Value of Doppler ultrasound in early detection of diabetic kidney disease: A systematic review and meta-analysis

Abstract

The diagnosis of diabetic kidney disease can be delayed by limitations of primary biomarkers, which are microalbuminuria and estimated glomerular filtration rate. A number of Doppler ultrasound studies have associated an increase in intrarenal vascular resistance with the disease, which makes ultrasound a potential adjunct tool for early diagnosis. However, there is inadequate evidence to establish the effectiveness of including Doppler ultrasound in the diagnostic process. This systematic review was therefore conducted to determine the value of using Doppler ultrasound in early detection of diabetic kidney disease. Electronic literature searches were carried out in PubMed, CINAHL, Web of Science and EMBASE. All published prospective studies with records of intrarenal Doppler ultrasound, microalbuminuria and estimated glomerular filtration rate were obtained, and their relationship as parameters for diabetic kidney disease assessed. The meta-analysis of Doppler ultrasound versus albuminuria shows insignificant statistical difference between high resistive index of ≥ 0.7 and albuminuria, with the resistive index being the favoured parameter on the forest plot, making Doppler ultrasound highly comparable with albuminuria for the detection of diabetic kidney disease. Again, there was a significant statistical difference between high intrarenal resistive index of ≥ 0.7 and low estimated glomerular filtration rate of < 60 mL/min/1.73 m², with the resistive index being the favoured parameter on the forest plot, making Doppler ultrasound a superior parameter compared with estimated glomerular filtration rate for early detection of diabetic kidney disease.

Rationale

Diabetic kidney disease (DKD) is a microvascular complication of the kidney, affecting up to 40% of the diabetic population.¹ It is characterised by albuminuria and progressive loss of renal function, leading to end-stage renal disease (ESRD) and cardiovascular disease² with the latter being more common in type 2 diabetes.³ Diabetic patients with DKD have a three-fold increase in mortality rate, with 16 years loss of life expectancy⁴ although this can be significantly reduced by early detection and treatment.⁵

One major challenge of early detection of DKD is the method of diagnosis. This is based on microalbuminuria and an estimated glomerular filtration rate (eGFR) of < 60 mL/min per 1.73 m². However, some recent studies found limitations in microalbuminuria for early detection because of a spontaneous remission rate of about 21–64% in some diabetic patients.⁶ Also, significant decline in eGFR is found in some diabetic patients with normal albumin excretion.^7–9 On the other hand, eGFR calculation using serum creatinine level is time consuming and also influenced by other extraneous factors like muscle mass and diet (particularly meat intake), which could affect its accuracy.¹⁰ Even though eGFR estimation using cystatin C is considered more accurate, it is unavailable in most laboratories.⁶ Furthermore, the detection of low eGFR implies about 50% loss of functional renal tissue^11,12 and hence cannot be a reliable maker of early nephropathy. Due to these limitations, there is a search for other alternatives to complement microalbuminuria for earlier detection of DKD, before the significant reduction in eGFR.

Over the past two decades, a number of primary studies have demonstrated an increase in intrarenal vascular resistance with the Doppler ultrasound resistive index (RI) in diabetic patients. However, no systematic review and meta-analysis has been performed on the value of Doppler ultrasound RI as a potential marker for early DKD detection.

Objectives

The main objective of this review was to determine the value of Doppler ultrasound for early detection of DKD in type 2 diabetes, by assessing the relationship of intrarenal RI versus albuminuria and eGFR.

Protocol and registration

The general methods of the review and inclusion criteria were specified in advance and registered with PROSPERO as CRD42020190551

Methods

Preferred reporting items of systematic reviews and meta-analysis (PRISMA)¹³ was the protocol used in preparing this review.

Eligibility criteria in literature search

Every type of primary study was eligible for inclusion, whether observational or randomised control trial. Primary studies that provided data on intrarenal Doppler ultrasound, albuminuria and eGFR in their diagnoses of DKD were included. Papers without any of the information listed above were excluded. Literature search was restricted to only human subjects. There were no date restrictions in the search process.

