Statistical Modelling of Waning Immunity After Shanchol™ Vaccination: A Prospective Cohort Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript describes a longitudinal study of the antibody response to the cholera vaccine Shanchol. The study is well done and the methods and results are clearly explained. I just have a couple of minor comments:

1. Line 97: Please be more definitive on what is meant by "recent" vaccination or illness.
2. Line 253: The authors state "...generally remained above the positivity threshold at day 90." This does not look to be true from Figure 2. The model prediction appears to drop lower than the positivity threshold just before 90 days in all cases and many of the measurements also appear to be lower than the positivity threshold.

Author Response

Reviewer #1

Minor comments:

1. Line 97: Please be more definitive on what is meant by "recent" vaccination or illness.

 

Response: We agree and have clarified the timeframe for “recent” in the manuscript. The revised text now reads: “…key exclusions included recent OCV receipt (within the past 12 months), current or recent diarrhoeal illness (within the past 2 weeks), recent antibiotic use (within the past 7 days), pregnancy, or participation in a similar vaccine study.” (see lines 102-103)

 

1. Line 253: The authors state "...generally remained above the positivity threshold at day 90." This does not look to be true from Figure 2. The model prediction appears to drop lower than the positivity threshold just before 90 days in all cases and many of the measurements also appear to be lower than the positivity threshold.

Response: Thank you for pointing this out. We have corrected the wording to accurately reflect the results. The sentence now reads: “…titers increased steeply during the first month, reached their maximum around 5–6 weeks, and then declined gradually, and by day 90 had fallen close to the positivity threshold.” (see lines 300-301)

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript addresses how rapidly vibriocidal antibody responses induced by Shanchol™ peak and wane in an outbreak-prone, cholera-endemic setting with heterogeneous prior vaccination exposure. The study is well organized and strengthened by its prospective cohort design, repeated serologic measurements, and the application of mixed-effects models to capture nonlinear antibody kinetics. Overall, the work is scientifically sound and relevant to public health practice. However, before publication, some clarifications and minor methodological and interpretative refinements are required.

Major points

1. The stated aim is to provide “setting-specific estimates of how rapidly vaccine-induced vibriocidal antibodies peak and wane.” This objective is clearly addressed in the Results and Discussion. However, at several points, the manuscript extends the interpretation toward booster effectiveness and protection durability, which are not directly measured. I suggest  explicitly distinguishing between: Antibody kinetics (the primary outcome), and Immunologic memory or clinical protection (secondary or inferential considerations).
2. The finding that prior vaccination (0, 1, or 2 doses) did not materially alter peak timing or decay rates is interesting and relevant. However, the discussion occasionally implies a lack of immunological memory in adults vaccinated four years earlier. Please clarify that similar kinetics across strata may reflect saturation effects, natural exposure, or ceiling effects in vibriocidal responses. Memory responses may manifest in ways not captured by serum vibriocidal titres alone (e.g., mucosal IgA, memory B cells).
3. The use of mixed-effects models with natural cubic splines is appropriate and well justified. The derivation of half-lives from model-based predictions is innovative and programmatically meaningful. However, it is necessary to clarify: Were random slopes for time considered, or only random intercepts? How sensitive were half-life estimates to knot placement? Were alternative functional forms (e.g., biexponential decay) explored or ruled out?
4. The limitation of follow-up to 90 days is acknowledged, but the manuscript still draws conclusions relevant to booster timing over “months.” However, it is necessary to temper conclusions by emphasizing that: Half-life estimates are derived from early post-peak decay. Longer-term persistence may follow a different kinetic phase (e.g., slower terminal decay).
5. The discussion is well referenced; however, the narrative is occasionally dense and could be more tightly focused on how external studies contextualize the present findings. I suggest grouping studies by vaccine type (killed vs. live), separating pediatric vs. adult evidence more explicitly and clearly stating why differences across studies are expected (age, endemicity, interval since priming).

