Diagnostic Tests - Science topic

As far as I know, it is by comparing AUC of the ROC curve to determine which test is the best.

Funnel plots assess publication bias. They have nothing to do with assessing heterogeneity. Heterogeneity and publication bias are unrelated. Publication bias refers to when small studies with negative and/or nonsignificant findings tend to not be published and thus, are often omitted from systematic reviews/meta-analyses. There are several ways for assessing the presence of publication bias (e.g., Begg & Mazumdar rank correlation, Egger's regression intercept) but I consider the funnel plot to be the most useful means (at least after you've looked at a few dozen funnel plots).

One newspaper in the United Kingdom is reporting that some travelers are doctoring emails from Covid-19 testing laboratories, printing them out and simply handing them to airline staff before boarding flights.

Some Air Passengers Are Faking Negative Covid-19 Test Results, Per U.K. Reports (forbes.com)

One of the most important reasons to measure palpebral length is that an excessive eyelid length can lead to entropion of the lateral lower eyelid in some animals. In others it leads to a macropalpebral fissure and a combination of entropion-ectropion. There are also cases of microblepharon. All of this is important clinically (i.e. to know how best to operate an animal with one or another type of entropion - different entropions require different surgical approaches) but also for breeding (i.e. to not allow affected animals to breed, or to breed them with unaffected animals to reduce the effect of the problem). There is an article on palpebral eyelid length of dogs and cats from 1992 (now considered a little old, but unique, so still relevant). You might also find interesting the guidelines of the ECVO that mention macroblepharon and microblepharon (through www.ECVO.org, under hereditary eye diseases (HED)). I have included the paper and the guidelines for your interest (I have highlighted the important part of the guidelines). Enjoy.

Dear Dr. Ssematimba!

I found the following resource for YOU:

1) Mellacher, P. Endogenous viral mutations, evolutionary selection, and containment policy design. J Econ Interact Coord (2022). https://doi.org/10.1007/s11403-021-00344-3, Open access: y

Yours sincerely, Bulcsu Szekely

Thank you Mwoya Byaro

Yes Patrick. Completely agree it would be innaproppiate to put NPV and PPV because the high variability in prevalanece of the outcome. Tough telling the reviewer its not a good thing to do. Thanks!

Regarding your problem, I can recommend the following article:

Battese, G.E., Coelli, T.J. Frontier production functions, technical efficiency and panel data: With application to paddy farmers in India. J Prod Anal 3, 153–169 (1992). https://doi.org/10.1007/BF00158774.

In addition to the previous reply, here are the normal variants : The mean value for RNFL thickness in the general population is 92.9 +/- 9.4 microns. Typically, a normal, non glaucomatous eye has an RNFL thickness of 80 microns or greater. An eye with an average RNFL thickness of 70 to 79 is suspicious for glaucoma.

Hi Ravindra Muley thanks for your question. You can adopt a variety of tests. For instance, the Breusch and Pagan test to trace heteroskedasticity, Wooldridge test for autocorrelation, Breusch-Godfrey test for autocorrelation test. You can also use the cross-sectional dependence (CD) test

Hello Max Pierini. In trying to find the best balance between Se and Sp, you are using Youden's Index (or some variation on it). Let me remind you of what Jochen Wilhelm said in his reply (with emphasis added):

The performance of the new method can be evaluated by a ROC analysis. You may decide on a useful combination of sensitvity and specificity you can get (standard indices like the Youden index or similar are ignoring all practical consequences of false positives and false negatives and should not be used).

The only change I would make to what Jochen said is that Youden's index (or similar) should only be used when false positives and false negatives are deemed equally costly. But you want to limit false negatives (because of lethality), so trying to strike a balance between Se and Sp does not make sense. Rather, you need a cut-point that guarantees whatever level of Se you deem necessary (IMO). HTH.

Hi Pragyan

following the delta/delta Ct procedures, try this file

all the best

fred

For classic frequentist analysis (i.e, panel model & time series); depends on the hypothesis.

There is no rules of thumbs casted on stone, that a certain number of independent variables should be the minimum in exploring the links between dependent variables and independent variables. But having 22 observations are enough and have high degree of freedom to conduct the study. However, avoid including too many variables in the same model (i.e 7 variables) may cause multicollinearity.

You can also use Bayesian estimate, in which the small samples sizes can be simulated to larger sample by MCMC....

