We assessed amoxicillin-treated corals from three offshore sites and four inshore sites on Florida’s Coral Reef (Table 2; Figure 1). The offshore sites were 7 – 10.5 km from land, had spur-and-groove morphology, and had corals treated at depths of 2.0 – 9.5 meters. In contrast, the inshore sites were 0.6 – 2.9 km from land, were patch reefs surrounded by sand and seagrass, and had corals treated at depths of 1.8 – 6.0 meters. Inshore reefs had higher coral density and coral species diversity than offshore reefs. Reefs were first visited and treated between January 2019 and August 2021. Time of treatment was based on when permits were authorized as well as when SCTLD outbreaks were observed. Once a reef was initially treated, it was revisited approximately one month afterwards, and then approximately every two months after that for full surveys and treatments if necessary.

SCTLD-affected corals were defined as those exhibiting acute to subacute disease lesions not accompanied by any microbial mat. The corals used in this study were initially treated when SCTLD was in an epidemic stage (which was present on inshore reefs until the 2023 bleaching event) in which reefs were characterized by infections across multiple susceptible species. When an SCTLD-affected coral was found, a numbered cattle tag was affixed for future identification. The coral was also mapped and photographed from multiple angles both before and after treatments were applied. All active SCTLD lesions were treated with 98% amoxicillin trihydrate powder hand-mixed in an 8:1 Base2b:amoxicillin by weight ratio. Mixing was done manually between 1 and 72 hours before application to the diseased corals. The mixed paste was packed into 60cc catheter syringes, then squeezed out in a ~1 cm wide band along the active lesion. The paste was pressed by hand onto the colony so that approximately 0.5 cm of treatment covered the live but diseased tissue and about 0.5 cm was pressed into the recently denuded skeleton to anchor the treatment onto the colony. In rare occasions where the treatment did not adhere well, such as in surge or occasionally on excessively mucous corals, small (~ 0.25 cm) pieces of modeling clay were placed sporadically atop the treatment to anchor it to the coral.

Divers used maps and datasheets with coral tag numbers, species, and directions to find each previously-tagged coral. From February 2019 to February 2022, every coral was visited approximately every two months for a health assessment and retreatment if needed. Each coral was coded as either “T” (treated), “NAD” (no active disease), “DEAD” (newly dead since the last monitoring event), or, rarely, “AL” (active lesions – not treated; this occurred only when the amount of remaining tissue was so small as to not be worth treatment efforts). Beginning in March 2022, methods shifted to allow for broader SCTLD detection and treatment coverage. Two sites – Sombrero Reef (offshore) and Cheeca Rocks (inshore) – retained the established level of intense colony-specific monitoring. At all other sites, divers used a systematic swim pattern to cover the entire reef to detect new lesions while still locating and assessing all previously tagged corals, but live corals without active SCTLD did not receive full monitoring. Corals which did have SCTLD were recorded as T or AL, and any dead corals were recorded as such. At all sites, any newly diseased corals found during each monitoring period were tagged, treated, and added to all future monitoring events. Monitoring time steps were rounded to the nearest number of 30 day months since initial treatment (e.g. a visit 14–45 days after initial treatment was rounded to 1 month, 46–75 rounded to 2 months, etc.).

To assess overall coral survivorship rates, we used the survfit function in the survival package (Therneau, 2024) to fit a global Kaplan-Meyer survival curve using the number of days observed and survivorship. The monitoring data were uninformative right-censored (i.e. corals were followed until they either died or the study ended). We focused on monitoring data for five species for which suitable replicates were available in both habitats: Orbicella faveolata (n = 1079), Montastraea cavernosa (n = 708), Pseudodiploria strigosa (n = 163), Colpophyllia natans (n = 802), and Diploria labyrinthiformis (n = 252). The date each coral was first observed dead was considered its date of death, which may have slightly overestimated survival rates for some corals by up to two months. Assessments of mortality for most sites were conducted from the initial treatment date (Table 2) through June 30, 2023. For Newfound Harbor, assessments past May 30, 2022 were excluded because a presumably different disease swept through Pseudodiploria clivosa colonies at the site (smooth linear margins that affected only P. clivosa colonies though other highly SCTLD-susceptible species were present, and that were unresponsive to treatments), thus confounding SCTLD survival data. We also generated survival curves for each species across each habitat. To assess the effects of coral species and habitat (inshore, offshore) on coral survival, we performed a series of log-rank tests using the survdiff function in the survival package (Therneau and Grambsch, 2000; Therneau, 2022) in R version 4.2.2 (R Core Team, 2022) followed by pairwise log-rank comparisons using the pairwise_survdiff function in the survminer package. The false discovery rate (FDR; Benjamini and Hochberg (1995)) was used to correct for multiple comparisons.

Health status and reinfection rates through time were assessed for all corals that met the following conditions:

For corals with more than one data point per quarter, we used the most detrimental condition as the point of record for that time point: Dead > Treated > Active lesion untreated > No active disease. We then grouped corals by each health status through each time point, and calculated the proportion that were lesion-free from each time point through the end of the first year since initial treatment.

A total of 2,476 corals were assessed for 1-year reinfection patterns after the initial treatment date (Table 2). These were from the five most common species at both inshore and offshore sites, with 1,655 from inshore sites and 821 from offshore sites. To assess how 1) species and 2) habitat (inshore/offshore) affected the likelihood of a coral remaining disease-free for at least one year after one visit (initial treatment only) or two visits (initial treatment + an additional visit and treatment if needed at 1–3 months), we created four generalized linear models (GLMs): one with species only, one with habitat only, one with both but without an interaction effect, and a final ‘full’ model that also included an interaction between species and habitat (Supplementary Table 1). We used a model selection approach (Anderson, 2008) leveraging Akaike’s Information Criterion (AIC; Akaike (1973)), to choose the optimal model for inference, selecting the model with the lowest AIC score (and thus best fit). This model was used to assess whether habitat or species were significant predictors of reinfection rates. We assessed predictive power by splitting the data into a 70/30 train/test split measuring area under the curve (AUC) of the predictive values. Statistical analyses were conducted in R v.4.4.1 (R Core Team, 2022) using the stats package for generalized linear models as well as the packages ROCR for the AUC (Sing et al., 2005), emmeans for post hoc analysis (Lenth, 2018) and multcomp for group significance designations (Hothorn et al., 2008).

