Accuracy of Smartphone Apps To Measure Pulse Oximetry

Some smartphones have a pulse oximeter function. Samsung, for example has an app called Digidoc, which uses the camera and the flash inbuilt into the phone to give a an arterial oxygen saturation. To understand how this works, it is perhaps important to understand how standard pulse oximetry functions, so that we may compare the two approaches.

The standard pulse oximeter is based on the principle that oxygenated haemoglobin (oxyHb) and deoxygenated haemoglobin (deoxyHb) absorb various wavelengths of light differently. Thus, as the graph below illustrates, oxyHb absorbs more infrared light than deoxyHb, while deoxyHb absorbs red light far better than oxyHb. (Pnemonic: SeXy DARLing- At SiX hundred wavelength, Deoxygenated haemoglobin Absorbs Red Light)

The pulse oximeter has a diode for emitting red light (wavelength 660 nm) and one that emits infrared light (wavelength 940 nm) separately. These are passed in turn through the finger inserted into the probe, and a screen below the finger (with its contained artery) measures the amount of light coming through. It also measures the amount of ambient light passing through the finger and subtracts it from the total light traversing the finger.

The oximeter cleverly ignores the "noise" from absorbance contributed by tissues such as skin, muscle etc (because these too absorb light) by only measuring light absorbance when it is pulsatile, i.e. generated by arterial pulsation. It thus ignores the blood flow in veins, which is of course non-pulsatile.

It is thus important to look at the graph (called a plethysmograph) of oxygen saturation generated by the oximeter or app. it should look something like this. If the graph looks like the tracing on the top, you are OK. If it looks like the one at the bottom, don't trust it. The plethysmograph is as important as your oxygen saturation reading. If the graph looks unfamiliar, the reading is not accurate. (Pnemonic used by anaesthetists: SpO2- See pleth before O2)

Thus it is intuitive that depending on the amount of oxygen present in arterial blood, the relative amount of oxyHb versus deoxyHb will vary. With this, will vary the ratio of cumulative red light absorbed versus the cumulative infrared light absorbed. Thus, at 100% arterial oxygen saturation, the only contribution will be from oxyHb, and the red:infrared absorbance ratio will be that of oxyHb. Conversely, at 0% oxygen saturation, the red:infrared absorbance ratio will be that for deoxyHb. At various intermediate saturations, the ratio will be between the 2 extremes, and the computer can easily tell what proportion of Hb is oxygenated from a given figure. This sort of analysis is called "ratio of ratios".

The pulse oximeter readings were standardised by measuring oxygen saturations in healthy volunteers given fixed amounts of oxygen to breath in a previously titrated oxygen/air mixture. However, oxygen saturations below 75% were not used due to reasons of safety and below this level, mathematical methods are used to calculate oxygen saturation.

Now to the Digidoc app used by Samsung. This is based on using white light (from the flash), which is of course a mixture of all visible lights. The camera measures the residual light coming through the finger. Instead of using the ratio of ratios method, the app relies on a "neural network" created from 38 normal volunteers, who were asked to breath room air for 30 minutes and then advised to hold their breath for as long as they could. Reference values were thus obtained for absorbance of white light for all levels of oxygen saturation between 85% and 100%. Saturation of oxygen rarely falls below 85% with breath holding.

It is important to state that the app is not accurate below 85% oxygen saturation due to the above reason. It should therefore not be used for detecting hypoxia, as in subjects with COVID-19. The purpose of the app is to establish baselines for a given person pursuing exercise and sports and compare their oxygen sats before, during and after exercise. Unfortunately, it is widely used by people with lung disease such as COPD, which it is not intended for.

There are 2 separate studies, one in adults and one in children, which compared the Samsung oximetry app to a standard pulse oximeter. The first study also compared the readings to arterial blood gas measurements. Both studies found that the app is reliable and accurate. The app also measures pulse rate accurately. However, using a probe that plugs into the smartphone (available to purchase) increases accuracy slightly.

