Clinical Relevance of Neutralizing Antibodies in Botulinum Neurotoxin Type A

Abstract

The precise definition of prevalence of neutralizing antibodies (NABs) affords cross-sectional testing of a cohort. But in most studies, only selected patients are tested. This leads to gross underestimation of NAB-prevalence, and the opinion that induction of NABs is a rare phenomenon in botulinum neurotoxin (BoNT)/A-therapy. However, recent cross-sectional studies report annual incidences between 1 and 2% in patients being treated with a complex protein (CP)-containing preparation. This implies that NAB-prevalence above 10% has to be expected in patients being treated for more than 10 years. High dose per session and long duration of treatment are relevant risk factors for induction of NABs. In patients exclusively treated with the CP-free incobotulinumtoxin A (incoBoNT/A) preparation Xeomin® no NAB-induction has been reported so far. In patients with NABs switching to incoBoNT/A may lead to a decline of NAB-titers. In patients with NABs under treatment with a CP-containing BoNT/A-preparation it may take years of treatment until a second treatment failure (STF) becomes clinical manifest. In a cohort of 59 patients with partial STF patients’ reports on the reduction of BoNT-activity predicted the presence of NABs better than treatment related data produced by the treating physicians.

1. Introduction

Botulinum neurotoxins (BoNTs) are bacterial products and large molecules which are usually embedded into even larger complex proteins (CPs) [1, 2, 3, 4]. In clinical practice, BoNTs have to be applied by injection. This causes activation of local dendritic cells, elicits hit-shock proteins, and leads to local inflammation. Therefore, the induction of antibodies (ABs) can hardly be avoided [5]. The analysis of lymphocytes in BoNT/A or BoNT/B long-term treated patients with movement disorders indicates that in most of these patients, T-cells have responded to the BoNT application [6].

Induced antibodies target epitopes of CPs and BoNT. Some of the ABs do not influence the biological activity of BoNTs, and others reduce or neutralize BoNT action [7]. Neutralizing antibodies (NABs) in immune-resistant patients target epitopes of the heavy [8] or the light chain [9] of the BoNT molecule.

In clinical practice, the relevant question arises whether a partial or complete secondary treatment failure (pSTF or cSTF) results from NAB induction.

2. The problem of precise determination of the prevalence of ABs and/or STF

For the determination of the presence of NABs, clinical as well as laboratory tests can be used (for a recent discussion, see [10]). Regardless of which test is addressed, little is known about the test-retest liability or variability of repeated measurements of a single serum or serial measurements within a single patient. Nevertheless, the precise determination of the prevalence of ABs heavily depends on the quality (sensitivity and specificity) of NAB testing.

Furthermore, by definition of prevalence, it is necessary to test all members of a cohort if the prevalence of a certain feature in a special cohort has to be determined. Thus, precise determination of the prevalence of neutralizing antibodies in a cohort of BoNT-treated patients implies that a cross-sectional NAB study has to be performed in this cohort.

In the majority of studies reporting on the prevalence of NABs, no cross-sectional testing has been performed. Instead, antibody rates or antibody frequencies (= number of NAB-positive patients/number of patients in the cohort) are determined resulting from NAB testing of selected patients. This procedure crossly underestimates the presence of NABs (comp. [11]).

In their meta-analysis of NABs in BoNT therapy, Fabbri et al. report on an overall NAB frequency of 3.5% in clinically responding patients. In responding patients with spasticity, only 0.7% were reported to be NAB-positive, in patients with dystonia 6.5%. These data are at clear variance with cross-sectional studies reporting 31 positive patients among 212 responding patients with cervical dystonia who were tested by means of the mouse hemidiaphragma assay (MHDA) (= 14.6%; [12]) and 14.3% MHDA-positive patients in long-term treated patients with spasticity [13].

An even more challenging problem is the determination of the temporal development of NAB prevalence. Antibody rates or antibody frequencies do not give any information on the temporal development of NABs, since this ratio between NAB-positive patients and all members of the cohort does not take into account the duration of BoNT treatment.

