Lyme disease is a bacterial infection carried by ticks. Caught early it is reliably cured by a couple of weeks of ordinary oral antibiotics. That sentence is true and it is also, on its own, misleading — because the disease is hard to diagnose at exactly the moment when diagnosis matters most, because it is spreading, and because a minority of people remain unwell after treatment in ways that remain genuinely unexplained and genuinely contested.
This review tries to hold all of that at once, without dismissing patients and without overclaiming.
Start here: what Lyme disease is
The illness was recognised in the 1970s from a cluster of children with what looked like juvenile arthritis in Connecticut, linked to a preceding expanding skin lesion [1]. Within a few years it had been tied to a tick-borne spirochaete [2], the organism was isolated from patients' blood [3], and the bacterium was formally named Borrelia burgdorferi [4]. The genome was sequenced in 1997 and turned out to be unusual: a small linear chromosome plus a remarkable collection of linear and circular plasmids [5] [6]. What we call Lyme disease in Europe is caused by several related species, distinguished in the early 1990s [7] [8], which is one reason the European and North American clinical pictures differ [9].
The bacterium lives in a cycle between ticks and small mammals — mainly mice — and human beings are an accident, a dead end for the organism [10] [11]. The tick is an Ixodes species: Ixodes scapularis in eastern North America [12], Ixodes ricinus in Europe [13].
The rash matters more than anything else in this review. Erythema migrans is an expanding red patch appearing days to weeks after a bite, sometimes but not always with the clearing centre that gives it the "bull's-eye" description [14]. In an endemic area it is close to diagnostic on sight — and, as the diagnosis section explains, it appears before the blood tests reliably turn positive. Untreated, the infection can disseminate to the joints, the heart and the nervous system, producing the arthritis that first identified it [1], a characteristic triad of neurological manifestations [15], and, in a minority, persistent joint infection detectable in synovial fluid [16].
Lyme disease is the most commonly reported vector-borne disease in the United States: 275,589 cases were reported to the CDC between 2008 and 2015, of which 208,834 were confirmed [17]. It is also geographically focal, concentrated in the Northeast, mid-Atlantic and upper Midwest — and that geography is moving, which the centerpiece takes up.
Pillar 1: measurement and diagnosis
Early disease is a clinical diagnosis, and this is the point most often missed
If you have erythema migrans in an area where Lyme disease occurs, you have Lyme disease, and you should be treated. You do not need a blood test, and a negative one should not stop treatment. The reason is a matter of timing, and it is well quantified.
Serology detects the immune response, not the bacterium, and the immune response takes weeks. In a prospective study of the standard tests, only about one third of 76 patients with erythema migrans had positive IgM or IgG seroreactivity in acute-phase samples; by convalescence three to four weeks later, almost two thirds did [18]. That is the single most important limitation of Lyme testing: in the earliest, most treatable, most clearly diagnosable stage, the test is most likely to be negative.
Two-tier testing, and what it is and is not good for
The standard approach was set at a national conference in 1994 and published the following year: a sensitive first-tier immunoassay, and, if that is positive or equivocal, a more specific Western blot with defined interpretation criteria [19] [20]. The design reflects a real problem — serological tests for Lyme cross-react with other spirochaetal infections [21] — and two-tier testing exists to buy back specificity that a single sensitive test would lose.
Where it works well is later disease. Of 44 patients with neurological, cardiac or joint involvement, all had a positive C6 peptide ELISA and 42 had IgG responses on two-tier testing [18]. Broader diagnostic reviews reach the same conclusion [22] [23].
Two further limitations deserve stating plainly, because they generate a great deal of confusion.
First, antibodies persist. A positive test tells you the immune system has met this organism; it does not tell you the infection is present now. Serology cannot distinguish active from past, treated infection.
Second, testing in the absence of a compatible illness and exposure is a poor idea. Any test applied where the disease is unlikely returns mostly false positives, and CDC surveillance guidance makes the corresponding clinical point from the other direction: suspicion should be based on local experience and a history of potential exposure rather than on surveillance incidence cutoffs, particularly as the disease emerges in new areas [17].
Exposure is part of the diagnosis
Because risk depends on the tick, so does the diagnosis: geography, season and attachment time all inform it. Attachment time is quantified. In a prospective study of tick bites, infection followed 3 of 15 bites with attachment of 72 hours or more (20 percent) versus 1 of 94 shorter attachments (1.1 percent) — an odds ratio of 23.3 [24]. This has a mechanistic basis: the spirochaete must switch its surface proteins during tick feeding before it can establish in a mammal, and that takes time [25].