IMO, this is an useful app for daily exercise and for sportsmen, just like the estimated VO2max on Garmin watches. However, it was never intended for medical usage.

References:
1. https://www.howequipmentworks.com/pulse_oximeter/#:~:text=The%20pulse%20oximeter%20works%20out,of%20infrared%20light%20absorbed%20changes.
2. https://s2.smu.edu/~eclarson/pubs/2018pulseox.pdf
3. https://pubmed.ncbi.nlm.nih.gov/29215972/
4. https://pubmed.ncbi.nlm.nih.gov/30904343/

What's The Diagnosis?

52-year-old man with one acute and one "chronic" finding. Leads V1-V6 are actually leads V1R through V6R (right chest leads). What's going on?

From Wave-Maven (ecg.bidmc.harvard.edu)

South Asian Health in Perspective

Why do highly educated, prosperous middle aged Indian men or women need to worry about health? They now have access to some of the best hospitals, highly qualified physicians, and can literally buy healthcare like a commodity. Surely, there is no cause for concern?

Sadly, there is. Many of the conditions I am going to discuss here are not diseases of poverty, but maladies associated with plenty.

In the two decades between 1990 and 2010, the proportion of deaths worldwide from non-communicable diseases (NCD) rose from 57% to 65%. Fully 80% of these occur in countries like India. And 90% of those deaths affect people below the age of 60.

The 4 biggest killers among NCD are CV disease, Cancer, COPD and Diabetes. Three of those are particularly relevant for Indians.

Of these, the one that towers above the rest is cardiovascular disease- read as a higher risk of MI and stroke. It's sobering to realise that the Indian subcontinent accounts for 25% of the world population but 60% of patients with heart disease.

A recent study (2010-2014), charmingly called the "MASALA study" (Mediators of Atherosclerosis in South Asians Living in America), showed that this cohort has a 4-fold higher risk of ASCVD than the general population. Furthermore, they develop this a decade earlier than others- often even before they have reached their 50s.

But it doesn't end there. They are more likely to require CABG to treat their IHD, and are likely to have poorer outcomes at CABG.

So what's driving this? The fact that Indians have one of the highest rates of Type 2 Diabetes Mellitus (T2DM) does not help. Worldwide, 1 in 11 people suffer from T2DM, i.e. 9%. The Indian government's own figures puts the prevalence at 11.8%. However, the prevalence is almost double in urban areas vis-a-vis rural settings, so that figure is an underestimate for most Indians reading this article.

The worrying thing is that many Indians who develop T2DM or IHD would not be classified as overweight or obese by international consensus definitions. Thus internationally, overweight is BMI>25, while obesity is defined as BMI>30. It has now been suggested by the AHA that we lower this to 23 and 27.5 for people of South Asian origin. Why is that?

The problem appears to be with the distribution of body fat in Indians. We tend to have less of our body weight contributed by muscle and more by fat. However, this fat is distributed around our abdominal organs- so called visceral fat and around our heart, where they contribute to an inflammatory phenotype in the blood vessels. Furthermore, Indians have almost half the brown fat- so called brown adipose tissue, of Caucasians, which is thermogenic and helpful in burning off the calories. Our basal metabolic rate is lower than other ethnicities. It appears that we accumulate fat- in the wrong places- literally, and struggle to burn it off. Thus T2DM and IHD both occur at lower BMI in Indians.

South Asians also have a higher prevalence of hypertension (not more than Blacks), high triglycerides, lower HDL, more LDL cholesterol, and a higher total cholesterol to HDL ratios, again, all at a lower body weight than other groups.

Diet clearly plays a part. Thus the reliance on unrefined bread- naan, etc, high usage of trans fatty acids through Vanaspati, and sources of saturated fats such as ghee and butter don't help. Adoption of Western diets both in India and in those living abroad, with a preference for high fat dairy, pizzas, potatoes and red meat makes for the worst of both worlds. Those who do best have a bicultural diet, adopting the best from both systems, combining a predominantly vegetarian diet with a high consumption of green leafy salads, whole grains, fruits, nuts and seeds, and chicken and fish.

