'Unchanged Melody (2)' ... further reflections on the 2015 Report
When it comes to measuring the progress of a disease three different methods of measurement are normally used: incidence, prevalence and mortality. In the case of TB these are invariably pegged against a consistent number of population – i.e. ‘per 100,000’ – so when we encounter such estimates we can understand them in terms of ‘how many per 100,000’.
Mortality rates are fairly obvious: it suggests how many individuals within an average 100,000 individuals might have been estimated to die of TB in the year in question.
Incidence and Prevalence, however, are slightly trickier.
Incidence refers to the number of individuals who develop active TB during a particular time period (in the case of TB reports this period is normally a year) and once again it’s represented in terms of an average 100,000 population. More often than not this measurement is used to suggest the likelihood that an individual in a population might be affected by the condition, but with TB it’s not quite that simple because we have two stages of disease – the latent phase and the re-activated one. So with TB it’s a little different, though incidence does (as we will shortly see) at least offer a useful idea of the rate of re-activation from earlier infections which were latent.
Prevalence refers to the total number of individuals in a population who have an active disease or health condition at any moment of time. With TB once again it’s represented in relation to an average 100,000 population. Sometimes you’ll see prevalence as being estimated against a period of time, and sometimes as being estimated against a specific point of time (referred either as ‘period-prevalence’ or ‘point-prevalence’). It’s the latter which is the case with TB numbers in WHO Reports. Essentially prevalence offers us more an idea of proportion (of how much of the population is has active disease) at a specific point of time, rather than a rate of infection or re-activation (which is what incidence measures). It’s normally calculated by multiplying the estimated incidence rate and the estimated average duration of the disease.
TB is a chronic disease that can last for years and years (the WHO reckons the average duration of untreated TB to be 3 years), but it can also take between 6 and 24 months to complete a course of treatment, so these two measures (incidence and prevalence) offer very different windows in terms of what’s happening in any community epidemic. Put simply, if we have a country population with an estimated incident rate of 100/100,000 at the same time as we have a global average was 25/100,000 we can say that the rate of re-activating TB is four times the global average; but at the same time we can also reasonably assume that we simultaneously have a local pool of TB-infected people without re-activated disease which may well also be four times higher than the global average (because this is where the new incident cases must be coming from).
Furthermore if there was no TB treatment available and we accept the WHO estimate that the average duration of untreated TB is 3 years, we would logically reckon that prevalence was 100x3 or 300/100,000 (effectively three times the incident rate). The ratio between prevalence and incidence would then be 3:1. But in some cases (in children for instance) TB can kill literally in days, so things are far from consistently simple. And with the intervention of effective drugs this ratio drops anyway substantially: instead of 3:1, globally it looks like it’s currently down to roughly 4:3.
You may have already realised that TB incidence tends to offer more an idea of comparison. Countries with higher incidence rates have more entrenched epidemics than those with lower ones. It also offers an idea of trend, however: if the incidence rate is rising it suggests that, for one reason or another, more individuals are developing active potentially infectious tuberculosis; if the incident rate is falling it suggests the opposite (and this could well be because of successful efforts at TB control). Prevalence, on the other hand, tends to offer us not just an idea of disease burden, but also of attributable community risk: if the prevalent proportion is high it suggests that the risk for an individual member of that community of being infected by TB is also high; if it’s low it similarly suggest that the individual risk of infection is low.
Unfortunately, TB isn’t a simple disease so these measurements are far from straightforward in application because many factors may be at play in a national epidemic. First of all, (thankfully) not all re-activated TB is infectious. Pulmonary TB (the most common type) is generally accepted to be the most infectious; extra-pulmonary TB (TB which has infected another organ and constitutes globally perhaps 10%-20% of the pandemic) generally isn’t; bovine TB (extra-pulmonary TB contracted from drinking unpasteurized milk or poorly cooked meat from an infected animal) generally isn’t infectious either. But even amongst pulmonary TB cases infectiousness is known to be highly variable. Generally ‘sputum-positive’ cases (those who are diagnosed by sputum microscopy when visible TB mycobacteria can be seen in a sample of their sputum by microscope) are predictably more infectious than ‘sputum-negative’ cases (those diagnosed either simply by symptoms or by X-ray because there are no signs of TB mycobacteria on their microscope slides) for instance. And such sputum-negative cases comprise about 40% of the estimated pandemic.