We still need more opinion on this matter

You are right that the prevalence of a condition in the tested population will have an important influence on the +PPV and –PPV. However, you can calculate another informative statistic – the clinical likelihood ratio (positive and negative). This is the effect of the test on the likelihood of the person having the condition. It simply compares the odds of the condition pre and post test. A positive test with a clinical likelihood ratio of 10, for example, means that the condition is ten times more likely to be present if the test is positive. And a negative likelihood ratio of 0·1 means that the test is only a tenth as likely to be present if the test is negative.

Likelihood ratios are based on odds, so they generalise across prevalences in a way that predictive values don't. This might help you deal with the prevalence effects.

When preparing a diagnostic test for misconceptions among students about cell division, it must be taken into account that the test paragraph measures the definition of the concept, the distinction of the concept from related concepts, and the application of the concept in a new situation to ensure the misunderstanding of the students

It is possible. I think the attached article may be useful for you.

Hi Diana Iskanderq ,

For the first question, it is mainly a matter of "True and false positive rates". Have a look to the paper in the link as an example, but you can search more specific papers.

For the second question, check the previous answer.

Good luck

I recently got indications of beta-hemolysis on blood agar made with sheep's blood in Alsever's solution with bacteria that should be gamma-hemolytic. The plates were fresh, as was the blood.

Unfortunately, I won't have fresh blood again for a few months, so I may not be able to verify these results for a while.

library(meta)

mymeta = metabin(TP, TP+FP, FN, FN+TN, sm = “OR”)

metabias(mymeta, method.bias = “deeks”)

funnel(mymeta, y.axis = “ess”)

@Pure tone threshold audiometry

Thank you for your valuable information.

I'm following the best answer.

Thank you Leonardo Mataruna for your help. I'm not really familiar with R, I'm using Rats, STATA, EVIEWS. I'll take a look at the reference you sent me.

Sensitivity and specificity are calculated on separate populations : those with and without the health state of interest. In each case, the appropriate sample size is for a single proportion. In Stata, this is power oneproportion. Your null proportion is the sensitivity/specificity of PCR, and your alternative proportion is the proportion that would represent a clinically significant difference. For example, calculation for 90% power to detect a test that's 92.5% sensitive or worse compared with PCR which is 97.5% sensitive.

. power oneproportion .975 .925, power(.9)

Performing iteration ...

Estimated sample size for a one-sample proportion test

Score z test

Ho: p = p0 versus Ha: p != p0

Study parameters:

alpha = 0.0500

power = 0.9000

delta = -0.0500

p0 = 0.9750

pa = 0.9250

Estimated sample size:

N = 166

Note that 80% power is not acceptable in real life settings. It guarantees a 20% chance of failure to detect! For real health research, 90% or 95% power is required – someone could die if you come to the wrong conclusion!

If you take a general population sample, a relatively small percentage will be true cases, so in order to get an adequate sample size for sensitivity you will have a sample that is far bigger than you need for specificity.

The solution is to sample for sensitivity separately, and to use identified cases to build up your sample quicker than you would with a general sample.

I would guess your standard error is very large? Perhaps because your sample is too small?

Dear Mihály Károlyi,

Perhaps,the thread started by Katrin Skerl "How can I compare 2 AUC values of different parameters of the same sample group (ROC analysis)?" on May 11, 2015 can help you.

Best regards

Dear Binyam Girma Sisay:

That is a usual problem we usually face when we intend to perform a meta-analysis of diagnostic tests. Though STARD initiative encourages for providing 2x2 table, many papers do not contain such simple and valuable information.

First option could be asking to the authors of the different papers to provide 2x2 table. Unfortunately, my experience is that many of them do not answer.

To perform the 2x2 Table, you would need: sample size, values of sensitivity and specificity and the prevalence of the disease. Prevalence can be calculated thorough predictive values.

The reference posted down provides a simple calculator to estimate 2x2 table. You only must fill the “Diagnostic Test” table and all the valuable information is automatically provided in “Observed Contingency Table”.

It also gives you the accuracy of the estimates (essentially, when it delivers whole numbers without decimals, the accuracy will be good. The obtained accuracy is provided in “Fisher Exact Test” table.

Best regards.

I appreciate everyone for their help! It was very enlightening and helped me reach a decision with my research group.

Quantitative

Attached file could be helpful :

Related to your query, I suggest you follow https://online.stat.psu.edu/stat509/node/152/

You need to construct a standard curve of at least five 10-fold serial dilutions of a known standard with predetermined copies/ml or IU/ml. After that it is a simple math to determine the concentration of your Ct from the curve. However, as per my knowledge the RT-qPCR COVID-19 test is qualitative in nature not quantitative. Otherwise, the kit manufacturers would have provided known standards for construction of a standard curve.

Following...

@Mohamed Abdel-Mabboud, thank you for the advice.