Vegetarianism itself however does not help as this is often accompanied by fried snacks, sweetened beverages, and high fat dairy.

High intensity statins should be prescribed for secondary prevention. Those with T2DM, aged 40-75 and LDL>70 mg/dl should have moderate intensity statins. With other risk factors in addition to T2DM, such as family history of ASCVD, LDL>160, metabolic syndrome, CKD, pre-eclampsia, RA or HIV and South Asian origin, high intensity statins should be used. This means that all Indians aged 40-75 who have T2DM should take high intensity statins.

Primary prevention is based on the estimation of 10-year ASCVD risk through an online calculator, with risk stratified into 3 categories. (<7.5%, 7.5-20%, and >20%). CT calcium scores may be used where doubt exists, such as the intermediate group.

Indians have an unique vulnerability to haemorrhagic stroke. This may be associated with a higher prevalence of hypertension and high salt intake, but there are other, yet undefined factors. Unfortunately, this means that people who need anticoagulation, specifically those with AF and CHADS2 or CHADSVasc score of >1, are often not treated with anticoagulants, despite the availability of DOACs. Instead, aspirin is overprescribed for this group.

Undiagnosed CKD, related to ASCVD, or indeed as a contributor to ASCVD (it cuts both ways) is a further problem. Indians are vulnerable to heat stress nephropathy, found in hot dry areas. This was first described in central America, and therefore began under the rubric of Meso-American nephropathy. It then appeared in Sri Lanka and in Andhra Pradesh, particularly in the Nellore district. It tends to predominate in rural areas and is therefore more common in farmers, with >60% prevalence in villages. It's thought to be related to excessive sweating, coupled with inadequate hydration, with contributions from rhabdomyolysis. The temperature in India has been increasing at the rate of 0.8 degrees Celsius annually in many areas, and it is not unusual to find a wet bulb temp of>35 degree Celsius a few days of the year, a threshold that equates with intolerable heat.

Next to the heart, perhaps the most threatened organ in Indians is the liver. There are 4 prime insults that leads to this risk- alcohol, obesity, Hepatitis B and Hepatitis C.

Alcohol use is common in India across all sections of the society, and is far more common among men. Unfortunately, the Asian liver is also more susceptible to the ravages of alcohol, with increased production of acetaldehyde. Thus, studies in the UK show that Asians, particularly Sikh men, have a higher risk of cirrhosis, compared with White men drinking equivalent amounts of alcohol. Asian women drink much less often, but are exquisitely vulnerable to alcohol induced liver disease.

The chronic viral hepatitides B & C are silent killers in India, as they are in the rest of the world. It is estimated that India has 57 million cases of Chronic Hepatitis B, more than a fifth of the worldwide burden of 257 million cases. The prevalence of Hepatitis C carrier status in the Indian populace is 1-2%, equating with a total case load of 13-26 million cases, out of a worldwide denominator of 140 million cases. As fully 70% of Hepatitis C cases lead to carrier status (55-85%), most subjects are unaware that they have Hepatitis C. The commonest genotype by far in India is Genotype 3, unfortunately also the strain that leads most commonly to liver scarring and hepatocellular cancer. Unlike in the West, chronic Hep C in india is mostly acquired by vertical transmission rather than IVDU or in MSM.

Those born between 1945 and 1965 have a higher risk of chronic Hepatitis C due to a birth cohort effect.

NAFLD, is as expected, common in India, given the high prevalence of correlates such as T2DM, hypertrigyceridaemia, and high BMI. There is very little awareness of this condition and it is particularly damaging in conjunction with Chronic Hepatitis or alcohol.

It would only be appropriate to end with a word on the health of Indian women. They suffer from the lack of a national breast screening programme (contrast that to mammography every 3 years from age 50 in the UK), cervical screening programme (cervical smears from age 25 in the UK), and a general lack of awareness of post menopausal osteoporosis and the role of HRT. There is also very little awareness of the dangers of ovarian cancer and its late presentation.