But to make things more complicated still, a TB infection doesn’t remotely mean automatic illness anyway. This is because not all of those infected by the disease go on to develop active (potentially infectious and lethal) TB – in fact normally only a small percentage do, between 5% and 10% of infected latent cases – and this re-activation can often take years to happen. For the vast majority the disease remains contained by the first response of the infected person’s immune system – in which case it’s called a ‘latent’ infection without illness or symptoms but with an on-going possibility of breaking out at some later date and re-activating into full-blown TB. If one happens to be HIV-positive as well, however, the odds are very heavily stacked against the disease remaining in this latent condition for very long – in fact the current guesstimate is that it becomes a frightening 26 times more likely to re-activate. If you think about this for a moment it makes this possibility much more like a stone certainty.
So far so good, if that’s the only way to describe such a depressing summary.
The missing measurement for MDR disease
Strangely no estimated prevalence rates for either MDR- or XDR-TB yet appear in the annual Reports. Generally these Reports offer absolute estimated numbers of MDR cases calculated directly from percentages of the wider pandemic which have been estimated themselves from prior surveys of drug-resistance in the countries concerned. So if there are estimated to be 10,000 new cases of TB in a country in the year in question, and 5% of its TB disease is reckoned to be MDR, then 500 new MDR cases can be assumed to be arising. Sometimes these estimates are the results of national surveys, and sometimes they are extrapolated from sub-national ones. And sometimes these estimated percentages are from surveys that are a decade or more old.
One general problem with these incidence/prevalence indicators is that their value can fluctuate when the base numbers are low, and this may possibly be why they’ve been avoided with MDR-TB so far. This concern is only really relevant, however, if there are any inherent risks of ‘false positive’ diagnoses – in other words if a medical test for the disease in question is known to produce a proportion of false positive diagnoses because only a few can then really throw the numbers. With diagnostic tests for TB, however, this isn’t the case - actually it’s the opposite. TB tests generally are inherently unreliable and are far more likely to produce false negative results (hence the necessary categorisation of ‘sputum-negative’ disease for when the disease is clinically diagnosed despite the absence of any confirmation by bacteriological tests). All widely available approved tests for TB is statistically likely to provide some false negative diagnoses, so it really doesn’t seem that this particular caution should be being applied with this disease. So we’re really not sure why prevalence of MDR-TB isn’t being assessed.
The implications with MDR- and XDR-TB
Meanwhile, the implications with MDR- and XDR-TB are interesting. As we’ve already seen in the case of the wider pandemic, the WHO hangs its hat on a prevalent-to-incident ratio of around 4:3. This is hardly controversial but it’s important to recognise that this only applies to drug-susceptible disease. Given the far lower detection rates with MDR-TB (at best a rate of 25% as opposed to 63% for DS-TB), along with the lower treatment success rates (50% as opposed to 86%) we think that it can be anticipated with some certainty that the existing ratio of prevalent to incident cases with MDR-TB must be much higher than the blanket 4:3. In fact we think that it could quite easily be 2:1. And if this is the case there could an awful lot more infectious prevalent MDR-TB cases out there than anyone is caring to discuss - possibly at least twice what anyone is reporting, and at least twice what anyone seems ready or willing to plan for.
We’re prepared to take a rough punt in this blog at what we may be looking at. We fully accept that we may end up some way off the mark with our numbers, not least because we don’t have access to the necessary data, but we think that this is a worthwhile exercise at this point of time given what’s at stake.
We can make a start by looking at what the most recent Report had to say about treatment outcomes in MDR cases (from its 2012 cohort, their most recent data). Here’s what it had to say on the matter: 50% of MDR cases who started treatment successfully completed it; meanwhile of the other 50%, 16% were known to have died, 16% were lost to follow up, 10% were reported as having failed their treatment, and for the remaining 8% there was no outcome information available.
From an epidemiological perspective we can actually be rather relieved that those 16% died, because at least we can then say that they no longer pose an infectious threat to anyone. We can’t, unfortunately assume anything similar about the other 34% - but let’s be epidemiologically hopeful and assume that the 8% who had no available outcome information shared a similar fate (or assume that they spontaneously recovered which does sometimes happen). At least then our rough calculations can’t be accused of being incautious. So this leaves 26% of our original treatment cohort remaining potentially infectious for as long as they keep breathing.