Centerpiece: a simple simulatable model of the risk season
Almost all human Lyme infection comes from one life stage. Nymphal ticks are tiny, easily missed, and abundant in early summer when people are outdoors; standardised drag sampling across 95 sites east of the 100th meridian established that human risk tracks the density of host-seeking infected nymphs [26]. So the risk season is the nymphal season — and the nymphal season is set by accumulated warmth, not by the calendar.
That is a measured relationship, not an assumption. Across 12 highly endemic states from 1992 to 2007, an earlier beginning to the Lyme disease season was associated with higher cumulative growing degree days through week 20, and the timing of the peak and the duration of the season were also associated with cumulative growing degree days [27].
The model follows directly. Daily mean temperature is taken as an annual sinusoid; growing degree days accumulate as the sum over days of whatever the temperature exceeds a base value; and nymphal activity is a pulse centred on a fixed accumulated total, multiplied by a propensity to quest that ramps in as the day warms above that base. Warming raises the temperature curve, so the accumulated total is reached on an earlier date and the whole window slides forward.
The illustrative parameters — the temperature cycle, the base temperature, the accumulated-warmth target and the pulse width — were chosen only to place the baseline season where drag sampling finds it, which they do: the modelled peak falls on 14 June.
The right-hand panel is the test. Sweeping the warming increment produces an advance in peak nymphal activity, and that advance can be compared against a number the model was never given. A 19-year field study of blacklegged ticks on small-mammal hosts found that warmer years were associated with a nearly three-week advance in nymphal and larval phenology relative to colder years, and projected that warming by the 2050s would advance average nymph activity by 8 to 11 days [28]. The model reproduces that 8-to-11-day band at warming of 1.5 to 2.1 degrees — a plausible increment, which is what makes the agreement worth something rather than being a coincidence of tuning.
The teaching point: human Lyme risk is concentrated in the nymphal-tick window, and warming moves that window earlier rather than simply making it bigger. That has a practical consequence — public-health messaging keyed to a fixed calendar month drifts out of alignment with the actual risk — and it connects to the two other ways climate is changing exposure. Tick ranges are expanding in North America [29] and Europe [30] [31] [32], with other species moving too [33]; and US surveillance shows case counts rising in states neighbouring the historically high-incidence ones [17].
Three honest limits. Warmth is not monotonically favourable: hot, dry weather during the nymphal questing period significantly reduced both questing nymph density and Lyme incidence in long-term endemic areas, so a model with only a warming term will overstate risk at the hot end [34]. Phenology varies geographically for reasons beyond temperature, including diapause and development-rate differences that require site-specific calibration [35], and peak nymphal host-seeking already occurs earlier in southern than northern sites [26]. And the density of infected nymphs depends on the ecology of the host community as well as on climate — host diversity and community composition alter Lyme risk [36] [37].
Pillar 2: treatment and prevention
Early Lyme disease is curable, and that is not a hedge
Oral antibiotics for early Lyme disease work. Doxycycline and cefuroxime axetil were directly compared in early disease and both were effective [38]; amoxicillin and azithromycin were compared in a double-blind trial in erythema migrans [39]; guideline syntheses set out the regimens and the evidence behind them [40] [41]. Even Lyme neuroborreliosis with encephalitic, myelitic or vasculitic features has been treated with oral doxycycline [42], and in late Lyme disease two ceftriaxone regimens have been compared directly [43].
Prophylaxis after a bite is not routine, but the attachment-duration data identify a small, genuinely high-risk subset — long-attached, engorged ticks — in whom it can be considered [24].
Persistent symptoms after treatment: what is actually known
This is the contested area, and it deserves to be set out carefully rather than summarised in a slogan.
The symptoms are real. In the two randomised trials that most directly addressed the question, baseline assessments documented severe impairment of health-related quality of life in patients with persistent musculoskeletal pain, neurocognitive symptoms or dysesthesia after documented, previously treated Lyme disease [44]. Whatever is causing that, the burden is not in doubt, and there are biological correlates under investigation, including elevated interleukin-23 in a subset of patients with post-Lyme symptoms after erythema migrans [45] and anti-neural antibody reactivity in patients with persistent symptoms [46].