Fully 800 million Indians are classed as anaemic. Fifty two percent of non pregnant women of reproductive age are anaemic. Iron deficiency anaemia (IDA) is the major aetiology, but other contributors such as poor diet, parasitic infections and haemoglobinopathies also apply. The intake of iron in diet is very low. IDA in turn leads to a higher risk of preterm labour, low birth weight and high infant mortality rate. Infants are at risk of developing IDA after 4 months of age.

However, apart from well known side effects such as fatigue, IDA has a sinister side to it. The microspherocytes that circulate in the blood of such patients are much less deformable, have higher viscosity, and are more likely to clog up capillaries. IDA also increases factor VIII levels & favours platelet aggregation, increasing the risk of both arterial and venous clots.

One large population based study involving over 200,000 patients showed that subjects with IDA were almost 50% more likely to suffer from ischaemic stroke (http://www.ncbi.nlm.nih.gov/pubmed/24349404). There is a well established association between IDA and cerebral venous thrombosis. The pro-coagulant effect of IDA is magnified in subjects who are otherwise predisposed to thrombotic events, such as Congenital Cyanotic Heart Disease and Hereditary Haemorrhagic Telangiectasia. In the former condition, severe polycythaemia is a physiological response to chronic deoxygentation, with haemoglobin levels often above 200g/l. Despite this, venesection is avoided unless symptoms of hyperviscosity such as headache, myalgia, or blurred vision are intolerable and the haematocrit is more than 65%, as the risk of causing IDA is unacceptable.

IDA is also the commonest secondary cause of Restless Legs Syndrome (RLS). This is a disorder where subjects have an unpleasant feelings in their legs in bed or while they are resting, relieved only by moving their legs. The condition affects sleep and many subjects are chronically sleep deprived. While most cases are primary, a significant proportion are iron deficient and respond to iron repletion.

The association of IDA with Restless Legs Syndrome led people to investigate its role in Parkinson's Disease. Reason? RLS responds well to anti-Parkinsonian drugs such as ropinirole. On such chance observations rests progress in Medicine. Sure enough, researchers established an association- IDA is more common in Parkinson's Disease although the association is nowhere as strong as between IDA and RLS. However, the challenge in Parkinson's Disease (PD) is to recognise it when it is not fully established...or even predict it years before the classical motor symptoms of rigidity, bradykinesia and tremor develop. It is now well recognised by neurologists that a constellation of non-motor symptoms- specifically anosmia, constipation, and in particular a strange phenomenon called Rapid-Eye-Movement Sleep Disorder (REM Sleep Disorder) develop in patients destined to suffer from PD up to 5-10 years before the motor symptoms set in. The last of these, REM Sleep Disorder is a fascinating example of corruption of normal physiological processes and deserves an explanation.

Normally, during REM sleep, which principally occurs in the second half of the night, dreams occur frequently, accompanied by a physiological paralysis of skeletal muscles. In REM sleep disorder, the paralysis of skeletal muscles are lost, so that the subject "acts out" his dreams (the disorder is much more common among men) with motor movements such as flailing arms and legs. Such movements can often be quite violent, and in some cases spousal injury has occurred. It has been estimated that REM sleep disorder affects up to half of all patients with PD. It is even more common in other α-synucleopathies such as Multiple System Atrophy and Lewy Body Dementia, affecting around 80% of patients. Thus, strictly speaking, non-motor symptoms are not unusual in PD and Parkinsons plus disorders at all. They are however less appreciated.

COVID-19- Lessons From Four Hundred Years Ago

If you are a newly infected COVID-19 patient (or any RNA virus, for that matter), you are more likely to do worse if you contracted the virus from a close relative than if you picked up the infection from an unrelated person.

So what support is there for this somewhat outlandish theory?