Information sources

Literature search was done using electronic databases PubMed (Medline), CINAHL, Web of Science and EMBASE. In addition to this, the reference lists of relevant papers were searched to identify papers that could be missed by electronic databases.

Search

The search terms for this review were based on the target population, the intervention being investigated, the comparator for this investigation and the outcome (PICO). These were identified along with their synonyms and abbreviations as follows:

Population – Diabetic patients; type 2 diabetes, T2D

Intervention – Intrarenal arterial Doppler; Intrarenal artery resistive index – Renal resistive index, renal Doppler, renal RI, RRI, renal arterial Doppler

Comparator – estimated glomerular filtration rate; albuminuria; proteinuria; albumin excretion rate; microalbuminuria; eGFR; AER

Outcome – Diabetic nephropathy; diabetic kidney disease; DN; DKD; renal failure

The search terms were used in combination with Boolean operators ‘AND,’ ‘OR’ and ‘NOT’ together with truncators in the multiple electronic databases. Electronic database search was done in a total of seven steps (Table 1 shows details of results).

Table 1.

Results obtained from the search through electronic databases on 4 July 2020.

Search number	Combinations	Number of results
		PubMed	CINAHL	Web of Science	EMBASE
1	Renal duplex AND diabetic nephropathy	45	2	46	261
2	Diabetes AND renal resistive index AND nephropathy NOT transplant NOT chronic kidney disease NOT renal artery stenosis	201	5	44	0
3	Intrarenal artery resist* AND Diabet* AND nephropathy	64	6	38	41
4	Renal artery resistive index AND diabetes AND (albuminuria OR proteinuria OR ACr)	33	1	15	37
5	Intrarenal artery resistive index AND diabetic nephropathy	35	13	27	20
6	Intrarenal artery resistive index AND diabe* AND albuminuria AND GFR	3	1	1	1
7	Ultraso* Doppler AND diabetic nephropathy	150	25	119	277
Total 1511 results		531 results	53 results	290 results	637 results

Study selection

Articles obtained from the electronic search were exported to a citation manager. The articles were first sorted by removing duplicates. The title of the study and abstract were screened with reference to the objectives of the study. The primary selection was based on papers that clearly classified the diabetic patients according to their albuminuric status, eGFR and RI. Articles that did not meet these criteria were rejected. The selection process was performed by two independent members of the research team. Any disparity between the two team members was resolved by the third member.

Data collection process

The data collection process is illustrated by the PRISMA flow chart (Figure 1). This process was performed independently by two members of the research team and cross-checked by the third member.

Figure 1.

The adapted PRISMA flow diagram.

Data items

Information from the studies were extracted using the following headings; author and date, study location or origin, age, total diabetic population, high RI population, normal RI population, albumin status distribution and their mean RI, eGFR status distribution, duration of diabetes, non-diabetic control groups and body mass indices and study design.

Risk of bias

In this systematic review and meta-analysis, risk of bias was minimised by registering the protocols of the review before starting the study. The clearly defined inclusion and exclusion criteria also reduced the risk of bias. In addition, the Quality Assessment of Diagnostic Accuracy Studies – 2 (QUADAS – 2) tool¹⁴ was used to assess for the risk of bias in all the studies. However, the exclusion of grey literature and papers not written in English implies that potentially useful papers may have been missed.

Data synthesis

All relevant data from eligible studies were extracted and sorted into various categories and subcategories (Table 2).This included a synthesis of the sum of diabetic population with high RI (Figures 2 and 4), the sum of diabetic population with evidence of albuminuria (Figures 2 and 4) and the sum of diabetic population with low eGFR. These homogenous data were used in the forest plots presented in Figures 2 to 4. Data were then entered in to Revman 5.3 review manager, to construct the forest plots, using the Mantel–Haenszel statistical method.

Table 2.

Characteristics of included studies.