Hi Mohammed,

I hope you are doing fine.

I agree with Dr Frain, considering the fact that you are new to the estimation technique, you need to consult someone close to you to take you through the entire process including the condition under which you can even use the ardl else the whole exercise you are undertaking will be very hectic for you.

Besides, The coefficient of determination (R squared) comes with the estimation of the ardl model (conditional ols) before the short and long run results.

I will also suggest you get good econometric text books and read for more insight. Good luck

Thank you

Dear Umair,

"The Oxford Handbook of Panel Data," edited by Badi Baltagi, Oxford University Press, 2015 shows a number of tests you can do. Some of them listed in the index include constant-correlated effects, cross-section dependence, incidental parameters, moment conditions, noncointegration, overidentifing restrictions, residual based, sphericity, state dependence, unit roots etc.

The cheaper kit can be made on the basis antigen-antibody reactions. This technique requires a number of samples available to the developers.

The specificity and sensitivity do not seem to affect the regression analysis. Once the regression is done it cannot account for the specificity and sensitivity of the dependent variable.

Please find attached an interesting paper:

Please see the link below I found it very useful https://labtestsonline.org/news/laboratories-working-expand-covid-19-testing

I'm working on rehabilitative programs to promote adaptive behaviors in individuals with severe to profound developmental disabilities and multiple disabilities. I usually use the VABS to assess the severity of the disability and found it valid and reliable. Accordingly, I feel that the VABS constitutes a solid way to evaluate multiple disabilities. I wonder whether a german version is actually available in the literature.

Choice of optimal cut off depends on the objective of your study. Do you require more sensitive or specific test? If both sensitivity and specificity are equally important, you should use cut off with 0.66 youdens with good sensitivity as well as specificity. Anyway, in both the scenarios, AUC is good because it is >0.7. there are many other aspects are also involved while choosing cut off such as cost. Following article may help you.

1- If you run Traditional Panel, then the following diagnostic tests have to run:

Hausman Test, Normality Test, Multicollinearity Test, Hitroskedasticity Test, Serial correlation. Linearity Test.

2- If you run panel ARDL, then in addition to the above tests, you have to run Unit root test.

All the best.

How would you like to define 'most cost-effective' for this? Do you have a willingness to pay in mind?

I would like to warn that if you are working with a group of women with a positive screening mammogram, then incidence and prevalence numbers for the general population are not relevant.

Are you data on sensitivity and specificity based on a population of women with positive screening mammogram? Might the different tests even be performed at the same time on the same women?

At this stage, it may be best to read up on cost-effectiveness in general.

There are books on the subject, e.g.:

Methods for the Economic Evaluation of Health Care Programmes (Oxford Medical Publications)

by Michael F. Drummond , Mark J. Sculpher , et al.

Cost-Effectiveness in Health and Medicine 2nd Edition

by Peter J. Neumann, Gillian D. Sanders, et el.

Decision Modelling for Health Economic Evaluation (Handbooks in Health Economic Evaluation) 1st Edition

by Andrew Briggs, Karl Claxton, Mark Sculpher

Decision Making in Health and Medicine: Integrating Evidence and Values

by M. G. Myriam Hunink , Milton C. Weinstein, et al.

ISPOR focuses on drugs, which works a bit differently than evaluation of tests, but they do make some good background readily available:

https://www.ispor.org/heor-resources/good-practices-for-outcomes-research/report/-in-category/categories/economic-evaluation

While old, I think that I should still recommend chapter 10 from my dissertation concerning specific aspects of breast cancer screening evaluation:

https://repub.eur.nl/pub/1377/

I also uploaded it to the ResearchGate web site:

Please, always try to confirm the diagnosis by microscopy, RDT or PCR. Never give Artesunate alone, always in a combination therapy, due to the emergence of possible resistance to artemisinins.

Hi Abhishek Aravind

As the goal is to establish inter-rater agreement between two instruments, I suggest the use of Weighted Kappa statistic. For detailed justification see:

For sample size calculations and pre-calculated sample size tables, refer to

Dear Aileen,

You may also want to check the NICE guidance for the Diagnostic Assessment Program.

Please check the Methods section of the programme manual (link to pdf at the bottom of the page).

Practical applications are published in HTA journal:

Good luck!