You will have to look back a long time- around 400 years, to find a plausible explanation for this. In the 16th century, settlers from Europe increasingly explored the Old World (a term that refers to the Americas and Australia, but mainly the former). Over the next century or so, denizens of the Old World, mainly American Indians living in South and Central America, died off. Approximately 56 million people had died by 1650. The population shrunk by 90%. Even as late as the the 1960s and 70s, the population in some Amazonion basins had a death rate of around 75%.

So what killed them?

A key piece of evidence comes from the work of Garenne and Aaby in 1990, in which they found that a child contracting measles from a family member faced twice the risk of death than a child picking up the infection from an unrelated passerby. While it is tempting to simply attribute this to the dosage of the virus being higher in the instances where it was picked up from a family member, studies with an attenuated measles virus have disproven this by showing that the dosage of this virus while infecting a new subject has no bearing on its outcome.

In a fascinating paper published in Science in 1992, Francis L Black, epidemiologist at New Haven, CT, contended that the American Indians were mainly killed off by RNA viruses. These viruses have poor proofreading in their RNA polymerase, which leads to numerous mutations during new viral RNA production, as discussed previously in this forum. They start mutating even while the infection is progressing within an individual host. The host counters this by presenting the neo-antigens in the grooves of MHC Class I molecules on antigen-presenting-cells (APCs) to CD8 T-lymphocytes. Cellular immunity results, eventually clearing the virus. When the virus infects a new host, the process starts all over again.

However, the virus adapts during this process as well. Like all RNA viruses, it has clever ways of stymieing the immune system by hiding antigenic epitopes inside the cell, rather than on the surface, preventing antigen presentation by APCs or mimicking host antigens. When the virus passes from one host to another, if the hosts are related and therefore share MHC phenotypes, the virus, being pre-adapted, finds it easier to multiply, as it has already found a way of bypassing the adaptive immune system. It does not have this advantage if it attacks a person unrelated to the original host.

And here, the genetic homogeneity of those original American Indians became their downfall. Just to illustrate, There are 3 classes of Class I HLA antigens- A, B, and C. Most epidemiologic data is available on the first two. Thus, serological studies have identified 40 A & B alleles in 1342 sub-Saharan Africans, 37 in 1069 Europeans, 34 in 4061 East Asians, but only 10 among 1944 South Americans, 14 in 12,243 Polynesians, and 10 in 5499 Papua New Guineans. All A & B sequences in the New World also occur in the Old World.

The more alleles in a population, the less is the frequency of an individual allele, and the less likely it is that the virus will find it in its next host. Thus if the allele frequency is q, there is a q^2 chance of finding it in the next host. Black worked out that there was a 32% chance that a virus passing between 2 South Americans would not find a new MHC phenotype at either the A or B locus of the new host, but only a 0.5% chance when it passed between 2 Africans.

Thus, the less the polymorphism of MHC alleles in the host group, the more dangerous the virus is, as evidenced by the demise of those Old World denizens, who lived in a cloistered community with no intermarriage.

The same principles apply to this pandemic, albeit in a different context. COVID-19 has a very high infectivity rate, R being >3. If one happens to pick it up from a close relative, one will fare worse due to the shared MHC alleles, and thus viral pre-adaptation, than if the infection came from an unrelated source.

A 400 year old tragedy may thus have lessons for the current pandemic.

References
1. Black F. why Did They Die. Science 1992;258:1739-40.
2. Garenne M and Aaby P, ibid 1990;161:1088.

Using Genomic Sequencing to Contain COVID-19

If you haven't noticed already, the game has moved on, subtly but surely. Countries are moving on from lock downs, even as cases are at best stable or slowly declining (UK, USA, Italy, Spain), or even surging in some (Brazil, India, Russia). There is now a tacit acceptance that the price of continued and total lock down is too steep, and that nations may have to accept some new cases, in order that society and economy as a whole, can reopen.

