Role of Hounsfield Unit in Predicting Outcomes of Shock Wave Lithotripsy for Renal Calculi: Outcomes of a Systematic Review

The search strategy was designed following a three-step approach recommended by Cochrane group [4].

An initial scoping review was performed in MEDLINE to identify similar studies to generate a comprehensive list of MeSH terms and keywords relating to the domains: (A) shock wave lithotripsy or lithotripsy or SWL or ESWL, (B) renal stone or renal calculi or renal calculous, (C) nephrolithiasis, (D) computed tomography or CT, and (E) Hounsfield unit or HU. Searches relating to the domains were combined with Boolean operators: AND, OR, and NOT. Database search was performed across the Ovid EMBASE, Ovid MEDLINE, and Scopus from inception till August 2022. Search results were limited to those published in peer-reviewed journal, available in English language, and involving adult human subjects (aged 18 years and above).

All relevant articles were exported into EndNote [5], wherein digital deduplication was performed, followed by manual deduplication. The final set of articles was exported to Rayyan [6] for title and abstract screening, pursued by full text screening.

Title and abstracts were double screened by two independent reviewers (MG/HJ) to ensure compliance with the study eligibility criteria. Full-text publications were assessed using standards for reporting of diagnostic accuracy (STARD) [7], to ensure all included studies were accurate and relevant to the purpose of this review. Eligibility was ensured if the studies fulfilled the following inclusion criteria: (A) addressed stone density or attenuation in cohorts of adult patients undergoing ESWL for renal stones, and previously diagnosed using non-contrast CT imaging; (B) empirical study design such as randomised control trials, non-randomised control trials, cross-sectional, cohort, or case control, excluding case studies, systematic reviews, and meta-analysis; (C) published since 1976 in a peer-reviewed journal; and (D) available in the English language.

Double data extraction was manually performed by two independent reviewers (MG/HJ) using a self-designed Microsoft Excel spreadsheet (data extraction tool), which was piloted before use in the study. The following data were recorded: general study characteristics (name of the author, year of publication, and study design), participant characteristics (total number of participants, gender and age distribution), treatment details (including initial imaging, treatment protocols, and outcome evaluation), and stone parameters (stone size, stone density, and HU).

Disagreements at screening or data extraction stage were resolved by discussion amongst the reviewers. If no consensus was reached, the final decision was taken by a third independent reviewer who acted as an arbiter (JP). Additionally, in case of the absence of full-text or requirement of further outcome data, two attempts were made to contact the study authors via email. If the authors did not respond or data remained unavailable within two weeks of sending the second email, the respective study was excluded at that stage.

The data was reviewed qualitatively and quantitatively after evaluating the direction size, homogeneity of effects amongst studies, and the strength of evidence. As different research outcomes could not be combined, results (including characteristics of individual studies) were displayed in a tabular or descriptive form (narrative synthesis). Moreover, heterogeneous participant categorisation existed across most studies, which rendered participant standardisation impossible, thus a weighted pooled estimate (meta-analysis) was not performed.

Measurement of stone attenuation varied between studies. All studies utilised the conventional HU scale. The CT window used to analyse stones varied between studies. In publications specifying CT window, the bone window was used in 14 studies [10, 11, 13, 15, 16, 22, 27, 28, 31, 34•, 35], and the abdominal (soft tissue) window was used in five [14, 20, 21, 26, 34•].

23 studies specified the method of measuring mean stone density (MSD) [11‐16, 18‐23, 25, 26, 29‐31, 32, 33, 34•, 35, 36]. The majority (n = 23) measured stones on the longest stone diameter (longitudinal or transverse planes), with multiple regions of interest (ROI), incorporating the stone but not the surrounding soft tissue. This was done creating single elliptical ROI, a squared 10 × 10-pixel map, or a freehand ROI drawn along the stone edge to take into account abnormal shapes. These studies used 3 ROI within the image, either over-lapping or non-overlapping, and taking the mean of the three results or the peak HU [9‐15, 20‐22, 24, 26, 29‐31, 34•, 35, 36]. Seven studies measured the stones on a single axial plane that displayed the stone at its maximal diameter [13, 18, 19, 27, 30, 34•, 35]. Whilst others took the average HU from ROIs in three separate axial planes: the upper pole, the stone at its maximal diameter, and the lower pole [11, 12]. A single study defined MSD as the average of the minimum and maximum HU readings [13]. Joseph et al. [14] and Pareek et al. [15] generated a pixel map of the largest stone dimension, measuring the maximal attenuation within ROI. Mean value was calculated using differing norms. Perks et al. [16] compared two methods of stone attenuation measurement, using a single elliptical ROI vs. 3 small, 0.005 cm² non-overlapping ROIs within a single axial slice. Despite the latter method being more time consuming, the two methods correlated highly (r² = 0.98, p < 0.001), and the authors proceeded to use the easier, single ROI.

Finally, newer 3-dimensional methods of measuring stone attenuation have been developed. Yamashita et al. [37•] measured MSD for 3-D stone images, comparing this to the two methods used by Perks et al. [16]: the single elliptical and 3 non-overlapping ROIs. For prediction of SWL outcome, the area under the curve (AUC) for the receiver operator characteristics (ROC) curve was 0.6330 for 3-D images, significantly higher than elliptical ROI in abdominal (0.5836) or bone (0.5797) windows, and average of 3 ROIs in abdominal (0.5756) or bone (0.5794) windows. Langenauer et al. similarly found improved SWL outcome predictive value for 3-D MSD vs. traditional MSD taken from a single CT slice (AUC 0.70 vs. 0.66) [38].