To analyze the temporal development of NAB formation, the Kaplan-Meier survival analysis has to be performed, calculating the prevalence of NAB-positive patients among those patients with a given duration of treatment. This approach demonstrates that up to 50% of the patients will become NAB-positive when treatment durations exceed 25 years or higher doses are used. In Figure 1, the probability to remain AB-negative is plotted against the duration of treatment in 595 patients. The prevalence of NABs was 82 out of 594 patients (= 13.8%) in the entire cohort, 7 out of 186 patients (= 3.8%) being treated with doses <500 uDU, 49 out of 312 patients (= 15.7%) being treated with doses between 500 and 1000 uDu, and 26 out of 96 patients (= 27%) when doses larger than 1000 uDU were used. To compare doses of different BoNT/A preparations and to determine unified dose units (uDU), aboBoNT/A doses were left unchanged and ona- and incoBoNT/A doses were multiplied by 3. These conversion ratios have been used by Fabbri et al. 2016 [11] and have been discussed by Contarino et al. [14]. The Kaplan-Meier analysis clearly reveals that there is a nonlinear decline of the probability to remain MHDA-negative with the duration of treatment, especially in the higher dose groups.

Figure 1.

Kaplan-Meier survival curve to remain AB-negative in patients with cervical dystonia being treated with BoNT/A. With increasing duration of treatment beyond 10 years, the probability to remain AB-negative declines rapidly down to values around 50%.

In summary, induction of NABs occurs frequently and inevitably progresses with the duration of treatment as long as complex protein-containing BoNT/A preparations are used. The induction of NABs is probably significantly lower in complex protein-free BoNT/A preparations [5, 15] (see Section 3 below).

The determination of incidence and prevalence of secondary treatment failure (STF) is even more complex. So far, there is no clear definition of STF. If no response to a BoNT injection can be detected neither by the patient nor the treating physician, the diagnosis of a complete treatment failure (cSTF) is comparably easy [16]. However, this is the end stage of a longer process, starting with an increasing reduction of the duration of action of a BoNT injection before the 4-week peak effect becomes reduced. This has been described in 2004 by Dressler in detail [1].

When patients are reinjected every 3 months, the reduction of the duration of action cannot be measured directly but results in a systematic worsening of disease severity (for details see [5, 17]). We, therefore, have proposed a formal definition of pSTF in patients with CD. If a patient has responded with more than three points on the TSUI scale [18] and then develops a systematic worsening over three injection cycles of more than two TSUI score points and reports a reduction of the effect of BoNT injection to head position, tremor, or pain, a pSTF has to be suspected.

Our experience is that such a formal definition of pSTF is helpful to detect patients with pSTF. Furthermore, in a cohort of 32 patients with pSTF according to this definition, 25 patients (= 78%) had a positive MHDA test [4]. This is definitely more than in a large cohort of patients with suspected pSTF without formal definition were only about 50% were MHDA-positive [19].

This led to the opinion that NAB-associated secondary nonresponse (SnR) is a rare occurrence, and SnR is more frequently due to an insufficient dose, inappropriate muscle selection, or improper injection technique or targeting [10, 20]. But to our opinion, these aspects of BoNT treatment do not suggest the development of pSTF but indicate insufficient and inappropriate BoNT treatment. pSTF can only be suspected when the patient worsens despite of therapy optimization.

In a cross-sectional study on 66 patients with CD who had started their BoNT therapy with abo- or onaBoNT/A, we have analyzed patients´ drawing of the course of disease (course of disease graphs (CoDGs)) after the onset of BoNT therapy over the entire duration of BoNT therapy. Five different response types could be distinguished: the rapid or golden responder (RR) type, the continuous response (CR) type, the poor response (PR) type, and the secondary treatment failure (STF) type I and II.

The RR type is characterized by a rapid response after the onset of BoNT therapy, followed by a further less rapid improvement. The CR type is characterized by a continuous improvement over the entire duration of treatment, The PR type is characterized by an improvement of less than 20%. Patients who drew a PR type CoD graph were primary nonresponders. The STF types I and II are characterized by an initial improvement followed by a secondary worsening. In the STF type II, the second period of improvement followed after the switch of the BoNT preparation.

Among the 66 patients, 17 patients (= 25.8%) produced a STF type. In Figure 2A, the mean CoD graph (full line) plus/minus 1 standard deviation (hatched lines of the 11 patients with a STF type I CoD graph is presented. In Figure 2B, the corresponding mean CoD graph (full line) plus/minus 1 standard deviation of the six STF-type II CoD graphs are presented.

Figure 2.