Minor points

1. Abstract: Check the description of sample size and clarify whether this refers to participants or total observations. Explicitly state that follow-up ended at 90 days.

2. The slower waning of Inaba responses is consistently reported. Briefly discuss whether this has known implications for protection or outbreak dynamics, or whether it remains primarily an immunologic observation.

3. HIV status was adjusted for, but its potential modifying effect on kinetics is not discussed. Even a brief statement noting the lack (or presence) of interaction would strengthen the analysis.

Author Response

Reviewer #2

Major points

1. The stated aim is to provide “setting-specific estimates of how rapidly vaccine-induced vibriocidal antibodies peak and wane.” This objective is clearly addressed in the Results and Discussion. However, at several points, the manuscript extends the interpretation toward booster effectiveness and protection durability, which are not directly measured. I suggest  explicitly distinguishing between: Antibody kinetics (the primary outcome), and Immunologic memory or clinical protection (secondary or inferential considerations).

 

Response: We agree and have clarified this point in the Discussion. We added in lines 404-407: “We emphasise that our study measured the kinetics of the vibriocidal antibody response over 90 days and did not directly assess immunologic memory or protection. While the waning of antibodies can inform booster scheduling, it does not equate to an absence of immunologic memory or a direct measure of clinical protection.”

 

1. The finding that prior vaccination (0, 1, or 2 doses) did not materially alter peak timing or decay rates is interesting and relevant. However, the discussion occasionally implies a lack of immunological memory in adults vaccinated four years earlier. Please clarify that similar kinetics across strata may reflect saturation effects, natural exposure, or ceiling effects in vibriocidal responses. Memory responses may manifest in ways not captured by serum vibriocidal titres alone (e.g., mucosal IgA, memory B cells).

Response: We agree and have clarified this point in the Discussion. We added in lines 417-419: “However, this similarity in kinetics across previously vaccinated and unvaccinated adults may reflect a ceiling in vibriocidal antibody levels or high baseline immunity from endemic exposure, rather than an absence of immunologic memory.”

 

2. The use of mixed-effects models with natural cubic splines is appropriate and well justified. The derivation of half-lives from model-based predictions is innovative and programmatically meaningful. However, it is necessary to clarify: Were random slopes for time considered, or only random intercepts? How sensitive were half-life estimates to knot placement? Were alternative functional forms (e.g., biexponential decay) explored or ruled out?

Response: We have added clarifications in the Methods to address these modeling questions.

In lines 154-155:“We also tested inclusion of a random slope for time; which improved the model fit and was retained.”

And in lines 157-161: “In sensitivity analyses, altering the placement of spline knots (by ±7 days) had a negligible impact on the estimated half-lives. Additionally, a bi-exponential decay model yielded similar half-life estimates for the initial waning phase and did not significantly improve the fit, supporting the use of the chosen spline model.”

Here are the model comparisons:

MODEL COMPARISON: AIC, BIC tests
Models	AIC	BIC	AICc	CAIC
LOG-LINEAR MIXED MODEL (EXPONENTIAL DECAY)	3604.73	3614.64	3604.74	3616.64
QUADRATIC (POLYNOMIAL) MIXED MODEL	3501.89	3516.76	3501.92	3519.76
CUBIC B-SPLINES (RANDOM INTERCEPT)	3464.36	3489.13	3464.41	3494.13
CUBIC B-SPLINES (RANDOM INTERCEPT & SLOPE)	3460.02	3474.89	3460.04	3477.89

 

 

3. The limitation of follow-up to 90 days is acknowledged, but the manuscript still draws conclusions relevant to booster timing over “months.” However, it is necessary to temper conclusions by emphasizing that: Half-life estimates are derived from early post-peak decay. Longer-term persistence may follow a different kinetic phase (e.g., slower terminal decay).