Author		Soyoye et al.34	Abdelhamid et al.17	Ozmen et al.18	Fallah et al.35	Thukral et al.16	Sperandeo et al.15	Taniwaki et al.36
Study location/origin		Nigeria	Egypt	Turkey	Iran	India	Italy	Japan
Total diabetic population		80	60	101	81	57	262	61
Population with high resistive index (i.e. RI ≥0.7)		N= 53	N = 40	N = 56	N= 54	N = 57	N = 117	N = 25
Population with normal resistive index (i.e. RI (i.e. < 0.7)		N= 27	N= 20	N= 45	N= 27	N= 0	N= 145	N= 36
Distribution of albuminuria status in diabetic population and their mean RI	Normoalbuminuria (N)	36 Mean RI = 0.7	20 Mean RI = 0.62	36 Mean RI = 0.69	27 Mean RI = 0.67	14 Mean RI = 0.74	145 Mean RI = 0.68	36 Mean RI = 0.695
	Microalbuminuria (N)	32 Mean RI= N/A	20 Mean RI = 0.72	37 Mean RI = 0.69	27 Mean RI = 0.71	14 Mean RI = 0.742	45 with normal serum creatinine and 72 with elevated serum creatinine Mean RI = 0.78 And Mean RI = 0.92	25 Mean RI= 0.703
	Macroalbuminuria (N)	12 Mean RI= N/A	20 Mean RI = 0.77	28 Mean RI = 0.72	27 Mean RI = 0.77	31 Mean RI = 0.825	N/A	N/A
eGFR status of the population	Low eGFR (N)	9	0	15	0	45	72	0
	Normal eGFR (N)	71	60	86	81	12	190	61
Duration of diabetes		0–30 years	–	8–20 years	10–12 years	2–7 years	7–20 years	7–13 years
Non-diabetic controls		N= N/A Mean RI:	N= N/A Mean RI:	N= N/A Mean RI:	N= N/A Mean RI:	N = 19 Mean RI:0.64	N= 100 Mean RI:0.60	N= N/A Mean RI:

Figure 2.

The forest plot of RI versus albuminuria.

Figure 4.

The forest plot of RI vs. GFR.

Figure 3.

The forest plot albuminuria vs. GFR.

Risk of bias across studies

In scrutinising for the risk of bias, it was assessed whether sonographers/physicians performing the intrarenal Doppler ultrasounds were blinded from the laboratory results on albuminuria and eGFR across studies.

Results

Study selection

A total of 1511 articles were identified through database searching as described, including PubMed, Web of Science and CINAHL and EMBASE. This reduced to 824 articles, after duplicates were manually removed in the citation manager (EndNote). The number of articles remaining for further screening by title and abstract was 33. The number of relevant articles for full text screening was 32, and eight articles were found to be eligible for inclusion in the systematic review, including one additional article obtained from the reference list of included articles (Figure 1). However, one was excluded due to inadequate information and, after contacting the author, raw data could not be obtained.

Study characteristics

Table 2 shows characteristics of the seven articles included in the review; three were from Asia, two were from Europe and two from Africa. A total of 702 diabetic patients were sampled in these primary studies. Out of this, 390 participants showed evidence of albuminuria (55.56%), including 272 with microalbuminuria (69.7%) and 118 (30.3%) with macroalbuminuria. Estimated GFR was low in 141 participants, which represents about 20% of the population. Intrarenal ultrasound Doppler reported a high RI in 402 participants, which represents 57.3% of the population.

Risk of bias within studies

The forest plot found fair homogeneity between RI and albuminuria (see Figure 2). The forest plot of Albuminuria vs. eGFR, and RI vs. eGFR found significant heterogeneity (Figures 3 and 4). However, funnel plot assessment of publication bias was considered less effective because there were less than 10 relevant papers included in the meta-analysis. Blinding of sonographers could be verified in four out of seven included papers. The rest did not indicate whether or not there was blinding.