The severity of cervical dystonia versus percentage of time elapsed from treatment onset. A. Mean course of disease graph (CoD graph; solid line) plus 1 standard deviation range (hatched lines) of 11 CD patients who had drawn a STF type I CoD graph after the onset of BoNT therapy. The severity of CD rapidly decreases initially but then worsens again. B. Mean course of disease graph (CoD graph; solid line) plus 1 standard deviation range (hatched lines) of six CD patients who had drawn a STF type II CoD graph after the onset of BoNT therapy. The severity of CD improved initially, then worsened again, but improved a second time after the switch of the BoNT preparation.

Figure 3.

Time to paralysis versus TSUI. A. Temporal development of the relation between TSUI scores and paralysis times in 2010, 2013, and 2017 in nine CD patients in whom the complex protein-containing preparation had not been switched, although the initial MHDA test was positive (paralysis time > 60 mins). Open circles indicate values of the investigation in 2010, full circles indicate values of the investigation in 2017, and no circles indicate values of the investigation in 2013. B. Temporal development of the relation between TSUI scores and paralysis times in 2010, 2013, and 2017 in nine CD patients in whom the complex protein-containing preparation was switched to incoBoNT/A in 2010. Open squares indicate values of the investigation in 2010, full squares indicate values of the investigation in 2017, and no squares indicate values of the investigation in 2013. Apart from one exceptional case, paralysis times of eight patients decrease between 2010 and 2017. In six out of nine patients, TSUI scores improve, but the improvement of the TSUI score is less pronounced compared to the improvement of the paralysis times.

The same task (to draw the course of disease after the onset of BoNT therapy) was analyzed in 34 patients who had exclusively been treated with incoBoNT/A. No patient in the incoBoNT/A monotherapy group produced a STF type I or II CoD graph. This difference in frequency of drawing STF type I or II graphs is significant (p < .05).

In summary, partial secondary treatment failure (pSTF) occurs more frequently than positive NAB tests suggest. This indicates that NAB tests are less sensitive to detect pSTF than careful clinical investigation and patient’s assessment of the efficacy of BoNT therapy.

3. Implications of the presence of NABs on long-term outcome

As mentioned, earlier, the presence of NABs does not imply that there is no clinical response at all. We have recently described a small cohort of nine CD patients with positive NAB testing in 2010 who did not want to be switched to another BoNT preparation. They were retested in 2013 and 2017. Their relation between TSUI scores and paralysis times in 2010, 2013, and 2017 are presented in Figure 3A (open symbols indicate data from 2010, full symbols indicate data from 2017, and missing symbols indicate data from 2013). Apart from two exceptional cases, little changes in paralysis times and TSUI scores can be observed.

The data of these nine CD patients in whom BoNT/A preparation had not been switched between 2010 and 2017 were compared to data of nine CD patients in whom the complex protein-containing BoNT/A preparation was switched to the complex protein-free incoBoNT/A preparation in 2010. The relation between TSUI scores and paralysis times of these nine switchers is presented in Figure 3B. Apart from one exceptional case, paralysis times improved in the switchers. In parallel, the TSUI scores also improved in six out of nine patients, but the improvement of TSUI scores was less pronounced than the improvement of the paralysis times (see Figure 3B).

In summary, so far little is known about the development of NAB titers when patients remain on their BoNT preparation under which they have developed NABs. However, there is increasing evidence that switching from a complex protein-containing preparation (abo- and onaBoNT) to a complex protein-free preparation (incoBoNT/A) may lead to a significant long-lasting improvement of paralysis times [4, 21].

4. No NAB induction in patients under incoBoNT long-term monotherapy

Soon after the treatment of patients with the old “Botox,” it became obvious that in a large percentage of patients, NABs were induced [22]. This led to a purification process of the “old” Botox® preparation, a reduction of the protein load of the “new” Botox® by a factor of 5–6 [10, 23], and an improvement of the antigenicity of the Botox® preparation [24]. The incidence of NAB induction reported for the “new” Botox® preparation was around 1%/year [24].

In 2005, incoBoNT was licensed for the treatment of CD [25]. By removal of botulinum neurotoxin complex proteins and elimination of biological inactive fragments of the BoNT molecule, the protein load of this preparation was reduced down to 0.55 ng compared to 4.8 ng of the Botox and 5.3 ng of the Dysport® preparation [3, 26]. But so far, no convincing study has been presented that the lower protein load of the incoBoNT preparation also leads to a significantly lower antigenicity compared to the other two BoNT/A preparations licensed in Europe.

