Use of a Smartphone Camera at the Bedside to Assess Adequacy of Kidney Biopsies

Kidney biopsies are the gold standard for diagnosis in evaluating native and transplanted kidney disorders. Obtaining adequate tissue samples is vital. Although the number of glomeruli required to make a diagnosis is variable, obtaining 15–20 glomeruli is widely considered adequate. Traditionally, real-time bedside adequacy of the biopsy sample has been evaluated under a dissecting or standard light microscope.¹

Due to the associated cost and manpower shortage, several institutions have abandoned tissue assessment at the bedside. In one such example, the rate of inadequate sampling went up from 12.5% to 21.6% after abandoning bedside microscopic evaluation.² Similarly, another study showed inadequate sampling of 5.7% with onsite microscopic evaluation and 22% without it.³ This relationship between onsite tissue evaluation and sample adequacy appears to be marked with the standard radiologist practice of using 18 G needles to obtain ultrasound-guided kidney biopsies.⁴

With improvements in technology, the use of smartphone-based evaluation is increasing in pathology. A study from Thailand showed the use of a smartphone-based magnifying device led to a transplant biopsy inadequacy rate of 7%, as compared with 21.3% without it.⁵ However, they pretested a 40× magnification device and evaluated the presence of only a single glomerulus in the specimen. We hypothesized that experienced interventional radiologists can visually determine specimen adequacy with the assistance of a smartphone camera (Figure 1).

Figure 1.: Appearance of biopsy tissue in smartphone photographs. (A) Predominant renal cortex. (B) Predominant renal medulla.

We performed a prospective cohort study on consecutive adult patients who underwent kidney biopsies in the department of interventional radiology at Geisinger Medical Center between August 1, 2019 and December 11, 2019. We compared the adequacy assessment made using a smartphone camera and tissue measurement to the assessment made by trained pathology technologists using light microscopes. Detailed methods are available in the Supplemental Methods. Tissue was considered adequate at bedside if the smartphone photograph showed speckled glomeruli with a core length approaching 1 cm. The standard of care microscopic evaluation determined the procedure completion. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC). Our research protocol was approved by the Geisinger Institutional Review Board.

As summarized in Table 1, 57 kidney biopsy core specimens were obtained for 20 consecutive patients. Of these, ten (50%) were native kidney biopsies (27 cores), and ten (50%) allograft biopsies (30 cores). In total, 80% of the biopsies were obtained using real-time ultrasound guidance and 20% using computed tomography guidance. The number of cores obtained ranged between two and six, with a mean of 2.9 per patient. Mean core specimen length was 1.5 cm. The mean age was 48 years. The most common biopsy indication was AKI (60%), followed by proteinuric CKD (30%), and nonproteinuric CKD (10%). No hemorrhagic complications occurred and recovery for all patients was uneventful.

Table 1. - Results of transplant and native kidney biopsies

Variable		Value
Biopsy details
Number of biopsies		20
Biopsy type, n (%)
Native kidney biopsy		10 (50)
Transplant biopsy		10 (50)
Imaging guidance, n (%)
Ultrasound		16 (80)
CT		4 (20)
Use of sedation, n (%)		18 (90)
Biopsy needle, n (%)
18 G		19 (95)
16 G		1 (5)
Total number of cores, n		57
Number of cores per patient, n (SD)		2.9 (1.1)
Mean length of core, cm (SD)		1.5 (0.4)
Clinical details
Mean age at biopsy, yrs (SD)		48 (20)
Indication
AKI, n (%)		12 (60)
Proteinuric CKD, n (%)		6 (30)
Non-proteinuric CKD, n (%)		2 (10)
eGFR before biopsy, mean (SD)		39 (23)
Proteinuria estimate (mg), mean (SD)		2005 (2302)
Bedside adequacy versus microscopic adequacy
Bedside, n (%)	Microscopic, n (%)
Adequate	Adequate	38 (67)
Inadequate	Inadequate	15 (26)
Adequate	Inadequate	3 (5)
Inadequate	Adequate	1 (2)
Positive agreement %		97.4
Negative agreement %		83.3
Overall agreement %		93
Cohen's kappa statistic		0.83
Final pathology results
Cortex as % of total sample, % (SD)		67 (32)
Glomeruli obtained per patient, n (SD)		26 (12)
Able to make diagnosis, n (%)		20 (100)
Diagnosis, n (%)
Glomerular disease		11(55)
Tubulointerstitial disease		4 (20)
Vascular disease		2 (10)
Acute rejection		3 (15)

CT, computed tomography.

