysis and included the following: hatched in the wild, imported into the ZSSD, combined import status–hatch type, years in the collection, exposure to a bird with intestinal lesions classified as significant, exposure to infected bird in a small enclosure, exposure when < 1 year of age, exposure to infected enclosure mate within 30 days of that mate's diagnosis, exposure to an infected conspecific, and exposure duration (Table 2). The final adjusted model included 5 of these variables (Table 3).

The final model revealed that cases were 7.5 times as likely to have been imported from outside sources (vs having been hatched at the ZSSD) than controls. Cases were 4.8 times as likely to have been exposed at a young age (< 1 year) than controls. Cases were 3.3 times as likely to have been exposed to an infected enclosure mate of the same species (vs a different species of bird) and 3.2 times as likely to have been exposed to an infected enclosure mate while sharing a small enclosure (vs a large enclosure) than controls. Finally, the likelihood of being a case increased with increasing exposure duration, which translated into an 87% increase in odds of mycobacteriosis after a year (365 days) of exposure. No effects attributable to limited discordant pairs, interactions, or additional confounding variables were identified.

Discussion

Management of mycobacteriosis in zoos has focused on the identification of infected birds and subsequent minimization of contact with other birds to reduce the potential for disease spread. In the present study, we found the incidence and lifetime prevalence of mycobacteriosis were quite low among birds that were known to have been exposed to an infected enclosure mate. Incidence was higher among birds in the present study population (7.9 cases/1,000 bird-years at risk) than in the subset of the source population that was not exposed (2.0 cases/1,000 bird-years at risk), which supports the importance of exposure to an infected enclosure mate identified in our previous study.⁵ Among birds that were in direct contact with other infected birds, 91% to 96% never developed disease. These findings suggested that transmission between birds in a shared enclosure may be rare. They also supported conclusions made by other investigators¹¹ who evaluated the molecular diversity among isolates from cohoused birds and found that sources other than bird-to-bird transmission may be important.

The discordance between exposure to an infected enclosure mate and subsequent development of disease has implications for population health management. Quarantine and depopulation methods based simply upon exposure to a bird in which mycobacteriosis was diagnosed may not be warranted. Additional temporal, host, and environmental factors that modify risk for individual birds should be considered when developing disease control plans for an avian collection.

Exposure to diseased enclosure mates occurred in all taxonomic groups in the source population. Significant differences in prevalence across exposed taxonomic groups were not evident, and the 95% CIs for all taxonomic prevalence estimates overlapped (Table 1); however, sample sizes became too limited to provide powerful comparisons when the study population was subdivided into 22 different taxonomic orders. Across taxonomic orders, prevalence did not correlate with the proportion of birds in the source population that were ever exposed; therefore, exposure to infected and potentially mycobacteria-shedding birds by itself did not explain species prevalence patterns. In another study,¹² the prevalence of mycobacteriosis in wildfowl was higher in species that feed through diving and dabbling than in those that graze for food on the ground, in species that were not native to the temperate climate in which they were housed than in those that were native, and in marine- and arboreal-dwelling species than in freshwater- and ground-dwelling species. In the study population for the present investigation, no such overall trends in life history characteristics were evident; feeding patterns, native clime, and habitat types varied among groups in which the prevalence of disease was high. Differences in immune mechanisms against and species susceptibility to mycobacteriosis have also been proposed to explain prevalence variations among groups of birds^3,13 but could not be assessed in our study. Expansion of the study to include known exposed birds from other zoo collections might provide a means to examine taxonomic differences, while controlling for differences in exposure opportunity.

Multivariable modeling identified factors important in describing infection patterns that persisted across species and age groups of birds. The most substantial increase in odds of mycobacteriosis was for birds imported from outside sources. Imported birds may experience stress related to shipping or differences in veterinary care or may have differences in exposure histories. When imported birds were further classified by hatch type, the effect was most pronounced for birds hatched in a zoo rather than in the wild. While this observed relationship may indicate an increased risk of infection around the time of hatching, the combining of hatch type with import status was not included in the final model because the significant relationship was based on data from sources external to the present study and could also reflect systematic differences in bird management across different institutions. Decreasing the number of imports or emphasizing quarantine and surveillance among imported birds may be important measures for reducing the overall incidence of disease.

Young birds are reportedly susceptible to mycobacterial infection,^14,15 and in a survey of pet birds,¹⁶ > 30% with mycobacteriosis were < 1 year of age. Most studies^1,5,12 and literature reviews^3,4 have revealed most diagnoses of mycobacteriosis are made in adult birds, but this higher likelihood of infection is often attributed to a longer follow-up period, providing more time for clinical expression of this slowly progressive disease.³ In our study, the median follow-up time since the first known exposure was 473 days (mean, 742 days) for birds exposed during their first year of life and 501 days (mean, 922 days) for birds that were exposed at later ages, indicating that opportunity for disease development was similar for both groups. Pedigree information for cases, controls, and infected enclosure mates was not complete enough to determine whether the association between disease and exposure at an early age was related to infected parents (via vertical or horizontal transmission through a high degree of contact) or other infected conspecifics. Whereas additional studies are warranted to better clarify the relationship between age and risk of developing mycobacteriosis, prevention of exposure for young birds may be an important control point.