However, it has not been reported that NABs or pSTF were induced in a patient who had exclusively been treated with incoBoNT/A [15]. Cases with NABs have been presented who had been treated only for a few treatment cycles with abo- or onaBoNT and then were switched to incoBoNT/A [1, 15], but NAB induction under incoBoNT/A monotherapy has not been observed.

This is in line with a recent observation that in 34 CD patients who were exclusively been treated with incoBoNT/A, no patient had a positive MHDA test [15] and no patient drew a STF type I or II CoD graph (see Section 2 above).

In summary, the antigenicity of the complex protein-free incoBoNT/A preparation appears to be very low, since no NAB induction has been observed in patients who have exclusively been treated with incoBoNT/A.

5. Prediction of the NAB presence from clinical treatment-related data

It has been demonstrated that long-term treated CD patients with NABs have a significantly worse TSUI score, were treated with significantly higher doses, and have higher pain scores of the CDQ24 than long-term treated CD patients with a negative MHDA [12, 27, 28]. The paralysis times of the MHDA are significantly correlated with the actual doses, the actual TSUI scores, and the CDQ24 pain scores [28].

To analyze which treatment-related parameter may be used to predict the presence of NABs best, NABs were determined in 59 patients with pSTF. Patients ‘assessments of the effect of long-term BoNT/A treatment were determined by asking patients to estimate the improvement of CD in percent of the initial severity of CD at the onset of BoNT therapy (IMPQ) and to mark the actual severity of CD in percent of the initial severity on a visual analog scale which yielded a second estimation of improvement (IMPD). The receiver operating characteristics (ROC) curves for the prediction of the presence of NABs by IMPQ and IMPD are presented in Figure 4A. The sensitivity and specificity of both parameters were around 0.7–0.8.

Figure 4.

The ROC curves for prediction of the presence of NABs by IMPQ and IMPD. A. ROC curves analyzing the relation between the presence of NABs and patients´ assessments of the improvement of CD since the onset of BoNT therapy: Solid circles indicate the ROC curve of IMPQ and light circles indicate the ROC curve of IMPD. For both parameters, sensitivity and specificity lie around 0.7–0.8. B. ROC curves analyzing the relation between the presence of NABs and treating physicians’ scoring of the actual severity of CD (ATSUI) and the improvement of CD since the onset of BoNT therapy (IMPTSUI): Solid squares indicate the ROC curve of ATSUI and light circles indicate the ROC curve of IMPTSUI. For both parameters, the sensitivity is around 0.5, whereas the specificity lies around 0.7–0.8.

The treating physicians scored the actual severity of CD by means of the TSUI score (ATSUI) and determined the improvement since the onset of BoNT therapy by calculating the difference between ATSUI and the initial severity of CD at the onset of BoNT therapy (IMPTSUI) in the same 59 patients. Similar ROC curves were calculated for ATSUI and IMPTSUI (Figure 4B). Treating physicians´ scoring predicted the presence of antibodies less well compared to the assessment of the patients. Sensitivity was lower than 0.6, and specificity was also between 0.7 and 0.8.

In summary, patients realize the NAB-induced reduction of the efficacy of BoNT injections quite well, probably better than treating physicians scoring of the treatment effect.

6. Conclusion

Induction of NABs occurs frequently (Figure 1), may become manifest after years of successful treatment (Figure 3), progresses with the duration of treatment (Figure 1), and has clinical implications. In patients with CD, it goes along with higher severity of CD, leads to more pain, and affords treatment with increasingly higher doses. Patients realize the reduction of efficacy of BoNT/A treatment quite well (Figure 4).

Since the induction of NABs has not been observed under monotherapy with incoBoNT/A and switch to incoBoNT/A may lead to clinical improvement in patients with pSTF after ona- and aboBoNT/A incoBoNT/A seems to have a very low antigenicity. We, therefore, recommend using the complex protein-free BoNT/A preparation incoBoNT/A from the very beginning of BoNT/A therapy to reduce the risk of antibody formation as low as possible.

Acknowledgments

We would like to thank the following individuals for their expertise and assistance throughout all aspects of our studies especially Prof. Dr. Philipp Albrecht for making substantial conceptual and design contributions, Dr. Marek Moll, Beyza Ürer, Raphaela Brauns, and Dietmar Rosenthal for contributing to gathering data conducting the data analysis and creating the tables and figures.

Conflict of interest

The authors declare no conflict of interest.