It is not a question of if but when. Cases will surge in many places. The pandemic hasn't peaked in the last 3 nations, and is hardly under control in the first group. Some countries have rolled out a "track and trace" mobile app, based on using bluetooth signal to inform if the phone in question has been in the vicinity of one owned by a COVID sufferer. For this, you have to download a track and trace app. If your phone data shows you are at risk, a "track and tracer" (25000 strong in the UK) will contact you, and ask you to self-isolate for 14 days. If you develop the disease, you self isolate for 7 days.

However, there is another way- something that makes the track and trace much, much more effective. This is the power of genomic screening of the COVID19 strains. RNA viruses undergo many more mutations than DNA viruses. One reason for this is the spontaneous deamination of cytosine to uracil. When this happens in DNA virus, this is quickly detected by the virus, and the base is corrected back to cytosine, as uracil is not normally present in DNA. However, this cannot happen in RNA viruses, as uracil is not a "foreign base", and will therefore not be "corrected". C to U mutations will therefore accumulate.

In fact, RNA viruses have over 10 times the rate of mutations as DNA viruses. This does not increase or decrease the pathogenicity of the virus appreciably, but it does mean that within a community or a nation, there might be several "strains" (with differing genome sequences) of the virus in circulation. As this usually affects only a single nucleotide, rather than blocks, it is called "intra-host single nucleotide variation", or simply iSNV. And this presents an opportunity for those seeking to track the spread of the virus.

Consider this. With a limited number of cases, you have the power to sequence the genome of the virus in every documented case within a matter of hours. Each cluster of cases will have a "signature" viral genome, because of the fact that the index case will have a viral strain with its own unique mutations. Thus, once the pandemic is stable and reasonably contained, scientists have within their power to look at the viral genome of a new case, and from a database of existing patients, pinpoint exactly from whom the infection was acquired. Thus, self isolation, instead of being a non-selective and disruptive tool, can be applied selectively and in a limited fashion, to maximum effect. The rest of the population can get on with their lives.

Some examples of common viral mutations might make this easier to understand. During the Zaire Ebola epidemic, it became clear that within the incorporated viral genome in host DNA, a disproportionately high number of mutations were thymine to cytosine (T>C). (Thymine does not appear in RNA, so this is referring to the host DNA that has incorporated the complementary sequence of the viral RNA). It turns out that the preponderance of T>C in virus infected cells is due to the action of an interferon inducible enzyme called "Adenosine deaminase acting in RNA 1", or simply ADAR1. (Interferons, as you know, are produced by the host in response to viral infections).

ADAR1 deaminates adenosine to inosine. Inosine is not a natural nucleotide, and is read as guanosine by the cell. Since the complementary base of guanine is cytosine, the thymine bases complementary to adenine are "corrected" to cytosine by single nucleotide excision and repair. Hence T>C.

This is not a pipe dream. Scientists in NZ, Australia and UK have an extensive database of viral genomes circulating in their respective nations. While the database is virtually 100% complete in NZ and Aus, the UK has data on 20% of viral genomes in circulation, given the very large number of cases. But they are getting there.

This, IMO, presents the only realistic way of opening up the society while continuing to promptly identify and isolate cases and their contacts.

Are Camelids the Key to Beating COVID-19?

The normal antibody (immunoglobulin) has two light chains and two heavy chains. Each chain, be it light or heavy, has a variable and a constant fragment. The variable fragment binds to the putative antigen, while the constant fragment carries the Fc receptor that lets cells like NK cells and neutrophils bind to the antibody. The constant portion also binds and activates complement through the classical pathway.

in 1984, Raymond Hamers at the VUB university in Brussels, while analysing the blood of dromedary camels for antibody response to a Trypanosomal species (the dromedary camel is the Arabian camel, with a single hump, as opposed to the double humped Bactrian camel, found in the plains of Central Asia), found to his surprise that the camel antibodies did not look like the human counterparts at all. They lacked the light chain altogether, and contained only the heavy chain, comprised of the variable and heavy fragments. Quite appropriately, these antibodies were named "camelids".

