Early diagnosis of pertussis in adults for outbreak management requires low cut-off levels for single IgA and IgG serological tests. We found that cut-offs of 24 U/ml for IgA and 27 U/ml for IgG led to a specificity of 90% and a sensitivity of 100% and 75%, respectively, during the first 2 weeks of pertussis. In the third week, the sensitivity was 100% for both tests. The sensitivity decreased slightly to 98% for IgA, and 95% for IgG during the first 87 days of clinical pertussis. After acute onset of pertussis, IgA remained above 24 U/ml for a mean duration of 7.2 months, and IgG remained above 27 U/ml for a mean duration of 5.1 months.
Although this study is limited by the relatively small number of subjects, the results are based on a pertussis epidemic in a defined community, with 100% participation. We are not aware of another study in which sensitivities of single IgA and IgG were evaluated in both the pre-clinical and clinical phases of pertussis.
The definition of pertussis was partly based on single IgA and IgG levels, which were also evaluated as diagnostic marker. This may have caused some incorporation bias.[42, 44] However, 42 of the 46 pertussis cases were based on at least 4-fold changing IgA or IgG levels in paired samples (Table 1). The lowest IgG level of the 4 cases identified with a single IgG sample was 376 U/ml and they had coughed between 44 – 263 days. From 3 of these 4 cases we were not able to obtain a third serum sample in week 60 (2 subjects died earlier with pertussis), in order to detect significant change. The fourth subject showed a 3.6-fold changing IgG level. In addition, we did not find other causes of this epidemic of cough in the ideal epidemic circumstances of a convent population with positive cultures for B. pertussis.
In our evaluation of spectrum bias in sensitivity and specificity, we showed that our choice of pertussis cases and controls was sound. Arguably, the use of our 35 internal controls may have led to underestimation of the specificity of single low IgA and IgG levels. Indeed 7 of the 35 internal controls had high IgA levels ranging from 24 – 74 U/ml and 2 internal controls had IgA levels above 74 U/ml. Six internal controls had IgG levels ranging from 27 – 90 U/ml and one internal control above 90 U/ml. We argue that IgA levels above 24 U/ml and IgG levels above 27 U/ml decreased the specificities obtained in the internal controls, compared with the specificities obtained in the external controls (Figure 2). Among all 47 subjects with no cough (Table 1), 21 (45%) had high single IgA and/or IgG levels. Because these 21 subjects with high single IgA and/or IgG levels had a serological indication of infection (antibody boosting) without symptoms, we consider the external control group, with GMC of IgA and IgG levels significantly lower than in the 46 pertussis cases, the better choice for calculating specificities.
The retrospective part of the study may have resulted in recall bias regarding onset of cough. The study however, was supported by a nun who was responsible for looking after all the nuns with pertussis during the epidemic. This registered nurse kept reliable clinical records including the history of coughing before and after the start of the study. We are therefore fairly confident that recall bias is limited.
IgA antibodies to B. pertussis antigens in whole-cell sonicate is known to lack specificity compared to IgA antibodies to pertussis toxin. Indeed, single high values of IgA and IgG antibodies to pertussis toxin indicate infections in adults, and IgA is more indicative of a recent antibody response, although less consistent than IgG. However, we excluded causes of the epidemic other than pertussis and used culture, PCR, and IgG against pertussis toxin to positively identify pertussis cases. In our external control group, the levels of IgA against whole cell sonicate showed higher specificities than similar IgG levels. In the original Dutch study of IgA antibodies, it was stated that IgA antibodies, which are not induced by vaccination, can be used as a reliable indicator of natural infection with B. pertussis in adults within one week of infection, especially if interpreted in connection with clinical findings. On the other hand it has been postulated that because of the prolonged antibody response, IgA is not such a useful marker for recent infection. We estimated however, that IgA reaches 100% sensitivity and 99% specificity level sooner than IgG, and that IgA and IgG remain above the 99% specificity level for a mean duration of 5.5 and 4.3 months, respectively, after the onset of pertussis.
IgG antibodies against the virulence factors pertussis toxin, pertactin, and fimbriae increase and decrease after both natural infection and vaccination.[31, 47–49] In bacteriologically proven pertussis cases, IgG antibodies declined more rapidly than IgA. This was confirmed in our study. IgG levels of at least 25 IU/ml were associated with B. pertussis infection in a previous study. In our study, a cut-off level of 27 U/ml for IgG resulted in a sensitivity of 100% the third week after the onset of pertussis symptoms, and remained at 97% up to the 13th week with a specificity of 90%.
We determined a sensitivity of 90% for 50 U/ml IgG and of 80% for 100 U/ml, which is comparable to a previous evaluation that determined a sensitivity of 89% for IgG levels above 50 U/ml and 76% for IgG levels of ≥ 100 U/ml. Also, our specificity of 99% for an IgG level of 90 U/ml is comparable to that found in the prior study which concluded that, independent of age, a cut-off level of 100 U/ml IgG showed a specificity of 99–100%. In the prior study, most patients reached IgG levels of 100 U/ml within 4 weeks of disease onset which persisted for 4.5 months. These findings are in line with our estimates that the IgG level increases from the detection limit (5 U/ml) to 100 U/ml in 6.7 days and persists at this level for 4.2 months. The rate of IgA and IgG increase underlines the importance of obtaining acute phase samples early in the disease in order to detect a significant increase, and consequently the importance of significantly decreasing levels for ultimately diagnosing pertussis if the first sample is not obtained early in the disease.[33, 34] Our outcomes are also in line with a study in which a cut-off point of 94 IU/ml for IgG pertussis toxin, with a sensitivity of 80% and a specificity of 93%, has been proposed. This 94 IU/ml is comparable with 76 U/ml in our study and considerable lower then the 125 IU/ml (= 100 U/ml) officially used in the Netherlands. In our study 94 IU IgG pertussis toxin had a sensitivity of 83% and a specificity of 98%.
Our findings for low IgA and IgG levels to diagnose pertussis in outbreak management are supported by findings from a pertussis outbreak in a boarding school in Australia, where IgA against whole-cell sonicate, and IgG against pertussis toxin proved useful for early diagnosis and outbreak management. In that study, it was concluded that the IgG level of 125 IU/ml (100 U/ml) was not sensitive enough to identify pertussis cases in their early stages for outbreak management.
Because IgA is not induced by vaccination against pertussis, it may be preferred over IgG in recently vaccinated subjects, as IgG is induced by vaccination with whole-cell vaccines against pertussis used in the Netherlands. Other vaccines may induce even higher IgG-pertussis toxin levels, since the response to pertussis toxin varies between different whole-cell vaccines and acellular vaccines. These IgG-pertussis toxin levels can reach levels higher than 100 U/ml.[48, 52–54]