Isaac

Dear Mohamad-Hani, maybe the following papers will help you on the subject:

Kunkel RS. Migraine aura without headache: benign, but a diagnosis of exclusion. Cleve Clin J Med 2005;72(6):529-34. https://mdedge-files-live.s3.us-east-2.amazonaws.com/files/s3fs-public/issues/articles/content_72_529.pdf

Tinsley A, Rothrock JF. What Are We Missing in the Diagnostic Criteria for Migraine? Curr Pain Headache Rep 2018;22(12):84. https://link.springer.com/article/10.1007%2Fs11916-018-0733-1

Yang H, Zhang J, Liu Q, Wang Y. Multimodal MRI-based classification of migraine: using deep learning convolutional neural network. Biomed Eng Online 2018;17(1):138. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6186044/pdf/12938_2018_Article_587.pdf

van der Meer HA, Visscher CM, Vredeveld T, Nijhuis van der Sanden MW, Hh Engelbert R, Speksnijder CM. The diagnostic accuracy of headache measurement instruments: A systematic review and meta-analysis focusing on headaches associated with musculoskeletal symptoms. Cephalalgia. 2019 Apr 18:333102419840777. doi: 10.1177/0333102419840777. [Epub ahead of print]. https://journals.sagepub.com/doi/pdf/10.1177/0333102419840777

Šukalo A, Merdžanović E, Alic A, Vrabac-Mujčinagić M, Alibašić E, Janković SM. Screening general practice patients for migraine without aura: construction and validation of the Balkan Migraine Screening Questionnaire (BMSQ). Med Glas (Zenica). 2019 Aug 1;16(2). doi: 10.17392/1021-19. [Epub ahead of print]. https://www.ncbi.nlm.nih.gov/pubmed/31127712

I think it depends on the reason for doing the analysis. Some organisms will continue to grow even at low temperatures so the relative abundance of all organisms will change with storage. If it is just to show that an organism is present then pcr only needs short fragments of dna so degradation will not do any harm ( small fragment dna may even amplify better than large dna as they melt easier in the early cycles). The main issue is whether large amounts of other organisms grew when they may produce enzymes that digest the dna of the organism that you are looking for since the skin sample will not have been sterile

Joshua;

When using sugar bait traps for general ant collecting I use 30-40 g/L solution. The threshold for diagnosing diabetes is about 0.126 g/L. It would be a simple lab experiment to determine the detection threshold of some of the nectar-gathering species of ants. In the southwestern US where I live there are an abundance of species that do that. Members of the genera Prenolepis, Myrmecocystus, Formica and Lasius are common and are readily maintained in the lab...and are commonly found in sugar bait traps.

See if you can find some data on that question. It's only a matter of personal curiosity because Nemanja is correct in his assessment.

Best regards, Jim Des Lauriers

Thank you for your reply Dilesha,

actually I came to the conclusion that the non-clustered robust estimator is known to be invalid for panel data. I have used the clustered standard error as it is, itself, also robust to heteroskedasticity.

Apparently, if you, incorrectly, give Stata the command -xtreg DV Ivs, fe vce(robust)-, Stata will actually disobey this command and substitute vce(cluster country_num) without even adverting to it. As such, results are identical for both commands.

Dear Sara

I hereby send you 3 articles dealing with the how to use the GRADE approach in diagnostic studies - and I agree with Gopalakrishna et al that it is a challange bu doable

Gopalakrishna G, Mustafa RA, Davenport C, Scholten RJ, Hyde C, Brozek J, Schünemann HJ, Bossuyt PM, Leeflang MM, Langendam MW. Applying Grading of Recommendations Assessment, Development and Evaluation (GRADE) to diagnostic tests was challenging but doable. J Clin Epidemiol. 2014 Jul;67(7):760-8.

Singh S, Chang SM, Matchar DB, Bass EB. Chapter 7: grading a body of evidence on diagnostic tests. J Gen Intern Med. 2012;27 Suppl 1(Suppl 1):S47-55.

Schünemann HJ, Oxman AD, Brozek J, et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies. BMJ. 2008;336(7653):1106-10.

Best Carsten

The Clinical picture seems to be urticaria. May be contact urticaria ( if localized only on the distal arms) or other causes if generalized. In any case firstly treat it by anti-histamine drugs and after almost two weecks of suspension consult an allergist and make the appropriate investigation on ethiology of hives.

Dear Adrienne,

that could actually happen. This can occur due to a choice of cut-offs for the two diagnostic tests.

Sensitivity and specificity depend upon a cut-off set by the diagnostician. The ROC (and thus the AUC) does not depend on your choice of a cut-off, because it is a depiction of the tests accuracy independent of a cut-off.

I find the following website very instructive in order to illustrate what I just wrote. It becomes clear, if one plays a little with the settings there.