Response: We agree and have further tempered our conclusions regarding durations beyond the 90-day follow-up. In the Discussion’s limitations section, we already noted that projections beyond 90 days are model-based extrapolations. We have now made this caution more explicit by adding that “Our half-life estimates pertain only to the initial, early decay phase.” (see line 467). We explicitly acknowledge that antibody waning could enter a slower “plateau” phase after the 90-day period, (see line 473) which we could not observe.

4. The discussion is well referenced; however, the narrative is occasionally dense and could be more tightly focused on how external studies contextualize the present findings. I suggest grouping studies by vaccine type (killed vs. live), separating pediatric vs. adult evidence more explicitly and clearly stating why differences across studies are expected (age, endemicity, interval since priming).

Response: We have restructured parts of the Discussion to improve clarity and flow, following the reviewer’s suggestions: 1) see lines 374-376 “Notably, Inaba titers waned more slowly than Ogawa titers. While this serotype difference was modest and has been observed in other studies, it likely represents an immunologic phenomenon without a clear impact on protection or outbreak dynamics.” 2) We elected to maintain the logic in lines 385-398 as this makes it easier in our reasoning in constrasting live vs killed. Same for separating pediatric vs. adult evidence as in lines 418-433.

 

Minor points

1. Abstract: Check the description of sample size and clarify whether this refers to participants or total observations. Explicitly state that follow-up ended at 90 days.

Response: We have revised the Abstract for clarity. 225 participants (not “225 samples/participants”), with a breakdown by prior dose group. We also explicitly noted the 90-day follow-up duration in the Abstract. For example, we added a phrase in the Methods portion of the Abstract: “…with serum collected at baseline and days 14, 28, 60, and 90 (end of follow-up).”

 

2. The slower waning of Inaba responses is consistently reported. Briefly discuss whether this has known implications for protection or outbreak dynamics, or whether it remains primarily an immunologic observation.

Response: We have added a brief discussion of the observed slower waning of Inaba vibriocidal titers compared to Ogawa. In the Results, we already noted the longer half-life for Inaba; in the Discussion, we now expand on this point (see lines 374-376) “Notably, Inaba titers waned more slowly than Ogawa titers. While this modest difference has been observed before, it likely represents an immunologic quirk; there is no clear evidence yet that it yields a major difference in clinical protection or alters serotype-specific cholera patterns.”

 

3. HIV status was adjusted for, but its potential modifying effect on kinetics is not discussed. Even a brief statement noting the lack (or presence) of interaction would strengthen the analysis.

Response: We have included a brief statement regarding HIV status in the Discussion (limitation lines 476-479) “Additionally, we did not observe a significant difference in antibody kinetics between HIV-positive and HIV-negative participants (no significant time×HIV interaction), although our study was not powered for a definitive subgroup analysis.”

Reviewer 3 Report

Comments and Suggestions for Authors

Journal: Vaccines

Manuscript ID: vaccines-411076

Title: Statistical Modeling of Waning Immunity after Shanchol™ Vaccination: A Prospective Cohort Study

 

This work consists of a prospective cohort kinetics analysis in Lukanga Swamps (Central Province, Zambia), enrolling adults (18–65 years) stratified by prior Shanchol™ exposure (0, 1, 25 or 2 previous doses). All participants received two Shanchol™ doses 14 days apart, with 26 serum collected at baseline and days 14, 28, 60, and 90. Ogawa and Inaba vibriocidal titres were measured using a complement-based assay and analysed on the log10 scale. Serotype-specific mixed-effects models with natural cubic splines for time (knots: 14, 28, 60 29 days) assessed trajectories by prior-dose strata, adjusting for age, sex, and HIV status. Peak timing and post-peak half-life were derived from model-based predictions with participant-level bootstrap CIs (1,000 replications).

Modelled titres for both serotypes rose steeply after vaccination, peaking around day 36–35 37 across prior-dose strata. Ogawa titres reached half of peak by about day 73–78, corresponding to post-peak half-lives of 37–41 days; Inaba declined more slowly with half-lives  of 42–46 days.