The 1990s saw the establishment of phage display libraries, which allow the manufacture of virtually any antibody in bacteria, by inserting the relevant sequence in bacteriophages, which then infect the bacteria, and uses the bacterial enzymes to make the protein whose sequence has been inserted into the phage. This technique is responsible for producing most monoclonal antibodies these days, having moved on from the days when antibody producing B-cells were immortalised by fusing them with rat myeloma plasmablasts, a technique described as hybridoma.

It is now possible to produce through such phage display techniques, not just whole immunoglobulin molecules, but parts thereof, such as a the Fab fragment (commercially marketed as Certolizumab), a single chain variable fragment, the camelid (commercial application Caplacizumab, used to treat acquired thrombotic thromobocytopenic purpura), or an isolated variable heavy chain fragment, called VHH. Please see Diagram.

It is the VHH, or the variable heavy chain single domain antibody that now provides promise for the treatment of COVID-19. While vaccines can take years, and canonical (standard) monoclonal antibodies around 6 months to prepare, VHH can be prepared very quickly- within weeks, and therefore are ideally suited for dealing with a pandemic. Please see the linked paper in Cell below:

https://www.cell.com/cell-host-microbe/fulltext/S1931-3128(20)30250-X

VHH has several advantages over other techniques. It ways around 15 kDa, around a tenth of the full immunoglobulin molecule. It can reach antigenic epitopes that the full immunoglobulin molecule cannot reach, such as hidden epitopes, a fact that is relevant with COVID-19. And it does not have the foreign antigenicity that full camelids have, thus reducing the risk that they would be rendered ineffective by the human immune system.

Is Re-infection By COVID-19 Possible?

Once you have had COVID19, can you be reinfected with the same virus?

This article in JAMA may provide some reassurance, although it's understandably based on very little data.

https://jamanetwork.com/journals/jama/fullarticle/2766097

If you wanted a summary, it's halfway down the article, in this line here:

To date, no human reinfections with SARS-CoV-2 have been confirmed.

However, as with everything COVID, answers may not be forthcoming for a very long time. Hence, you look at the literature to see if there are predictors of long term immunity after infections, and a couple of facts begin to emerge.

In general viruses (excluding Influenza, which is prone to antigenic shift and drift) tend to cause long lasting immunity- the half life for antibody levels is between 50-200 years, from a 26 year long study of subjects following infections by 6 viruses- Vaccinia (the virus used for small pox vaccine), measles, mumps, rubella, Varicella zoster (chicken pox), and EBV. This may explain why there has never been a sustained second epidemic by a non-influenza virus, i.e. SARS, MERS, Marburg, Zika, Lassa Fever etc (don't misinterpret this as meaning that these viruses cannot infect naive subjects).

https://www.nejm.org/doi/full/10.1056/NEJMoa066092

The same does not apply to antibody responses to protein antigens, specifically for bacteria. For example, the half life for antibodies to Tetanus and Diphtheria is only 11 years-a fifth of that to the virus with shortest lived immunity.

Thirdly, women tend to have longer lasting protection than men.

https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-018-0568-8

Fourth, if you have low antibody levels to start with, for example in a condition called Common Variable Immunodeficiency, a predominantly IgM response to the infective agent would suggest that the response will not be long lived. This may sound strange, but is intuitive, as IgM is the first antibody isotype produced for any sort of humoral (antibody mediated) response. It is then gradually followed by an IgG response. If IgG (particularly IgG1- there are 4 subclasses of IgG-1, 2, 3 and 4) levels do not increase appreciably, the response will not be long lived. Therefore in general, a high IgM level and a low IgG level- particularly that for the IgG1 component- is a worrisome feature.

https://www.jacionline.org/article/S0091-6749(18)30560-8/fulltext

And finally, some people use memory B cells as a surrogate marker for long lasting immunity. This is not a correct assumption. Long lasting humoral immunity can be memory B cell dependent and memory B cell independent, and each exists independently of the other.

If you wanted a straight yes or no answer to the question as to whether infection with COVID-19 is likely to lead to long lasting immunity- I would say, based on the available evidence, the answer is "Yes".