Results on RI vs. albuminuria

The meta-analysis of RI versus albuminuria shows insignificant statistical difference between high RI of ≥ 0.7 and albuminuria, with RI being the favoured parameter on the forest plot for the detection of DKD (see Figure 2).

Results on albuminuria vs. eGFR

The meta-analysis of albuminuria versus eGFR shows a significant statistical difference between albuminuria and low eGFR of < 60 mL/min/1.73 m², with albuminuria being the favoured parameter on the forest plot for the detection of DKD (Figure 3).

Results on RI vs. eGFR

The meta-analysis of RI versus eGFR shows a significant statistical difference between high RI and low eGFR of < 60 mL/min/1.73 m², with the RI being the favoured parameter on the forest plot for the detection of eGFR (Figure 4).

Discussion

The main purpose of this study was to determine the usefulness of intrarenal ultrasound Doppler for early detection of DKD in type 2 diabetes. This objective was carried out by assessing the relationship between intrarenal Doppler RI and the primary biomarkers for DKD, which are albuminuria and eGFR. Using intrarenal RI of ≥ 0.7 as a Doppler ultrasound indicator for DKD, a meta-analysis of RI versus albuminuria and RI versus eGFR was conducted to determine the usefulness of RI in predicting DKD.

The results show that, in type 2 diabetic patients with DKD, there is an insignificant statistical difference between albuminuria and intrarenal RI of ≥ 0.7 as predictors of DKD (Figure 2). We found that 57% (402/702) of the population had high intrarenal RI of ≥0.7 (Figure 2), which was very close to the albuminuria population of 56% (390/702), with almost all albuminuria cases recording high RI. In most of the studies included in this review, diabetic patients with normo-albuminuria had RI < 0.7. This suggests that the rise in RI was largely associated with albuminuria. Again, < 0.7 RI was found in all non-diabetic patients who were presented as healthy control-group by some studies,¹⁵,¹⁶ which also suggests that normal kidneys will usually have an intrarenal RI of < 0.7.

This review also observed a consistent increase in RI with increasing levels of albuminuria in all included studies. Abdelhamid et al.¹⁷ showed a consistent rise in RI from 0.62 in normo-albuminuria, to 0.72 in microalbuminuria, and to as high as 0.77 in macroalbuminuria. This trend was also reported by Fallah et al., who obtained an RI value of 0.67 in normo-albuminuria, 0.71 in microalbuminuria and 0.77 in macroalbuminuria. Again, not only was a similar trend reported by Sperandeo et al., but they also found a marked RI increase of 0.91 in a section of their study population whose serum creatinine had increased. Although the RI of normo versus micro albuminuria did not show significant change in the studies of Thukral et al.¹⁶ and Ozmen et al.,¹⁸ they both reported a remarkable rise in the RI of all macroalbuminuria cases, which still supports that the RI increases with increasing albuminuria. This review has also shown that high RI in type 2 diabetes can rarely be associated with normo-albuminuria, but almost always associated with micro- and macroalbuminuria. This makes the RI a useful DKD indicator in type 2 diabetic patients who are at the verge of significant renal damage. It is believed that the increase in RI occurs because of possible pathophysiologic changes in type 2 diabetes. The exact mechanism of elevated RI in DKD is unknown. However, several authors in recent years have suggested possible reasons for RI elevation. Some believe that the non-specific renal scarring process which is characterised by interstitial fibrosis, loss of capillaries and glomeruli often lead to a reduced number of intrarenal vessels and filtration area, which can cause an increased parenchymal vascular resistance.¹⁹ Also, arteriosclerotic^20,21 and tubulointerstitial lesions^22,23 as well as diabetic glomerulopathy do occur in DKD which could also contribute to an increased intrarenal arterial resistance. Experimental and clinical studies provide compelling evidence that changes in RI are determined predominantly by systemic haemodynamic rather than isolated renal pathology.^24–26 However, a direct causal link with renal factors is demonstrated, particularly where vascular compliance and interstitial pressure are altered and as such is a useful indicator of early stage renal microvascular damage.^26–28 There is consensus that vascular change accompanies renal damage which this meta-analysis has confirmed from intrarenal ultrasound Doppler findings (Figure 2).