In this cholera-endemic adult population, Shanchol™ induced vibriocidal responses that peaked at ~5 weeks and waned over the following 5–7 weeks, with broadly similar kinetics regardless of prior vaccination and slightly slower decay for Inaba than Ogawa.

The goal was to characterize the vibriocidal antibody kinetics over 90 days following oral cholera vaccination in a previously vaccinated cholera-endemic population. Comparing these antibody kinetics between individuals with different prior OCV exposure (e.g. zero, one, or two doses received in the past) can reveal whether previously vaccinated persons mount more rapid or durable responses upon boosting than OCV-naïve persons.

Ethical approval was approved on the 26th of December 2016. This work started a decade ago.

Vibriocidal antibody kinetics against V. cholerae Ogawa and Inaba were modelled using repeated-measures mixed-effects models, complemented by model-based estimates of antibody half-life.

Time since first vaccination (in days) was entered flexibly using a natural cubic spline with internal knots at 14, 28, and 60 days, chosen a priori to capture the rapid rise and early decline in titres within the first three months

MAJOR

This study is a longitudinal follow-up with repeated vibriocidal measurements at clearly defined post-vaccination time points (baseline, days 14, 28, 60, 379 and 90), characterizing serotype-specific kinetics and to summarize clinically interpretable parameters (timing of peak response and post-peak half-life) rather than relying only on single time-point contrasts.

The prospective study was well-defined, the statistical analysis is appropriate, it is well justified, the modeling captured the antibody titres for both Inaba and Owaga for a period of 90 days, the results are sound. The Discussion points out the strengths and weaknesses of the study.

I missed the equations of the Maximum Likelihood estimates.

References are appropriate.

My only concern is about the role of asymptomatic carriers and that authors claim that their results cannot be generalized.

Asymptomatic carriers of cholera are individuals infected with Vibrio cholerae who show no symptoms but still shed the bacteria in their feces, playing a crucial role in sustaining transmission and initiating epidemics, especially in endemic regions where they can comprise a large portion (up to 80%) of infected people, continuing to contaminate water and food sources for days to weeks after infection. While symptomatic cases are obvious, these silent carriers are vital for the bacteria's survival and spread, making them significant public health concerns. 

Some references of asymptomatic carriers in cholera transmission are given below:

1. Dizon J.J., et al. Studies on Cholera Carriers. Bull. World Health Org. 1967, 37, 737-743.
2. Wang, J. Mathematical Models for Cholera Dynamics—A Review. Microorganisms 2022, 10, 2358. https://doi.org/10.3390/microorganisms10122358
3. Ansari N, Ozgur S S, Bittar N, et al. (January 24, 2024) An Interesting Case of Asymptomatic Cholera in the Setting of Large Bowel Obstruction.

 

4. Andrews, J.R.; Basu, S. The transmission dynamics and control of cholera in Haiti: An epidemic model. Lancet 2011, 377, 1248–1255.

 

• Relying only on symptomatic cases misses a large part of the infected population, making it harder to track the true spread and plan interventions like vaccinations, note this NIH study and this Beacon Bio report. 

 

 

 

Author Response

Reviewer #3

 MAJOR

1. I missed the equations of the Maximum Likelihood estimates.

Response: We did not report the MLE for the chosen cubic B-spline random intercept and slope model – our focus was to use the MLE to estimate the wanning parameters. See the Stata 19 output below:

mixed ln_ogawatiter c.day s_day* || id: c.day, covariance(unstructured)

 

note: s_day1 omitted because of collinearity.

