Among matched stone patients with equal probability of spontaneous passage or early intervention, treatment failure rates were similar with both approaches. While we found no difference in the primary outcome, early intervention was associated with increased morbidity, evidenced by substantially more ED revisits and hospitalizations. Early intervention was also associated with higher system utilization: by definition, 100% of early intervention patients had stone intervention, with 21.8% requiring a rescue procedure or hospitalization within 60 days. In contrast, 21.8% of matched controls required intervention or hospitalization during their trial of passage.
Acknowledging that intervention benefits patients with refractory symptoms and large or infected stones , our findings infer that there are many patients for whom procedure-related morbidity exceeds that related to spontaneous passage. Previous studies confirm that ureteroscopic intervention carries morbidity [7, 18, 19, 21], and our data show that one-third of intervention patients experienced difficulty sufficient to trigger ED revisits, a 10.6% increase relative to controls. Our experience is that they did so because of clot or stent pain, post-op ureteral edema or spasm, incomplete stone removal, infection or, frequently, ongoing pain concerning to patients who expected their symptoms to resolve rapidly after intervention. The fact that intervention patients suffered higher ED revisit and hospitalization rates than matched controls suggests that intervention is being employed not only in patients who benefit, but also in patients who might do well with trial of spontaneous passage.
Other researchers have used matching approaches to compare conservative with interventional management. Hollingsworth studied insurance data, finding that intervention was associated with higher costs, similar post-index hospitalization rates and fewer ED visits . Because data were insurance-based, modeling covariates included age, employment, insurance type and region. Important determinants like stone size, location and hydronephrosis were unaccounted for, and it is unlikely these were equally distributed across the medical and surgical groups. Further, this study excluded women and patients who underwent stenting, which is a common cause of morbidity. Dauw used propensity-matching to compare early intervention with medical expulsive therapy, finding that disability and costs were higher for intervention patients, but this study also lacked stone-related covariables, instead matching patients on age, sex, insurance type, region, employment and Charlson score .
Our study is concordant with population-level data showing that ED revisits and readmissions (morbidity markers) increase in parallel with early intervention rates [15, 16]. Based on these findings, physicians can advise patients that surgery is not always a better option, despite its intuitive appeal. Although this study indicates that some patients will experience worse outcomes with early intervention, it does not clarify which patients may benefit and which will not. We found that stone size and (proximal) location were the main determinants of treatment failure (supplemental material, Appendix Table 1), therefore, refer readers to Canadian and European guidelines that propose 5 mm and 6 mm treatment thresholds, respectively [9, 11]. Adopting this approach would identify 11–21% of patients as potential early intervention candidates . We also feel that a history of prior successful spontaneous stone passage, favorable response to ED treatment, distal stone location, and patient preference for non-intervention should be considered in emergency management decisions.
Strengths and limitations
We studied a large consecutive sample from nine hospitals in two cities, including all patients who, based on stone size, practice guidelines, and site preferences, might be exposed to either treatment approach. Our findings apply to patients with 2.0–9.9 mm ureteral stones in primary care and emergency settings. They are not relevant to patients who have already failed a trial of spontaneous passage. Although we were able to incorporate the main determinants of stone passage in our analyses, our retrospective design precluded detailed description of patient comorbidities. We studied an interventional approach, not a single procedure, and outcomes may differ based on the operator and specific procedure performed. Notably, we did not isolate the effect of ureteral stenting. Our methods mimic those of Shah et al.  who defined passage success by the absence of subsequent intervention, but we did not verify stone passage, nor assess patient experience, quality of life or functional status. We defined adverse events based on ED revisits and hospitalizations, which are markers of patient distress [15, 16], but some patients may have suffered significant morbidity without seeking hospital-based care. Men were over-represented in our sample, because they more often undergo CT imaging .
Propensity matching is an effective way to balance control and intervention groups and reduce treatment selection bias, but it falls short of randomization, particularly if there are important unmeasured covariables. Propensity matching also depletes sample size, which is unfortunate but unavoidable, because the treatment groups have many members for whom conditional probability of intervention differs substantially (Fig. 2). Despite sample size shrinkage, our final sample is large relative to most renal colic studies, it exceeded our sample size estimate, and cohorts were well balanced on all important determinants. Finally, this research is subject to the limitations of observational research, and a randomized trial would be helpful to confirm our findings. Such a trial should assess cost-effectiveness, quality of life and functional status, clarify patient subgroups most likely to benefit, and address the importance of other predictors such as previous stone experience, prior interventions and renal comorbidities.