Prolonged antibiotics do not help. Those same two trials randomised patients to 30 days of intravenous ceftriaxone followed by 60 days of oral doxycycline, or matching placebos. They were stopped at a planned interim analysis because a significant difference was highly unlikely to emerge. Among seropositive patients on antibiotics, 37 percent improved, 29 percent were unchanged and 34 percent worsened; on placebo, 40, 26 and 34 percent — a p-value of 0.96 [44]. A separate trial of additional antibiotics assessing cognitive function reached a comparable conclusion [47]. This is a well-tested negative result, and it matters because prolonged antimicrobial treatment carries real harm.
The frequency depends heavily on which population you look at. In a prospective cohort of 128 patients with culture-confirmed early Lyme disease followed for a median of 15 years, 14 (10.9 percent) were regarded as having possible post-treatment symptoms, but only 6 (4.7 percent) had them documented at their last visit; 9 of the 14 had a single symptom, and none of the 6 was considered functionally impaired [48]. That is a very different picture from a referral clinic, and both pictures are real — they are describing different groups.
"Chronic Lyme disease" and post-treatment Lyme disease syndrome are not the same thing. The latter term is applied to persistent symptoms following documented, treated infection. The former is frequently applied to persistent pain, fatigue and neurocognitive symptoms in people with no evidence of previous acute Lyme disease at all, and it is often followed by prolonged treatment with multiple antimicrobials [49]. Someone can be genuinely, severely unwell and not have an ongoing borrelial infection; those are separate claims, and conflating them serves nobody. The syndrome after documented infection is now studied in its own right as a model for persistent symptoms following infection generally [50].
What remains unknown is the mechanism. Persistent infection, immune dysregulation, residual antigen and post-infectious injury have all been proposed, and none is established.
Prevention, and a vaccine that existed and was withdrawn
Avoiding bites and removing ticks promptly is the main lever, and the attachment-time data are why prompt removal works [24].
A vaccine did exist. LYMErix used recombinant outer-surface protein A, and the target is unusual and elegant: OspA is expressed by the spirochaete in the tick rather than in the mammal, so antibodies taken up in the blood meal neutralise the bacteria inside the feeding tick — a transmission-blocking vaccine that works before the infection is established [51] [52]. In a double-blind trial of 10,305 adults in endemic areas, efficacy was 68 percent in the first year and 92 percent in the second year among those who received a third dose [53], with a second large trial reported alongside it [54]. The FDA approved it in 1998. Three years later the manufacturer withdrew it voluntarily amid media coverage, fears about side effects and declining sales [55]. The episode is usually taught as a case study in risk communication rather than in immunology, and the OspA target itself was never the problem — it remains the basis on which Lyme vaccination is being attempted again.
Pillar 3: what is unresolved
Diagnostics remain the weakest link. The gap that matters is a test that is positive in the first days of infection and that distinguishes active from past infection — neither of which serology can do [18] [19]. Direct detection has been disappointing: PCR on ticks was an unreliable predictor of human infection [24], although PCR on synovial fluid does detect the organism in Lyme arthritis [16].
The biology of persistence is genuinely open. Borrelia evades immunity by promiscuous recombination at a variable-surface-antigen locus [56], its lipoproteins drive inflammation through Toll-like receptor signalling [57], and combinations of drugs have been shown to eradicate stationary-phase organisms in vitro that single agents do not [58] — a finding about a culture dish, not a demonstration that persistent infection causes human symptoms, and it should not be over-read.
Prevention has room to move. The OspA transmission-blocking approach is validated, both in mechanism [51] [52] and in field efficacy [53] [55]; what it now needs is deployment. Tick-side targets such as the receptor the spirochaete uses to attach in the tick gut are also being explored [59], and there is a broader literature on tick vaccines [60].
And exposure is a moving target. Range expansion and season shift both push in the same direction [29] [28] [32], which means the answer to "is this Lyme?" depends on a geography and a calendar that are themselves changing [17].
Dig deeper in lmmol
Lyme disease's closest relative in this collection is dengue: both are vector-borne, both are expanding as the climate warms, and both are governed by a transmission season that temperature sets rather than the calendar — though dengue's temperature dependence acts on a mosquito's ability to transmit, while Lyme's acts on when a tick is out looking for a host. Malaria shares the same vector-borne arithmetic with a third vector. For the threshold mathematics that decides whether any of these sustains transmission at all, see H5N1 avian influenza and measles. The chronic-symptom controversy here has an instructive parallel in the contested effect sizes of depression, and Lyme arthritis — an inflammatory arthritis triggered by infection — is worth reading against rheumatoid arthritis. Late Lyme carditis and neuroborreliosis connect to heart failure and, for the general problem of attributing persistent cognitive symptoms, Alzheimer's disease. The full collection is at health.