 

Mixed-effects ML regression                          Number of obs    =  1,049

Group variable: id                                   Number of groups =    225

                                                     Obs per group:

                                                                  min =      1

                                                                  avg =    4.7

                                                                  max =      5

                                                     Wald chi2(2)     = 161.34

Log likelihood = -1727.0097                          Prob > chi2      = 0.0000

 

--------------------------------------------------------------------------------

 ln_ogawatiter | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]

---------------+----------------------------------------------------------------

           day |   .0384338   .0032357    11.88   0.000     .0320919    .0447756

        s_day1 |          0  (omitted)

        s_day2 |   -.066939   .0052813   -12.67   0.000    -.0772901   -.0565878

         _cons |   3.421683   .1093354    31.30   0.000     3.207389    3.635976

--------------------------------------------------------------------------------

 

------------------------------------------------------------------------------

  Random-effects parameters  |   Estimate   Std. err.     [95% conf. interval]

-----------------------------+------------------------------------------------

id: Unstructured             |

                    var(day) |   3.27e-07          .             .           .

                  var(_cons) |   1.893676          .             .           .

              cov(day,_cons) |   .0007867          .             .           .

-----------------------------+------------------------------------------------

               var(Residual) |   .9563035          .             .           .

------------------------------------------------------------------------------

LR test vs. linear model: chi2(3) = 636.79                Prob > chi2 = 0.0000

 

Note: LR test is conservative and provided only for reference.

 

2. References are appropriate.

Response: We appreciate the reviewer’s positive feedback regarding the references. We have carefully maintained the quality and relevance of our citations in the revised manuscript. In cases where we added new text (for example, in response to Comment 3 below), we have included additional relevant references as suggested by the reviewer. We are glad the references were found to be appropriate, and we will continue to ensure they meet the highest academic standards.

3. My only concern is about the role of asymptomatic carriers and that authors claim that their results cannot be generalized. Asymptomatic carriers of cholera are individuals infected with Vibrio choleraewho show no symptoms but still shed the bacteria in their feces, playing a crucial role in sustaining transmission and initiating epidemics, especially in endemic regions where they can comprise a large portion (up to 80%) of infected people, continuing to contaminate water and food sources for days to weeks after infection. While symptomatic cases are obvious, these silent carriers are vital for the bacteria's survival and spread, making them significant public health concerns. 

 

Some references of asymptomatic carriers in cholera transmission are given below:

1. Dizon J.J., et al. Studies on Cholera Carriers. Bull. World Health Org. 1967, 37, 737-743.
2. Wang, J. Mathematical Models for Cholera Dynamics—A Review. Microorganisms 2022, 10, 2358. https://doi.org/10.3390/microorganisms10122358
3. Ansari N, Ozgur S S, Bittar N, et al. (January 24, 2024) An Interesting Case of Asymptomatic Cholera in the Setting of Large Bowel Obstruction.
4. Andrews, J.R.; Basu, S. The transmission dynamics and control of cholera in Haiti: An epidemic model. Lancet 2011, 377, 1248–1255.

Relying only on symptomatic cases misses a large part of the infected population, making it harder to track the true spread and plan interventions like vaccinations, note this NIH study and this Beacon Bio report. 

Response: We fully agree with the reviewer’s concern about asymptomatic carriers and the limitations of our study’s generalisability. We acknowledge that asymptomatic carriers play a crucial role in cholera dynamics and can constitute a majority of infections in endemic settings. We have revised the Discussion section to acknowledge this important limitation. We added lines 483-490 (“Another important limitation is that our model and analyses considered only reported symptomatic cholera cases. In cholera epidemics, however, asymptomatic carriers (i.e., individuals infected with V. cholerae who do not develop symptoms) can make up a large fraction of all infections and yet continue to shed the bacteria, thereby sustaining transmission (Dizon et al. (1967), Andrews & Basu (2011),  Wang 2022, Ansari N et al 2024 ). This means that the true infection count is higher than reported cases and interventions or predictions based only on symptomatic cases may underestimate the epidemic potential. Again, relying solely on symptomatic case data may limit generalisability to the entire infected population.”

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made appropriate revisions and provided clear and satisfactory explanations in response to my previous comments. These changes have improved the manuscript. I now recommend the publication in its present form.