Best,

Daniel

As the biology of a patient has some influence on each of your results, those estimates are in a way dependent on each other. Let's say a subject has low health - this will cause his antibody results be lower in both results compared to someone who is healthy. I assume that the test you used depends on random samples (at least) because of that matter.

My suggestion is to split your subjects sample randomly and use one part for diagnostic test 1, the second part of the sample for test 2. Then compare those results as you did. (be sure to have a large enough sample and to split the sample randomly or else the effect will be biased)

I hope this helps you a little.

Why not use culture or PCR as a gold standard for calculations?

Sensitivity and specificity, positive and negative predictive values, and positive and negative likelihood ratios are common indicators of diagnostic test accuracy. However, as illustrated by Glas et al., these paired indicators are not valid for comparing multiple diagnostic tests especially if one test was not superior to the other in both indicators. In addition, these parameters cannot be analysed in the traditional framework of the network meta-analysis or indirect comparison models. Alternatively, you can use the diagnostic odds ratio (DOR) as the effect estimate for your Meta-analysis. DOR is the single global indicator for diagnostic test performance and is defined as the ratio of the odds of positive test results in subjects with the disease compared to the odds in those without the disease. This effect estimate can be calculated according to the following equations.

📷

📷

Hi Niladri, thanks for your suggestion. I think I felt lazy to do the coding... Based on your suggestion, I will try to do it. In case I do not succeed in Stata, I will just look to R or WinBUGS! Thank you.

Hi Ajit, yes that is the traditional method for determining a test's sensitivity and/or specificity (also known as true positive rate/recall and true negative rate, depending on the discipline).

I do plan on calculating that and I was looking for a more rigorous statistical test to assess the new test's performance.

confidence interval of 99% represents close to 3 standard deviations, while 95% is 2. An 99% CI with 5% error is about 3.3% at 95% CI, which is within rounding error (all figures provided are severely rounded) of the first test.

I am assuming that the figures quoted are for the test as used (including the fixed number of sub-samples or replicates used per sample).

If CIs are based on different numbers of replicates then it may be useful to work backwards to find the precision for comparing systems - one system may have less precise measurements but compensate by taking more measurements to achieve the same level of confidence in the result.

If as well as confidence intervals the sensitivity and specificity are similar then the choice is likely to come down to practical considerations and require a benefit/risk analysis. Is there a different in capital cost, running cost, throughput, scalability etc.

Decision making regarding therpeutic strategies in cancer patients is complicated. How you can identify the mutations, how much you rely on the results, heterogeneity of tumoral cells, cost benefit of the treatment, insurance coverage, patient cooperation and many other factors.

Following

Thank you VERY MUCH Zoltan.

Following

NGS sequencing of a clinical exome gene panel (TS1) and analysis limited to 8 genes associated with PKD (BICC1, LRP5, NOTCH2, PKD1, PKD2, PKHD1, PRKCSH, SEC63). If the potential candidate variant is detected in PKD1 gene (it has 6 highly homologous pseudogenes) then it is confirmed by long-range PCR of selected region, followed by exon-specific Sanger sequencing. If the variant is detected in any other gene it is confirmed by classic PCR followed by Sanger seq.

Hi,

Currently, International Working Group (IWG) and National Institute on Aging Alzheimer's Association (NIA-AA) has proposed several biomarkers as diagnostic criteria which includes cerebro spinal fluid (CSF) amyloid beta (Aβ) and tau, atrophy on MRI, glucose metabolism on [18F]-fluorodeoxyglucose positron emission tomography (FDG-PET) and fibrillar Aβ burden on amyloid PET.

There are few regions, which acts as definite Hallmark regions for AD. You could refer to below attached papers.

Structural MRI and Amyloid PET Imaging for Prediction of Conversion to Alzheimer's Disease in Patients with Mild Cognitive Impairment: A Meta-Analysis

Biomarker-based prediction of progression in MCI: Comparison of AD signature and hippocampal volume with spinal fluid amyloid-β and tau

I hope these article will help you.

Further, regarding the medication you could look into these articles

Drugs for Alzheimer’s Disease: Are They Effective?

Gud Luck !!!!

Since spironolactone is contraindicated during pregnancy, a case report on primary hyperaldosteronism during pregnancy reported a favourable outcome for both mother and baby after switching to amiloride.

good question I follow

In general siloxanes deriving from the column might cause electrical leaks across insulators. Therefore the conditioning of the column (which should not be connected to the MS !) is a crucial point.

Troubleshooting the internal ion source is more difficult. I have no further advices for that. The remark of Christopher Lee regarding a kind of jet separator is worth thinking about.

Coronary angiography is the gold standard test.

Antigen detection ELISA is rapid. we are using