It may be remembered from our earlier blog (‘Unchanged Melody (1)‘) that there were 123,000 MDR cases that were diagnosed and notified in 2014. Of these, however, only 110,000 were enrolled on treatment. So using our percentage we’ve just decided upon of cases who remained infectious, we have 26% of these 110,000 (or 28,600) as the first contribution to a cumulative woodpile of prevalent infectious cases to consider for this year. But we can also immediately add those 13,000 who were diagnosed but never enrolled (123,000 less 110,000). And we can also add those who were estimated to have been new cases of MDR disease but who never got notified at all (480,000 less 123,000 or 357,000).
So here’s the sum of prevalent cases rolling out from the 2014 cohort:
398,600 prevalent potentially infectious MDR cases last year
Let’s just call this 400,000 for simplicity’s sake from here on. The million dollar question now is this: how long might they survive?
Well, let’s check back on the South African study which tracked XDR patients who were discharged from hospital after failing at least two years of all available treatment. These unhappy souls were then seen to survive for an average of 19 months before dying. This same survival period was recorded as being three years in Cambodia in another study. But we also know that these numbers should be treated with huge uncertainty if we are to view them globally, so let’s more conservatively suggest that the average survival rate of these 400,000 infectious cases might be just 12 months (once again because we don’t want to be seen as being incautious in our rough calculations).
On this basis we can anticipate our 400,000 potentially infectious cases from 2014 might just survive on average for the duration of 2015 – and in line with accepted infection-rates for TB (that a single infectious TB case infects 10-15 further cases in a twelve month period), they can thus be anticipated to infect roughly around 5 million people with their MDR strains in this period.
We should add at this point that some disease-duration calculations used by the WHO do seriously challenge this idea if the case concerned is also HIV positive. Without treatment such cases are reckoned by the WHO to last anywhere between a few days to two-and-a-half months, but since even treated HIV cases are only reckoned to have the disease for a maximum of a year when the treatment for MDR lasts for two we’d simply suggest that these numbers may need reviewing. Furthermore, given that South Africa (where those patients were seen to survive for 19 months) is a high incident HIV country, we think this is a reasonable viewpoint.
So this is pretty serious stuff, although we also know that these prevalent cases won’t all be infectious which gives us cause to be careful what conclusions we might draw. It does suggest though (if the proportion of MDR disease is remaining essentially unchanged) that several million could be being added to the existing growing pool of MDR latent infection every year – the size of which we really still have absolutely no idea about at all. And between 5% and 10% of this pool can be expected to develop re-activated infectious MDR-TB in the course of time (though this will be a much higher percentage if there’s much HIV around).
If we’re thinking of 5 million new latent infections each year, however, then happily these numbers seem to roughly tally with the WHO’s estimate of 480,000 new MDR cases each year since 10% of 5 million is 500,000, or almost exactly the same number as the WHO estimate. In other words we can reasonably suggest that the cycle of MDR disease is indeed stable as is being reported by the WHO. But unfortunately the reality doesn’t tally with this reckoning because, whilst our 5 million new infections were calculated from those who were estimated to be new incident cases in 2014 they only relate to the number infected during 2015 (during the year which we are suggesting the 400,000 might survive). As such it needs to be added to by two other numbers:
one is the possible number of people who would have been already infected by all 480,000 of these ‘new’ 2014 cases before they were ever likely to have presented themselves for diagnosis at a TB clinic;
and the other is the number who might have already been infected by them during the year in question (2014).
TB is well recognised to be a chronic disease, and the nature of those it most frequently infects (along with their general limited access to health care) means that the disease is frequently very well advanced before the sufferer presents for diagnosis, often only when he or she is finally really struggling to work. There are plenty of variables inherent in this for sure, but it’s been suggested that in low-income populations it can take as long as three years from onset of symptoms before patients even present themselves to a clinic.
Generally the WHO reckons that the duration of an average TB case is around 3 years from start to finish, either before dying or spontaneously recovering. So let’s stick with this estimate for now, and suggest that this “pre-presenting” period might be on average just a single year (much as we’ve suggested is the average survival time at the other end of the treatment period). In this case all of our 480,000 ‘new’ cases estimated to have appeared in 2014 may well have already latently infected over 5 million others before the calculations we’ve just run above have even come into play. And if so we can effectively quite easily double our 5 million number at a simple stroke, and then quite reasonably suggest that, unless only 5% of such latent infections re-activate, the WHO’s estimate of a stable 480,000 new cases of MDR-TB each year might not be quite such a reasonable number after all.