The assessment of SWL outcome differed from study to study, and therefore, makes comparison challenging. Most of the studies defined SWL ‘success’ as patients who were either stone-free or had ‘clinically insignificant fragments’. This outcome differed between studies, with fragments < 3 mm, < 4 mm, and < 5 mm termed ‘clinically insignificant’ in 9, 13, and 2 studies, respectively. Two studies defined SWL success as complete clearance of stone fragmentation with no further intervention required [17•, 18].

Furthermore, there were differences in SWL protocols, including both shocks delivered and number of SWL sessions. Between 2000 and 4000 shocks were delivered per SWL session. Most units protocolled a low voltage start of SWL treatment with escalation to a maximum of 5–6 kV.

Twelve studies assessed SWL success following a single session, whereas 11 allowed for 2–3 sessions. Three studies allowed more than 3 sessions per patient, whereas two publications were unclear as to the number of sessions allowed.

Statistical analyses differed from study to study. All studies performed analyses, which included comparisons of MSD in success or failure groups, comparisons of SWL outcomes between patient groups, dichotomised by a MSD cut-off value, and/or multivariate analyses to identify significant predictors of SWL outcomes.

Focusing on studies that analysed MSD and included multivariate analysis, 26 studies were identified. 24 studies examined MSD as a continuous variable; 3 studies found no association between MSD and SWL outcome on univariate analysis [18, 19, 27] and were excluded from further multivariate analysis. On multivariate analysis, 17 studies showed increasing MSD to be a significant predictor of SWL failure.

Nineteen studies determined an appropriate MSD cut-off and dichotomised the patient cohort into two groups. Of these, 16 found higher MSD to be a significant predictor of SWL failure. Finally, two publications examined MSD as both a continuous and dichotomised variable. MSD was not significant on univariate and multivariate analysis [19, 20].

Multiple studies have varied in their assessment of MSD. Associated parameters such as peak HU density and heterogeneity indices (SHI) have also been used to try to improve sensitivity for SWL success. Whilst the majority of studies examined only MSD, a smaller proportion, five studies [9, 10, 21, 28, 34•] evaluated additional HU factors. These studies found peak HU and SD to be significantly lower in patients with successful SWL, with a cut-off peak stone attenuation (PSD) of 900HU and mean SD of < 750HU as cut-off values for SWL success.

Researchers have examined the value of CT parameters in assessing internal structural heterogeneity of stones and its impact on SWL outcomes. The cut-offs are shown in Table 2 with values ranging from as low as 482 HU up to 1000 HU. Multiple way points have been proposed to delineate SWL success probability with 15 studies included in this systematic review having cut-off values of 750–900 and 1000 HU using ROC curves [11‐14, 16, 19, 20, 23, 25, 26, 28, 29, 31, 33, 35].

Several studies have found an association between stone attenuation and size (Table 3) [9, 34•, 35, 36]. To correct for this correlation, HU density was proposed as an alternative to MSD or PSD [9, 35, 36]. Three specific methods of measuring HU density were seen in this review. The most common method used in 3 studies is mean HU divided by maximum stone diameter (HU/mm). The stones in the successful treatment group had a lower HU density of almost 15 points, but not reaching significance. The authors suggest homogeneity of stones may explain this [36].

Stone heterogeneity index (SHI) was identified to be a better indicator of success in patients with larger stones [22, 34•]. SHI designated as the standard deviation of HU was found to be a truer reflection of SWL success in patients with larger stones [10‐20 mm], with higher heterogeneity likely to result in clearance. This new ratio was seen to have a higher negative predictive value than stone attenuation alone in predicting stone composition. ROC curve study suggested a cut-off value of 213 (AUC 0.60-CI 0.531–0.673) [20]. SHI may suggest stone intrinsic diversity of composition implying increased fragility.

Four studies developed predictive cut-off values using receiver operating characteristic (ROC) curve analyses utilising MSD or PSD to identify optimal cut-off value for independent predictors. Kaya et al. [27] reported on 51 patients evaluating serum creatinine, stone size, stone attitude, and skin to stone distance (SSD). The ROC curves revealed serum creatinine level (AUC: 0.681), stone size (AUC: 0.767), stone attitude (AUC: 0.672), HU (AUC: 0.722), and SSD (AUC: 0.672) as significant predictive factors for SWL outcome. The values were 0.86 mg/dl, 10 mm, 0.65, 618HU, and 9.2 cms, respectively. Based on ROC curve analysis, cut-off values for ESWL success were considered to range between 500 and 900HU [9, 16, 27, 35, 36].

There are large numbers of single-centre retrospective case series and cohort studies within the SWL literature. This constitutes poor evidence on which to base recommendations. First, a standardised protocol for measurement of radiological stone parameters and patient outcome reporting should be established, to allow for accurate comparison between studies. Second, trends towards prospective, multi-centre collaborative studies should be encouraged to accurately assess true predictive factors for SWL, including type of lithotripter used, modality of imaging, and criteria to claim stone clearance or adequate fragmentation. In addition, the development of machine learning evaluation of multiple variables should help develop better prediction models for ESWL success.

The strength of this review is the systematic approach used to review the literature on HU and SWL outcomes. However, an obvious weakness is the dependence on primary studies, which included many retrospective, non-randomised studies. These studies were potentially prone to bias in both patient selection and outcome reporting. Furthermore, there was significant heterogeneity in the method of stone attenuation measurement, outcome measurement, and patient follow-up with similar outcomes in both retro- and prospective studies.