Exposure to an infected bird of the same species was more common among cases than controls in the present study. The increased risk related to conspecific exposure may have resulted from higher degrees of contact within species than between species. However, the findings could also implicate exposure to a common environmental source of mycobacteria that is more readily acquired by a particular species (because of differences in species susceptibility) or more readily encountered (because of the tendency for conspecifics to be housed together or to congregate in particular areas of their enclosure). Matching by taxonomic group and age may have led to an underestimation of the true magnitude of this effect by making exposure to conspecifics more similar for cases and controls. Targeting disease surveillance and monitoring efforts toward same-species contacts of birds with mycobacteriosis may help to identify birds that are at a high risk of developing disease in the future.

In the present study, cases were more likely to have been exposed in small enclosures than were controls. The matching strategy may have led to an underestimation of this effect by indirectly minimizing variation in enclosure sizes within case and control sets. Smaller enclosures may have facilitated an increase in contact (or time spent) with either shedding enclosure mates or an area contaminated with mycobacteria. Differences in sizes of enclosures may also subject birds to different degrees of stress or monitoring by keepers, different environmental characteristics (eg, enclosure substrate and water source), and different methods of bird management (eg, birds in smaller enclosures may be more intensely managed). Most of these characteristics could not be evaluated in our study, but birds exposed to infected enclosure mates in small enclosures appeared to be moved more often than birds exposed only in larger enclosures, although the difference was not significant (P = 0.060). Results of our previous study⁵ suggested the number of moves a bird underwent was a significant predictor of developing mycobacteriosis.

Although duration of exposure to infected enclosure mates was a significant risk factor for developing mycobacteriosis, the clinical relevance of this factor for managing disease is questionable because of the small effect size and low incidence of mycobacteriosis in our bird population. In the matched case-control analysis, each species and age group of birds may have had a different risk of developing disease, but the β coefficients and ORs derived from the multivariable model can be used to estimate general effects of making management changes.¹⁷ For example, given the overall estimated incidence for the exposed cohort (0.008 cases/y), the incidence would be 1.87 times as high (0.015 cases/y; 1.5 cases/100 birds) among birds housed together for 1 year and 0.011 cases/y (1.1 cases/100 birds) for those paired for 6 months. This would be < 1 additional case among 100 birds housed in this manner. However, for groups of birds with a higher incidence of disease, this absolute increase in number of cases might be less acceptable. For example, if the incidence was 10 times as high (0.08 cases/y), then the number of cases would increase to 15 cases/100 birds (0.15 cases/y) with a year's shared exposure. The effect of cumulative exposure on incidence could not be attributed to a sole transmission route in our study. These findings suggested that short periods spent with diseased birds or environments in which diseased birds are housed may not substantially impact the number of new cases of mycobacteriosis in populations in which the incidence of the disease is low. This is an important area for future research because temporarily pairing an infected and uninfected bird for breeding might both preserve valuable genetic diversity as well as increase the size of rare or endangered bird populations. This proposed management option could be particularly useful if the offspring could be protected from exposure to disease. To develop better predictions for risk of developing avian mycobacteriosis, cohort studies are needed with longitudinal, time-dependent analyses in which risk and timing of exposure are incorporated.

Interestingly, no significant associations were detected for exposures during intervals in which shedding was predicted to be high, such as sharing of an enclosure with a diseased bird immediately before its diagnosis was made or with birds in which intestines contained severe lesions or high numbers of mycobacteria. This lack of associations further suggested that infection may not be easily transmitted between enclosure mates. Recent molecular epidemiologic studies^11,18 revealed that the environment rather than an infected enclosure mate may be an important source of exposure because multiple genetic strains of mycobacteria were associated with single outbreaks of mycobacteriosis. The ability to persist for long periods in the environment, which is common among mycobacteria,^19,20 could also lead to cumulative or prolonged exposure following high, short-term environmental contamination. Mycobacteria present in the environment may have accounted for the numerous infections among unexposed birds in the source population in the present study and may have contributed to infections among birds in the study population.

Use of a matched case-control study nested within a cohort study allowed a focused investigation of the effects of direct contact with diseased birds. The low disease prevalence and incidence across multiple taxonomic groups limited the analysis to data from 79 case-control sets. The use of multiple controls per case and risk-set sampling methods²¹ (ie, matching cases to controls that were at-risk at the same age as their matched case) helped to maximize statistical efficiency and power, while minimizing the potential for misclassification among the controls. However, additional studies in other settings are needed to expand the number and range of evaluated characteristics of infected birds. If some controls had unrecognized infection at the time they were selected as controls, the resulting misclassification error would tend to bias identified associations toward the null hypothesis. Therefore, the magnitude of risk may have been underestimated. If a large portion of controls were misclassified in this way, it could mean that the risk factors we identified may emphasize factors associated with the progression of disease (ie, causing mild infections to develop into more severe, detectable disease) rather than acquisition of infection.²¹

Overall, the results of the present study suggested zoos should avoid use of blanket postexposure culling and quarantine policies that would negatively affect the genetic contributions of rare and endangered species because many birds exposed to mycobacteriosis are likely to remain uninfected. The effect of disease control efforts aimed at mitigating risk associated with management, environmental, and individual characteristics should be explored in addition to those aimed at preventing direct contact with infected birds.