Out of 57 specimen cores, 38 (67%) were considered adequate by both the radiologist and pathology technician, and 15 (26%) were considered inadequate by both. The positive adequacy agreement rate was 97.4% and negative adequacy agreement rate was 83.3%. Cohen’s kappa statistic to measure the level of agreement between smartphone-assisted bedside evaluation and microscopic evaluation was 0.83 (95% confidence interval, 0.63 to 1.0). The sensitivity analysis including only ultrasound-guided biopsies showed similar results: positive adequacy agreement rate was 97.1%, negative adequacy agreement rate was 93.3%, and Cohen’s kappa statistic was 0.90. On the final pathology reports, 67% of the total specimens comprised of kidney cortex and the rest were predominantly medulla. The mean glomerular yield was 26. There was adequate tissue for light microscopy, electron microscopy, and immunofluorescence in every patient. The most common pathologic diagnosis was glomerular disease (55%), followed by tubulointerstitial disease (20%), acute rejection (15%), and vascular disease (10%).

The three core specimens rated adequate by radiologists and inadequate by microscopic evaluation were obtained on two native kidney biopsies. They were predominantly medullary samples (up to 90%) on review of the final pathology. In both patients, a clear diagnosis (minimal change disease and IgA nephropathy respectively) was possible. The only specimen rated inadequate by radiologists and adequate by microscopic evaluation was also a native kidney biopsy specimen, and a medullary predominant (60%) sample. Again, it led to a definitive pathologic diagnosis (scleroderma microangiopathy).

Our approach is simple, low cost, and easily reproducible. We demonstrated excellent agreement between smartphone-assisted and microscopic evaluation of kidney biopsy adequacy. Although there are clear differences in technique and operators, our method appears to be at par, or superior to, previously tested specialized devices.⁵ Smartphones, owned by 85% of Americans, are ubiquitous, cost effective, and user friendly.⁶ The majority of kidney biopsies now are performed by radiologists, who commonly lack access to onsite adequacy evaluation.⁷ In this setting, applying our approach may lead to a decreased need for repeat biopsies, while maintaining a good diagnostic yield and reducing the overall cost of care.

Given the small sample size, our approach certainly requires robust testing and head-to-head comparison before generalization. Development and validation of quantitative thresholds for smartphone-based adequacy assessment in larger studies is a logical next step. Quantitative thresholds would also allow independent observer evaluation for perception bias: is a radiologist more likely to consider a specimen adequate in patients at high risk of bleeding or requiring numerous needle passes? However, our study is hypothesis generating and presents a possible alternative to bedside microscopy in resource-constrained health care systems. This approach should be tested in larger multicenter studies for validation.

Disclosures

A. Chang reports having consultancy agreements with Novartis; reports receiving research funding from Novo Nordisk as an investigator in a sponsored study; reports receiving honoraria from Reata; reports being a scientific advisor or member of Reata and Relypsa; and reports having other interests/relationships with National Kidney Foundation with grant support from the National Kidney Foundation Patient Network. P. Anand reports having consultancy agreements with CareDx, Natera, and Veloxis; reports receiving research funding from CareDx and Natera; reports receiving honoraria from Caredx, Natera, and Veloxis; reports being a scientific advisor or member of Natera; and reports speakers bureau with Caredx, Natera, and Veloxis. S. Sharma is employed by Arkana Laboratories. All remaining authors have nothing to disclose.

Funding

None.