Overall coral survivorship as assessed by the global Kaplan-Meyer curve was 92% at one year, 89% at two years, and 86% at three years. Survival rates did, however, vary between habitats (inshore vs. offshore) and across species (Figure 2).

Survival differed significantly by both habitat (χ² (3004) = 15.94, p = 0.0001) and species (χ² (3004) = 447.69, p < 0.0001) when considered globally. The impact of habitat on survivorship was mediated by species. Specifically, differences in survival rates between habitats were driven by two species – O. faveolata (χ² (1079) = 8.93, p = 0.003) and C. natans (χ² (804) = 26.59, p < 0.0001) – which both had higher survivorship inshore than offshore (Figure 2). The other three species did not have significantly different survivorship between habitat types, though D. labyrinthiformis trended towards significance (χ² (252) = 3.44, p = 0.06). Survivorship of all species at inshore reef sites remained above 60% through 1200 days (3.25 years). At offshore reef sites, O. faveolata, M. cavernosa, and P. strigosa had survival rates above 60% at 1600 days (4.4 years), but only about 1/3 of C. natans and D. labyrinthiformis survived that long.

Survivorship between species differed significantly, both globally and within habitat types (Figure 3). Specifically, at inshore sites, O. faveolata and M. cavernosa had significantly higher survivorship than all other species (all p < 0.00001; see Supplementary Table 2 for full statistical values), while D. labyrinthiformis had significantly lower survivorship than all other species (all p <0.005). Pseudodiploria strigosa and C. natans had significantly lower survivorship than the top two performing species, and significantly higher survivorship than D. labyrinthiformis, but did not differ from each other. At offshore sites, O. faveolata and M. cavernosa had significantly higher survivorship than all other species (all p < 0.0001), and C. natans and D. labyrinthiformis had significantly lower survivorship than all other species (all p < 0.002). Pseudodiploria strigosa had significantly lower survivorship than the top two performing species and significantly higher survivorship than the bottom two. (Figure 3).

Across all individuals, 36% (897/2476) remained disease free for at least one year after the initial treatment. An additional 18% (for a total of 54% (1336/2476)) remained disease free through at least the first year when new lesions, if present, were treated during the 1–3 month follow-up. The remaining 46% of corals all developed additional lesions after the initial 1–3 months, although the percentage of corals remaining disease-free continued to rise through the first year (Figure 4).

For comparing habitat and species effects on corals remaining disease-free after only the initial treatment, the best model included both variables but no interaction factor (Supplementary Table 1). Both habitat (p < 0.001) and species (p < 0.001) were significant variables, though the model was not particularly predictive (AUC = 0.61).

At inshore reef sites, 36% of assessed corals remained disease free after the initial treatment, compared to only 29% of offshore corals (Figure 5). Globally (across inshore and offshore habitats combined), 48% of M. cavernosa remained disease free, significantly more than C. natans (33%; p < 0.0001), O. faveolata (29%; p < 0.0001), and D. labyrinthiformis (37%; p = 0.02). Additionally, significantly more P. strigosa remained disease free (44%) than C. natans (33%; p = 0.02). Within both inshore and offshore habitats, M. cavernosa colonies were less likely to develop lesions than both C. natans and O. faveolata. When comparing species and habitats together, every one of the five species was less likely to develop additional lesions within the inshore habitat compared to the offshore after initial treatment (all p < 0.0001). Additionally, offshore O. faveolata were more likely to develop lesions than every inshore species (all p < 0.02; Supplementary Table 3), offshore C. natans were more likely to develop lesions than every inshore species except O. faveolata (all p < 0.01), and offshore D. labyrinthiformis were more likely to develop lesions than inshore P. strigosa and M. cavernosa (both p < 0.03).

For assessing the likelihood of a coral remaining disease free for at least one year after the initial treatment plus a retreatment if necessary at 1–3 months, the best model included habitat, species, and the interaction of factors. Again, this model was not particularly predictive (AUC = 0.61). However, habitat was a significant variable, with 55% of inshore corals remaining disease free compared to 44% of offshore corals (p < 0.001; Figure 5). Species was also a significant predictor (p < 0.001), as was the interaction between species and habitat (p = 0.03). On inshore reefs, 67% of M. cavernosa remained disease-free after the initial plus 1–3 month checkups, a value significantly greater than for C. natans (52%; p < 0.0001) and D. labyrinthiformis (53%; p = 0.03). Also on inshore reefs, significantly more O. faveolata remained disease-free (64%) than C. natans (52%; p = 0.007). On offshore reefs, 40% of O. faveolata and 61% of M. cavernosa remained disease free, values significantly higher than for D. labyrinthiformis (27%; p < 0.001 and p = 0.006 respectively). When comparing species and habitats together, inshore O. faveolata and M. cavernosa were significantly more likely to remain disease-free than offshore C. natans, D. labyrinthiformis, and O. faveolata (all p < 0.02). Inshore P. strigosa were significantly more likely to remain disease-free than offshore D. labyrinthiformis and O. faveolata (both p < 0.01). And inshore C. natans were significantly more likely to remain disease-free than offshore O. faveolata (p = 0.001; Supplementary Table 3).