If the re-activation rate were a blanket 10% (still bearing in mind that the re-activation rate is reckoned to be 26 times higher amongst HIV co-infections this percentage hardly seems unreasonable in a country in southern Africa) then we can anticipate that, with a possible 5 million new DR latencies being spawned before the year in question and a further 5 million being spawned the year thereafter we’re looking at a prevalence of possibly 1 million potentially infectious cases. In other words, the prevalent:incident ratio of MDR could quite reasonably be assumed to be 2:1
But we should also consider how many might be have been infected by them during the year 2014 itself. In other words, if we just look at the last three years, we might even go so far as to suggest the following:
2012 5 million new latent infections
2013 5 million new latent infections
2014 5 million new latent infections
So just in the last three reported years the pool of latent MDR-TB may have increased by 15 million cases just from the 2014 cohort (not even contemplating the inclusion of those infected by the survivors from the 2013 cohort which could be adding a further 5 million). In which case we might even be looking at a prevalence:incidence ratio that’s near to 3:1 which is what we originally reckoned for TB disease when it’s left untreated.
Of course it probably won’t be because we haven’t as yet factored in the fact that probably only half of these surviving MDR cases will be infectious (by having sputum-positive pulmonary TB). We can hope that the other half won’t be, which means that the real ratiomight be nearer 3:2 (at least if we ignore the HIV factor). But even 3:2 is significantly different to 4:3 which is the ratio of ‘all-TB’ which seems to amount to a stable pandemic.
The MDR-TB pandemic surely cannot be as stable as is being claimed, can it?
The risk of re-activation of a latent infection is sometimes calculated as follows: 5% for the first two years, and then 0.1% for each year thereafter (though we stress once again that these sorts of percentages in no way apply to the immune suppressed, and reiterate that this re-activation is reckoned to be a huge 26 times more likely if the patient is co-infected with HIV). But of course there are some other counter-balancing factors to this. One is infection control within the community – and of course this could make a huge difference to keep the re-infection rates down. Another is that these events are taking place in hotspots, so the pool of potential infectees has to become relatively more finite in the course of time as more and more might be infected. In other words it might even be that the rate of onward infection of 10-15 new cases will in the course of time reduce. But this should hardly be any cause for complacency.
But what this all suggests is something simple. It’s also a little terrifying: if there is any prevalent MDR disease in a country that also has high rates of HIV then the possibility of its drug-resistant epidemic rising alarmingly is significantly more than in a country with little HIV. In fact it’s as good as a stone certainty and it should be treated as such.
But let’s remember something else: the WHO believed that the estimates of MDR disease last year were “essentially unchanged from those published in recent global reports” (Global Tuberculosis Report 2015, page 56). Frankly, if any of the above is correct, this assessment does seem unlikely. Any conclusions we might be drawing from this year’s Report can equally be extrapolated backwards for the last few years as well, the only difference being that the numbers put on treatment were less each year making any final extrapolations a little worse than we’ve made them for the most recent 2014 cohort. This simple cumulative effect seems to be something which is currently simply being ignored which has to further suggest that the current estimate of unchanging numbers is mistaken.
But here’s a rather inconvenient question to consider: why is it that the estimated rates of MDR disease in new TB cases in South Africa are being published as being just half of the global average when there’s been an MDR (and XDR) epidemic of sizeable proportions in this country for well over a decade? South Africa is a high incident HIV country with a known MDR and XDR problem and it's also recognised as being the furnace for the African TB epidemic. These are the percentages that appeared in the WHO Report: the global estimated percentage of new cases of TB that were MDR in 2014 was 3.5% whilst the published estimate for South Africa was 1.8%; the global estimated percentage of retreatment cases that were MDR was 20%, whilst for South Africa it was just 6.7%.
Improbable percentages? We certainly think so. But then you may well not trust our calculations either because they’re too rough and ready. Truthfully we can’t be sure of them ourselves. But what we do know with absolute certainty is that most of the measurements of this disease are still far too fraught with uncertainty for anyone to be confident about what’s really going – and we think that these things need to change rather fast because there are an awful lot of lives at stake.