The meta-analysis of RI vs. eGFR has also revealed a statistically significant difference between the two biomarkers, with RI being the predominant biomarker for DKD (Figure 4). However, there was high heterogeneity in the forest plot because patients had different durations of diabetes (Table 2), which implies varying degrees of renal damage in cases of DKD. As a consequence, the forest plot (Figure 4) shows that only 20% (141/702) had low eGFR, compared to 57% that had high RI. The typical natural history of DKD begins with microalbuminuria as the first indicator, which later progresses to macroalbuminuria that in turn precedes eGFR decline.¹² In view of this natural history of DKD, and the increasing trend observed in RI, which is high in microalbuminuria and higher in macroalbuminuria, one can safely suggest that the increase in RI also precedes eGFR as indicator of DKD. It therefore implies that there were nearly 36–37% cases of DKD with evidence of albuminuria and high RI which had not yet shown eGFR decline. This implies that RI could serve as a better compliment of albuminuria for earlier detection before eGFR decline. In addition, the insignificant statistical difference in RI versus albuminuria, which showed only 1% difference in population (Figure 2), suggests that there were some 1% cases of DKD with high RI, which had not yet been detected by albuminuria.

In recent times, some studies have reported a non-albuminuric DKD phenotype, which is often confusing, and has sparked a lot of debate on the underlying pathophysiology.²⁹ In view of this, accurate diagnosis of DKD can be delayed. Our meta-analysis highlights the potential of Doppler ultrasound as an additional parameter for early detection of DKD. The current treatment option for DKD includes intensive blood glucose control, blood pressure control and renin–angiotensin system blockade, which are effective in preventing the development of microalbuminuria.² However, intensive glucose control put patients at risk of hypoglycaemia leading to increased mortality.³⁰ In view of this, the complimentary role of ultrasound Doppler RI could help in carefully selecting diabetic patients who are at higher risk of severe renal damage to undergo intensive glucose control, hence reducing the risk of hypoglycaemia in diabetic patients with lower RI, insignificant albuminuria and normal eGFR value.

The use of Doppler ultrasound is however associated with a number of limitations. It requires a standardised scanning technique by skilled personnel who have enough time to meticulously scan the interlobar, segmental or arcuate arteries for RI calculations.³¹ Additional skills for optimising spectral waveform are also necessary, such as adjusting the pulse repetition frequency, wall filter, and gain settings for a good image (Figure 5). Ultrasound is also limited by high attenuation in obese persons. Another limitation of intrarenal arterial RI is the fact that it may also increase in other health conditions, such as older age,³² hypertension³³ and the use of vasoactive agents. This makes it unreliable for diagnosing DKD, unless it is used in conjunction with microalbuminuria.

Figure 5.

Intrarenal Doppler ultrasound showing high RI in a DKD patient.

The strength of this study lies in the robust and strict methodology in ensuring that all literature published in English, which included eGFR, albuminuria and RI in their analysis were included and duly analysed. A major limitation to our study is, due to the strict inclusion criteria set, studies that did not evaluate all three parameters were excluded. This has a potential of excluding relevant studies that did not analyse all three parameters simultaneously. Also, papers published in other languages and in the grey literature were not included in the review, which could lead to loss of relevant data. Well-designed randomised controlled trials are however recommended to elucidate the exact clinical value of RI.

Conclusion

This systematic review and meta-analysis finds insignificant statistical difference between albuminuria and intrarenal RI of ≥0.7 for the detection of DKD in type 2 diabetic patients, with as many patients with albuminuria having high intrarenal RI of ≥0.7. In addition, a progressive increase in RI was observed in albuminuric DKD, with macroalbuminuria having a higher RI than microalbuminuria. It therefore suggests that an intrarenal RI of ≥ 0.7 can be used for earlier detection of DKD when used in conjunction with albuminuria.